pyqstrat package

Submodules

pyqstrat.pq_utils module

class pyqstrat.pq_utils.Paths(base_path=None)[source]

Bases: object

Conventions for where to read / write data and reports

create()[source]
Return type

None

pyqstrat.pq_utils.bootstrap_ci(a, ci_level=0.95, n=1000, func=<function mean>)[source]

Non parametric bootstrap for confidence intervals :type a: ndarray :param a: The data to bootstrap from :type ci_level: float :param ci_level: The confidence interval level, e.g. 0.95 for 95%. Default 0.95 :type n: int :param n: Number of boostrap iterations. Default 1000 :type func: Callable[[ndarray], ndarray] :param func: The function to use, e.g np.mean or np.median. Default np.mean

Return type

Tuple[float, float]

Returns

A tuple containing the lower and upper ci

>>> np.random.seed(0)
>>> x = np.random.uniform(high=10, size=100000)
>>> assert np.allclose(bootstrap_ci(x), (4.9773159, 5.010328))
pyqstrat.pq_utils.date_2_num(d)[source]
Return type

Union[int, ndarray]

pyqstrat.pq_utils.day_of_week_num(a)[source]

From https://stackoverflow.com/questions/52398383/finding-day-of-the-week-for-a-datetime64 Get day of week for a numpy array of datetimes Monday is 0, Sunday is 6

Parameters

a (Union[datetime64, ndarray]) – numpy datetime64 or array of datetime64

Returns

Monday is 0, Sunday is 6

Return type

int or numpy ndarray of int

>>> day_of_week_num(np.datetime64('2015-01-04'))
6
pyqstrat.pq_utils.day_symbol(day_int)[source]
Return type

Union[str, ndarray]

pyqstrat.pq_utils.get_child_logger(child_name)[source]
Return type

Logger

pyqstrat.pq_utils.get_empty_np_value(np_dtype)[source]

Get empty value for a given numpy datatype >>> a = np.array([‘2018-01-01’, ‘2018-01-03’], dtype = ‘M8[D]’) >>> get_empty_np_value(a.dtype) numpy.datetime64(‘NaT’)

Return type

Any

pyqstrat.pq_utils.get_main_logger()[source]
Return type

Logger

pyqstrat.pq_utils.get_paths(base_path=None)[source]
Return type

Paths

pyqstrat.pq_utils.get_temp_dir()[source]
Return type

str

pyqstrat.pq_utils.has_display()[source]

If we are running in unit test mode or on a server, then don’t try to draw graphs, etc.

Return type

bool

pyqstrat.pq_utils.in_ipython()[source]

Whether we are running in an ipython (or Jupyter) environment

Return type

bool

pyqstrat.pq_utils.infer_compression(input_filename)[source]

Infers compression for a file from its suffix. For example, given “/tmp/hello.gz”, this will return “gzip” >>> infer_compression(“/tmp/hello.gz”) ‘gzip’ >>> infer_compression(“/tmp/abc.txt”) is None True

Return type

Optional[str]

pyqstrat.pq_utils.infer_frequency(timestamps)[source]

Returns most common frequency of date differences as a fraction of days :type timestamps: ndarray :param timestamps: A numpy array of monotonically increasing datetime64

>>> timestamps = np.array(['2018-01-01 11:00:00', '2018-01-01 11:15:00', '2018-01-01 11:30:00', '2018-01-01 11:35:00'], dtype = 'M8[ns]')
>>> print(round(infer_frequency(timestamps), 8))
0.01041667
Return type

float

pyqstrat.pq_utils.is_newer(filename, ref_filename)[source]

whether filename ctime (modfication time) is newer than ref_filename or either file does not exist >>> import time >>> import tempfile >>> temp_dir = tempfile.gettempdir() >>> touch(f’{temp_dir}/x.txt’) >>> time.sleep(0.1) >>> touch(f’{temp_dir}/y.txt’) >>> is_newer(f’{temp_dir}/y.txt’, f’{temp_dir}/x.txt’) True >>> touch(f’{temp_dir}/y.txt’) >>> time.sleep(0.1) >>> touch(f’{temp_dir}/x.txt’) >>> is_newer(f’{temp_dir}/y.txt’, f’{temp_dir}/x.txt’) False

Return type

bool

pyqstrat.pq_utils.linear_interpolate(a1, a2, x1, x2, x)[source]
>>> assert(linear_interpolate(3, 4, 8, 10, 8.9) == 3.45)
>>> assert(linear_interpolate(3, 3, 8, 10, 8.9) == 3)
>>> assert(np.isnan(linear_interpolate(3, 4, 8, 8, 8.9)))
>>> x = linear_interpolate(
...    np.array([3., 3.]),
...    np.array([4., 3.]),
...    np.array([8., 8.]),
...    np.array([10, 8.]),
...    np.array([8.9, 8.]))
>>> assert(np.allclose(x, np.array([3.45, 3.])))
Return type

Union[ndarray, float]

pyqstrat.pq_utils.millis_since_epoch(dt)[source]

Given a python datetime, return number of milliseconds between the unix epoch and the datetime. Returns a float since it can contain fractions of milliseconds as well >>> millis_since_epoch(datetime.datetime(2018, 1, 1)) 1514764800000.0

Return type

float

pyqstrat.pq_utils.monotonically_increasing(array)[source]

Returns True if the array is monotonically_increasing, False otherwise

>>> monotonically_increasing(np.array(['2018-01-02', '2018-01-03'], dtype = 'M8[D]'))
True
>>> monotonically_increasing(np.array(['2018-01-02', '2018-01-02'], dtype = 'M8[D]'))
False
Return type

bool

pyqstrat.pq_utils.nan_to_zero(array)[source]

Converts any nans in a numpy float array to 0

Return type

ndarray

pyqstrat.pq_utils.np_bucket(a, buckets, default_value=0, side='mid')[source]

Given a numpy array and a sorted list of buckets, assign each element to a bucket.

Parameters
  • a (np.ndarray) – The numpy array of values

  • buckets (List[Any]) – (list) List of buckets

  • default_value – Used when we cannot assign an element to any bucket if side is ‘left’ or ‘right’

  • side (str) – If set to mid, we use the midpoint between buckets to assign elements ‘left’, assignment <= element ‘right’, assignment >= element Default: ‘mid’

Return type

ndarray

Returns

np.ndarray of same length as a

>>> a = np.array([1, 5, 18, 3, 6, 10, 4])
>>> buckets = [4, 8, 12]
>>> assert np.alltrue(np_bucket(a, buckets, side='left') == np.array([0,  4, 12,  0,  4,  8,  4]))
>>> assert np.alltrue(np_bucket(a, buckets, default_value=25, side='right') == np.array([4,  8, 25,  4,  8, 12,  4]))
>>> assert(np.alltrue(np_bucket(a, buckets) == np.array([4,  4, 12,  4,  8, 12,  4])))
pyqstrat.pq_utils.np_find_closest(a, v)[source]

From https://stackoverflow.com/questions/8914491/finding-the-nearest-value-and-return-the-index-of-array-in-python Find index of closest value to array v in array a. Returns an array of the same size as v a must be sorted >>> assert(all(np_find_closest(np.array([3, 4, 6]), np.array([4, 2])) == np.array([1, 0])))

Return type

Union[int, ndarray]

pyqstrat.pq_utils.np_get_index(array, value)[source]

Get index of a value in a numpy array. Returns -1 if the value does not exist.

Return type

int

pyqstrat.pq_utils.np_inc_dates(dates, num_days=1)[source]

Increment the given date array so each cell gets the next higher value in this array >>> dates = np.array([‘2021-06-01’, ‘2021-06-01’, ‘2021-08-01’, ‘2021-04-01’], dtype=’M8[D]’) >>> check = np.array([dates[2], dates[2], np.datetime64(‘nat’), dates[0]]) >>> assert np.array_equal(np_inc_dates(dates, 1), … np.array([‘2021-08-01’, ‘2021-08-01’, ‘NaT’, ‘2021-06-01’], dtype=’M8[D]’), equal_nan=True) >>> assert np.array_equal(np_inc_dates(dates, 2), … np.array([‘NaT’, ‘NaT’, ‘NaT’, ‘2021-08-01’], dtype=’M8[D]’), equal_nan=True) >>> assert np.array_equal(np_inc_dates(dates, -1), … np.array([‘2021-04-01’, ‘2021-04-01’, ‘2021-06-01’, ‘NaT’], dtype=’M8[D]’), equal_nan=True) >>> assert np.array_equal(np_inc_dates(dates, -2), … np.array([‘NaT’, ‘NaT’, ‘2021-04-01’, ‘NaT’], dtype=’M8[D]’), equal_nan=True)

Return type

ndarray

pyqstrat.pq_utils.np_parse_array(s, dtype=<class 'float'>)[source]

Create a 1 or 2 d numpy array from a string that looks like: [[2. 5. 3. 0. 0.]

[3. 5. 0. 4. 3.]]

or [2. 5. 3. 0. 8.]

>>> x = np_parse_array('[[2. 5. 3. 0. 0.]\n [3. 5. 0. 4. 3.]]')
>>> assert np.allclose(x, np.array([[2., 5., 3., 0., 0.], [3., 5., 0., 4., 3.]]))
>>> x = np_parse_array('[3 4. 5]')
>>> assert np.allclose(x, np.array([3, 4., 5]))
Return type

ndarray

pyqstrat.pq_utils.np_rolling_window(a, window)[source]

For applying rolling window functions to a numpy array See: https://stackoverflow.com/questions/6811183/rolling-window-for-1d-arrays-in-numpy >>> print(np.std(np_rolling_window(np.array([1, 2, 3, 4]), 2), 1)) [0.5 0.5 0.5]

Return type

ndarray

pyqstrat.pq_utils.np_round(a, clip)[source]

Round all elements in an array to the nearest clip

Parameters
  • a (ndarray) – array with elements to round

  • clip (float) – rounding value

>>> np_round(15.8, 0.25)
15.75
pyqstrat.pq_utils.np_uniques(arrays)[source]

Given a list of numpy arrays that may have different datatype, generate a structured numpy array with sorted, unique elements from that list >>> array1 = np.array([‘2018-01-02’, ‘2018-01-03’, ‘2018-01-02’, ‘2018-01-03’], dtype=’M8[D]’) >>> array2 = np.array([‘P’, ‘P’, ‘P’, ‘C’]) >>> x = np_uniques([array1, array2]) >>> assert len(x) == 3 >>> assert x[0][0] == np.datetime64(‘2018-01-02’) >>> assert x[0][1] == ‘P’

Return type

ndarray

pyqstrat.pq_utils.percentile_of_score(a)[source]

For each element in a, find the percentile of a its in. From stackoverflow.com/a/29989971/5351549 Like scipy.stats.percentileofscore but runs in O(n log(n)) time. >>> a = np.array([4, 3, 1, 2, 4.1]) >>> percentiles = percentile_of_score(a) >>> assert(all(np.isclose(np.array([ 75., 50., 0., 25., 100.]), percentiles)))

Return type

Optional[ndarray]

pyqstrat.pq_utils.remove_dups(lst, key_func=None)[source]

Remove duplicates from a list :type lst: Sequence[Any] :param lst: list to remove duplicates from :type key_func: Optional[Callable[[Any], Any]] :param key_func: A function that takes a list element and converts it to a key for detecting dups

Return type

MutableSequence[Any]

Returns

A list with duplicates removed. This is stable in the sense that original list elements will retain their order

>>> print(remove_dups(['a', 'd', 'a', 'c']))
['a', 'd', 'c']
>>> print(remove_dups(['a', 'd', 'A']))
['a', 'd', 'A']
>>> print(remove_dups(['a', 'd', 'A'], key_func = lambda e: e.upper()))
['a', 'd']
pyqstrat.pq_utils.resample_trade_bars(df, sampling_frequency, resample_funcs=None)[source]

Downsample trade bars using sampling frequency

Parameters
  • df (pd.DataFrame) – Must contain an index of numpy datetime64 type which is monotonically increasing

  • sampling_frequency (str) – See pandas frequency strings

  • str (resample_funcs (dict of) – int): a dictionary of column name -> resampling function for any columns that are custom defined. Default None. If there is no entry for a custom column, defaults to ‘last’ for that column

Returns

Resampled dataframe

Return type

pd.DataFrame

>>> import math
>>> df = pd.DataFrame({'date': np.array(['2018-01-08 15:00:00', '2018-01-09 13:30:00', '2018-01-09 15:00:00', '2018-01-11 15:00:00'], dtype = 'M8[ns]'),
...          'o': np.array([8.9, 9.1, 9.3, 8.6]),
...          'h': np.array([9.0, 9.3, 9.4, 8.7]),
...          'l': np.array([8.8, 9.0, 9.2, 8.4]),
...          'c': np.array([8.95, 9.2, 9.35, 8.5]),
...          'v': np.array([200, 100, 150, 300]),
...          'x': np.array([300, 200, 100, 400])
...         })
>>> df['vwap'] =  0.5 * (df.l + df.h)
>>> df.set_index('date', inplace = True)
>>> df = resample_trade_bars(df, sampling_frequency = 'D', resample_funcs={'x': lambda df,
...   sampling_frequency: df.x.resample(sampling_frequency).agg(np.mean)})
>>> assert(len(df) == 4)
>>> assert(math.isclose(df.vwap.iloc[1], 9.24))
>>> assert(np.isnan(df.vwap.iloc[2]))
>>> assert(math.isclose(df.l[3], 8.4))
pyqstrat.pq_utils.resample_ts(dates, values, sampling_frequency)[source]

Downsample a tuple of datetimes and value arrays using sampling frequency, using the last value if it does not exist at the bin edge. See pandas.Series.resample

Parameters
  • dates (ndarray) – a numpy datetime64 array

  • values (ndarray) – a numpy array

  • sampling_frequency (str) – See pandas frequency strings

Return type

Tuple[ndarray, ndarray]

Returns

Resampled tuple of datetime and value arrays

pyqstrat.pq_utils.resample_vwap(df, sampling_frequency)[source]

Compute weighted average of vwap given higher frequency vwap and volume

Return type

Optional[ndarray]

pyqstrat.pq_utils.series_to_array(series)[source]

Convert a pandas series to a numpy array. If the object is not a pandas Series return it back unchanged

Return type

ndarray

pyqstrat.pq_utils.set_defaults(df_float_sf=9, df_display_max_rows=200, df_display_max_columns=99, np_seterr='raise', mpl_figsize=(8, 6), jupyter_multiple_display=True)[source]

Set some display defaults to make it easier to view dataframes and graphs.

Parameters
  • df_float_sf (int) – Number of significant figures to show in dataframes (default 4). Set to None to use pandas defaults

  • df_display_max_rows (int) – Number of rows to display for pandas dataframes when you print them (default 200). Set to None to use pandas defaults

  • df_display_max_columns (int) – Number of columns to display for pandas dataframes when you print them (default 99). Set to None to use pandas defaults

  • np_seterr (str) – Error mode for numpy warnings. See numpy seterr function for details. Set to None to use numpy defaults

  • mpl_figsize (Tuple[int, int]) – Default figure size to use when displaying matplotlib plots (default 8,6). Set to None to use defaults

  • jupyter_multiple_display – If set, and you have multiple outputs in a Jupyter cell, output will contain all of them. Default True

Return type

None

pyqstrat.pq_utils.set_ipython_defaults(jupyter_multiple_display=True)[source]
Return type

None

pyqstrat.pq_utils.shift_np(array, n, fill_value=None)[source]

Similar to pandas.Series.shift but works on numpy arrays.

Parameters
  • array (ndarray) – The numpy array to shift

  • n (int) – Number of places to shift, can be positive or negative

  • fill_value (Optional[Any]) – After shifting, there will be empty slots left in the array. If set, fill these with fill_value. If fill_value is set to None (default), we will fill these with False for boolean arrays, np.nan for floats

Return type

ndarray

pyqstrat.pq_utils.str2date(s)[source]

Converts a string like “2008-01-15 15:00:00” to a numpy datetime64. If s is not a string, return s back

Return type

Optional[datetime64]

pyqstrat.pq_utils.strtup2date(tup)[source]

Converts a string tuple like (“2008-01-15”, “2009-01-16”) to a numpy datetime64 tuple. If the tuple does not contain strings, return it back unchanged

Return type

Tuple[Optional[datetime64], Optional[datetime64]]

pyqstrat.pq_utils.to_csv(df, file_name, index=False, compress=False, *args, **kwargs)[source]

Creates a temporary file then renames to the permanent file so we don’t have half written files. Also optionally compresses using the xz algorithm

Return type

None

pyqstrat.pq_utils.touch(fname, mode=438, dir_fd=None, **kwargs)[source]

replicate unix touch command, i.e create file if it doesn’t exist, otherwise update timestamp

Return type

None

pyqstrat.pq_utils.zero_to_nan(array)[source]

Converts any zeros in a numpy array to nans

Return type

ndarray

pyqstrat.pq_types module

class pyqstrat.pq_types.Contract[source]

Bases: object

static clear()[source]

When running Python interactively you may create a Contract with a given symbol multiple times because you don’t restart Python therefore global variables are not cleared. This function clears global Contracts

Return type

None

contract_group: ContractGroup
contracts_by_symbol: Mapping[str, Contract]

A contract such as a stock, option or a future that can be traded

static create(symbol, contract_group, expiry=None, multiplier=1.0, properties=None)[source]
Parameters
  • symbol (str) – A unique string reprenting this contract. e.g IBM or ESH9

  • contract_group (ContractGroup) – We sometimes need to group contracts for calculating PNL, for example, you may have a strategy which has 3 legs, a long option, a short option and a future or equity used to hedge delta. In this case, you will be trading different symbols over time as options and futures expire, but you may want to track PNL for each leg using a contract group for each leg. So you could create contract groups ‘Long Option’, ‘Short Option’ and ‘Hedge’ and assign contracts to these.

  • expiry (Union[datetime64, datetime, None]) – In the case of a future or option, the date and time when the contract expires. For equities and other non expiring contracts, set this to None. Default None.

  • multiplier (float) – If the market price convention is per unit, and the unit is not the same as contract size, set the multiplier here. For example, for E-mini contracts, each contract is 50 units and the price is per unit, so multiplier would be 50. Default 1

  • properties (Optional[SimpleNamespace]) – Any data you want to store with this contract. For example, you may want to store option strike. Default None

Return type

Contract

expiry: Optional[datetime64]
multiplier: float
properties: SimpleNamespace
symbol: str
class pyqstrat.pq_types.ContractGroup[source]

Bases: object

A way to group contracts for figuring out which indicators, rules and signals to apply to a contract and for PNL reporting

add_contract(contract)[source]
static clear()[source]

When running Python interactively you may create a ContractGroup with a given name multiple times because you don’t restart Python therefore global variables are not cleared. This function clears global ContractGroups

Return type

None

contracts: MutableSet[Contract]
contracts_by_symbol: Mapping[str, Contract]
static create(name)[source]
Parameters

name (str) – Name of the group

Return type

ContractGroup

get_contract(symbol)[source]
name: str
class pyqstrat.pq_types.LimitOrder(contract, timestamp, qty, limit_price, reason_code='none', properties=None, status='open')[source]

Bases: Order

__init__(contract, timestamp, qty, limit_price, reason_code='none', properties=None, status='open')[source]
Parameters
  • contract (Contract) – The contract this order is for

  • timestamp (datetime64) – Time the order was placed

  • qty (float) – Number of contracts or shares. Use a negative quantity for sell orders

  • limit_price (float) – Limit price (float)

  • reason_code (str) – The reason this order was created. Prefer a predefined constant from the ReasonCode class if it matches your reason for creating this order. Default None

  • properties (Optional[SimpleNamespace]) – Any order specific data we want to store. Default None

  • status (str) – Status of the order, “open”, “filled”, etc. (default “open”)

class pyqstrat.pq_types.MarketOrder(contract, timestamp, qty, reason_code='none', properties=None, status='open')[source]

Bases: Order

__init__(contract, timestamp, qty, reason_code='none', properties=None, status='open')[source]
Parameters
  • contract (Contract) – The contract this order is for

  • timestamp (datetime64) – Time the order was placed

  • qty (float) – Number of contracts or shares. Use a negative quantity for sell orders

  • reason_code (str) – The reason this order was created. Prefer a predefined constant from the ReasonCode class if it matches your reason for creating this order. Default None

  • properties (Optional[SimpleNamespace]) – Any order specific data we want to store. Default None

  • status (str) – Status of the order, “open”, “filled”, etc. Default “open”

class pyqstrat.pq_types.Order[source]

Bases: object

class pyqstrat.pq_types.OrderStatus[source]

Bases: object

Enum for order status

FILLED = 'filled'
OPEN = 'open'
class pyqstrat.pq_types.Price(timestamp, bid, ask, bid_size, ask_size, valid=True, properties=None)[source]

Bases: object

>>> price = Price(datetime.datetime(2020, 1, 1), 15.25, 15.75, 189, 300)
>>> print(price)
15.25@189/15.75@300
>>> price.properties = SimpleNamespace(delta = -0.3)
>>> price.valid = False
>>> print(price)
15.25@189/15.75@300 delta: -0.3 invalid
>>> print(price.mid())
15.5
ask: float
ask_size: int
bid: float
bid_size: int
static invalid()[source]
Return type

Price

mid()[source]
Return type

float

properties: Optional[SimpleNamespace] = None
set_property(name, value)[source]
Return type

None

spread()[source]
Return type

float

timestamp: datetime
valid: bool = True
vw_mid()[source]

Volume weighted mid >>> price = Price(datetime.datetime(2020, 1, 1), 15.25, 15.75, 189, 300) >>> print(f’{price.vw_mid():.4f}’) 15.4433 >>> price.bid_size = 0 >>> price.ask_size = 0 >>> assert math.isnan(price.vw_mid())

Return type

float

class pyqstrat.pq_types.ReasonCode[source]

Bases: object

A class containing constants for predefined order reason codes. Prefer these predefined reason codes if they suit the reason you are creating your order. Otherwise, use your own string.

BACKTEST_END = 'backtest end'
ENTER_LONG = 'enter long'
ENTER_SHORT = 'enter short'
EXIT_LONG = 'exit long'
EXIT_SHORT = 'exit short'
MARKER_PROPERTIES = {'backtest end': {'color': 'green', 'size': 50, 'symbol': '*'}, 'enter long': {'color': 'blue', 'size': 50, 'symbol': 'P'}, 'enter short': {'color': 'red', 'size': 50, 'symbol': 'P'}, 'exit long': {'color': 'blue', 'size': 50, 'symbol': 'X'}, 'exit short': {'color': 'red', 'size': 50, 'symbol': 'X'}, 'none': {'color': 'green', 'size': 50, 'symbol': 'o'}, 'roll future': {'color': 'green', 'size': 50, 'symbol': '>'}}
NONE = 'none'
ROLL_FUTURE = 'roll future'
class pyqstrat.pq_types.RollOrder(contract, timestamp, close_qty, reopen_qty, reason_code='roll future', properties=None, status='open')[source]

Bases: Order

A roll order is used to roll a future from one series to the next. It represents a sell of one future and the buying of another future.

__init__(contract, timestamp, close_qty, reopen_qty, reason_code='roll future', properties=None, status='open')[source]
Parameters
  • contract (Contract) – The contract this order is for

  • timestamp (datetime64) – Time the order was placed

  • close_qty (float) – Quantity of the future you are rolling

  • reopen_qty (float) – Quantity of the future you are rolling to

  • reason_code (str) – The reason this order was created. Prefer a predefined constant from the ReasonCode class if it matches your reason for creating this order. Default None

  • properties (Optional[SimpleNamespace]) – Any order specific data we want to store. Default None

  • status (str) – Status of the order, “open”, “filled”, etc. (default “open”)

class pyqstrat.pq_types.RoundTripTrade(contract, entry_order, exit_order, entry_timestamp, exit_timestamp, qty, entry_price, exit_price, entry_reason, exit_reason, entry_commission, exit_commission, entry_properties, exit_properties, net_pnl)[source]

Bases: object

contract: Contract
entry_commission: float
entry_order: Order
entry_price: float
entry_properties: Optional[SimpleNamespace]
entry_reason: str
entry_timestamp: datetime64
exit_commission: float
exit_order: Order
exit_price: float
exit_properties: Optional[SimpleNamespace]
exit_reason: str
exit_timestamp: datetime64
net_pnl: float
qty: int
class pyqstrat.pq_types.StopLimitOrder(contract, timestamp, qty, trigger_price, limit_price=nan, reason_code='none', properties=None, status='open')[source]

Bases: Order

Used for stop loss or stop limit orders. The order is triggered when price goes above or below trigger price, depending on whether this is a short or long order. Becomes either a market or limit order at that point, depending on whether you set the limit price or not.

__init__(contract, timestamp, qty, trigger_price, limit_price=nan, reason_code='none', properties=None, status='open')[source]
Parameters
  • contract (Contract) – The contract this order is for

  • timestamp (datetime64) – Time the order was placed

  • qty (float) – Number of contracts or shares. Use a negative quantity for sell orders

  • trigger_price (float) – Order becomes a market or limit order if price crosses trigger_price.

  • limit_price (float) – If not set (default), order becomes a market order when price crosses trigger price. Otherwise it becomes a limit order. Default np.nan

  • reason_code (str) – The reason this order was created. Prefer a predefined constant from the ReasonCode class if it matches your reason for creating this order. Default None

  • properties (Optional[SimpleNamespace]) – Any order specific data we want to store. Default None

  • status (str) – Status of the order, “open”, “filled”, etc. (default “open”)

class pyqstrat.pq_types.Trade(contract, order, timestamp, qty, price, fee=0.0, commission=0.0, properties=None)[source]

Bases: object

__init__(contract, order, timestamp, qty, price, fee=0.0, commission=0.0, properties=None)[source]
Parameters
  • contract (Contract) – The contract we traded

  • order (Order) – A reference to the order that created this trade. Default None

  • timestamp (datetime64) – Trade execution datetime

  • qty (float) – Number of contracts or shares filled

  • price (float) – Trade price

  • fee (float) – Fees paid to brokers or others. Default 0

  • commision – Commission paid to brokers or others. Default 0

  • properties (Optional[SimpleNamespace]) – Any data you want to store with this contract. For example, you may want to store bid / ask prices at time of trade. Default None


pyqstrat.pq_io.df_to_hdf5(df, filename, key, dtypes=None, optimize_vlen_str=True)[source]

Write out a pandas dataframe to hdf5 using the np_arrays_to_hdf5 function

Return type

None

pyqstrat.pq_io.hdf5_to_df(filename, key)[source]

Read a pandas dataframe previously written out using df_to_hdf5 or np_arrays_to_hdf5

Return type

DataFrame

pyqstrat.pq_io.hdf5_to_np_arrays(filename, key)[source]

Read a list of numpy arrays previously written out by np_arrays_to_hdf5 :type filename: str :param filename: path of the hdf5 file to read :type key: str :param key: group and or / subgroups to read from. For example, “g1/g2” will read from the subgrp g2 within the grp g1

Return type

List[Tuple[str, ndarray]]

Returns

a list of numpy arrays along with their names

pyqstrat.pq_io.np_arrays_to_hdf5(data, filename, key, dtypes=None, optimize_vlen_str=True, compression_args=None)[source]

Write a list of numpy arrays to hdf5 :type data: List[Tuple[str, ndarray]] :param data: List of numpy one dimensional arrays along with the name of the array :type filename: str :param filename: filename of the hdf5 file :type key: str :param key: group and or / subgroups to write to. For example, “g1/g2” will write to the subgrp g2 within the grp g1 :type dtypes: Optional[Dict[str, str]] :param dtypes: dict used to override datatype for a column. For example, {“col1”: “f4”} will write a 4 byte float array for col1 :type optimize_vlen_str: bool :param optimize_vlen_str: if set, for every variable length string array, i.e dtype = ‘O’, we try to find the maximum string length

and if it is < 100, we write out fixed length strings instead of variable length. This is much faster to read and process

Parameters

compression_args (Optional[Dict[Any, Any]]) – if you want to compress the hdf5 file. You can use the hdf5plugin module and arguments such as hdf5plugin.Blosc()

Return type

None

pyqstrat.pq_io.test_hdf5_to_df()[source]

pyqstrat.holiday_calendars module

class pyqstrat.holiday_calendars.Calendar(holidays)[source]

Bases: object

EUREX = 'eurex'
NYSE = 'nyse'
__init__(holidays)[source]

Do not use this function directly. Use Calendar.get_calendar instead :type holidays: ndarray :param holidays: holidays for this calendar, excluding weekends :type holidays: np.array of datetime64[D]

static add_calendar(exchange_name, holidays)[source]

Add a trading calendar to the class level calendars dict

Parameters
  • exchange_name (str) – Name of the exchange.

  • holidays (ndarray) – holidays for this exchange, excluding weekends

Return type

None

add_trading_days(start, num_days, roll='raise')[source]

Adds trading days to a start date

Parameters
  • start (Union[datetime64, Series, ndarray, str, datetime, date]) – start datetimes(s)

  • num_days (Union[int, ndarray]) – number of trading days to add

  • roll (str) – one of ‘raise’, ‘nat’, ‘forward’, ‘following’, ‘backward’, ‘preceding’, ‘modifiedfollowing’, ‘modifiedpreceding’ or ‘allow’} ‘allow’ is a special case in which case, adding 1 day to a holiday will act as if it was not a holiday, and give you the next business day’ The rest of the values are the same as in the numpy busday_offset function From numpy documentation: How to treat dates that do not fall on a valid day. The default is ‘raise’. ‘raise’ means to raise an exception for an invalid day. ‘nat’ means to return a NaT (not-a-time) for an invalid day. ‘forward’ and ‘following’ mean to take the first valid day later in time. ‘backward’ and ‘preceding’ mean to take the first valid day earlier in time. ‘modifiedfollowing’ means to take the first valid day later in time unless it is across a Month boundary, in which case to take the first valid day earlier in time. ‘modifiedpreceding’ means to take the first valid day earlier in time unless it is across a Month boundary, in which case to take the first valid day later in time.

Return type

Union[datetime64, ndarray]

Returns

The datetime num_days trading days after start

>>> calendar = Calendar.get_calendar(Calendar.NYSE)
>>> calendar.add_trading_days(datetime.date(2015, 12, 24), 1)
numpy.datetime64('2015-12-28')
>>> calendar.add_trading_days(np.datetime64('2017-04-15'), 0, roll = 'preceding') # 4/14/2017 is a Friday and a holiday
numpy.datetime64('2017-04-13')
>>> calendar.add_trading_days(np.datetime64('2017-04-08'), 0, roll = 'preceding') # 4/7/2017 is a Friday and not a holiday
numpy.datetime64('2017-04-07')
>>> calendar.add_trading_days(np.datetime64('2019-02-17 15:25'), 1, roll = 'allow')
numpy.datetime64('2019-02-19T15:25')
>>> calendar.add_trading_days(np.datetime64('2019-02-17 15:25'), -1, roll = 'allow')
numpy.datetime64('2019-02-15T15:25')
static get_calendar(exchange_name)[source]

Get a calendar object for the given exchange:

Parameters
  • exchange_name (str) – The exchange for which you want a calendar. Calendar.NYSE, Calendar.EUREX are predefined.

  • calendar (If you want to add a new) –

  • function (use the add_calendar class level) –

Return type

Calendar

Returns

The calendar object

get_trading_days(start, end, include_first=False, include_last=True)[source]

Get back a list of numpy dates that are trading days between the start and end

>>> nyse = Calendar.get_calendar(Calendar.NYSE)
>>> nyse.get_trading_days('2005-01-01', '2005-01-08')
array(['2005-01-03', '2005-01-04', '2005-01-05', '2005-01-06', '2005-01-07'], dtype='datetime64[D]')
>>> nyse.get_trading_days(datetime.date(2005, 1, 1), datetime.date(2005, 2, 1))
array(['2005-01-03', '2005-01-04', '2005-01-05', '2005-01-06',
       '2005-01-07', '2005-01-10', '2005-01-11', '2005-01-12',
       '2005-01-13', '2005-01-14', '2005-01-18', '2005-01-19',
       '2005-01-20', '2005-01-21', '2005-01-24', '2005-01-25',
       '2005-01-26', '2005-01-27', '2005-01-28', '2005-01-31', '2005-02-01'], dtype='datetime64[D]')
>>> nyse.get_trading_days(datetime.date(2016, 1, 5), datetime.date(2016, 1, 29), include_last = False)
array(['2016-01-06', '2016-01-07', '2016-01-08', '2016-01-11',
       '2016-01-12', '2016-01-13', '2016-01-14', '2016-01-15',
       '2016-01-19', '2016-01-20', '2016-01-21', '2016-01-22',
       '2016-01-25', '2016-01-26', '2016-01-27', '2016-01-28'], dtype='datetime64[D]')
>>> nyse.get_trading_days('2017-07-04', '2017-07-08', include_first = False)
array(['2017-07-05', '2017-07-06', '2017-07-07'], dtype='datetime64[D]')
>>> nyse.get_trading_days(np.datetime64('2017-07-04'), np.datetime64('2017-07-08'), include_first = False)
array(['2017-07-05', '2017-07-06', '2017-07-07'], dtype='datetime64[D]')
Return type

Union[int, ndarray]

is_trading_day(dates)[source]

Returns whether the date is not a holiday or a weekend

Parameters

dates (Union[datetime64, Series, ndarray, str, datetime, date]) – date times to check

Return type

Union[bool, ndarray]

Returns

Whether this date is a trading day

>>> import datetime
>>> eurex = Calendar.get_calendar(Calendar.EUREX)
>>> eurex.is_trading_day('2016-12-25')
False
>>> eurex.is_trading_day(datetime.date(2016, 12, 22))
True
>>> nyse = Calendar.get_calendar(Calendar.NYSE)
>>> nyse.is_trading_day('2017-04-01') # Weekend
False
>>> nyse.is_trading_day(np.arange('2017-04-01', '2017-04-09', dtype = np.datetime64)) 
array([False, False,  True,  True,  True,  True,  True, False]...)
num_trading_days(start, end, include_first=False, include_last=True)[source]

Count the number of trading days between two date series including those two dates

>>> eurex = Calendar.get_calendar(Calendar.EUREX)
>>> eurex.num_trading_days('2009-01-01', '2011-12-31')
772.0
>>> dates = pd.date_range('20130101',periods=8)
>>> increments = np.array([5, 0, 3, 9, 4, 10, 15, 29])
>>> import warnings
>>> import pandas as pd
>>> warnings.filterwarnings(action = 'ignore', category = pd.errors.PerformanceWarning)
>>> dates2 = dates + increments * dates.freq
>>> df = pd.DataFrame({'x': dates, 'y' : dates2})
>>> df.iloc[4]['x'] = np.nan
>>> df.iloc[6]['y'] = np.nan
>>> nyse = Calendar.get_calendar(Calendar.NYSE)
>>> np.set_printoptions(formatter = {'float' : lambda x : f'{x:.1f}'})  # After numpy 1.13 positive floats don't have a leading space for sign
>>> print(nyse.num_trading_days(df.x, df.y))
[3.0 0.0 1.0 5.0 nan 8.0 nan 20.0]
Return type

Union[float, ndarray]

third_friday_of_month(month, year, roll='backward')[source]
>>> nyse = Calendar.get_calendar(Calendar.NYSE)
>>> nyse.third_friday_of_month(3, 2017)
numpy.datetime64('2017-03-17')
Return type

datetime64

pyqstrat.holiday_calendars.get_date_from_weekday(weekday, year, month, week)[source]

Return the date that falls on a given weekday (Monday = 0) on a week, year and month >>> get_date_from_weekday(1, 2019, 10, 4) numpy.datetime64(‘2019-10-22’)

Return type

datetime64

pyqstrat.holiday_calendars.read_holidays(calendar_name, dirname=None)[source]

Reads a csv with a holidays column containing holidays (not including weekends)

Return type

ndarray

pyqstrat.account module

class pyqstrat.account.Account(contract_groups, timestamps, price_function, strategy_context, starting_equity=1000000.0, pnl_calc_time=900)[source]

Bases: object

An Account calculates pnl for a set of contracts

__init__(contract_groups, timestamps, price_function, strategy_context, starting_equity=1000000.0, pnl_calc_time=900)[source]
Parameters
  • contract_groups (Sequence[ContractGroup]) – Contract groups that we want to compute PNL for

  • timestamps (ndarray) – Timestamps that we might compute PNL at

  • price_function (Callable[[Contract, ndarray, int, SimpleNamespace], float]) – Function that returns contract prices used to compute pnl

  • strategy_context (SimpleNamespace) – This is passed into the price function so we can use current state of strategy to compute prices

  • starting_equity (float) – Starting equity in account currency. Default 1.e6

  • pnl_calc_time (int) – Number of minutes past midnight that we should calculate PNL at. Default 15 * 60, i.e. 3 pm

add_trades(trades)[source]
Return type

None

calc(timestamp)[source]

Computes P&L and stores it internally for all contracts.

Parameters

timestamp (datetime64) – timestamp to compute P&L at. Account remembers the last timestamp it computed P&L up to and will compute P&L between these and including timestamp. If there is more than one day between the last index and current index, we will include pnl for at the defined pnl_calc_time for those dates as well.

Return type

None

df_account_pnl(contract_group=None)[source]

Returns PNL at the account level.

Parameters

contract_group (Optional[ContractGroup]) – If set, we only return pnl for this contract_group. Otherwise we return pnl for all contract groups

Return type

DataFrame

df_pnl(contract_groups=None)[source]

Returns a dataframe with P&L columns broken down by contract group and symbol

Parameters

contract_group – Return PNL for this contract group. If None (default), include all contract groups

Return type

DataFrame

df_roundtrip_trades(contract_group=None, start_date=None, end_date=None)[source]

Returns a dataframe of round trip trades

Parameters
  • contract_group (Optional[ContractGroup]) – Return trades for this contract group. If None (default), include all contract groups

  • start_date (Optional[datetime64]) – Include trades with date greater than or equal to this timestamp.

  • end_date (Optional[datetime64]) – Include trades with date less than or equal to this timestamp.

Return type

DataFrame

df_trades(contract_group=None, start_date=None, end_date=None)[source]

Returns a dataframe of trades

Parameters
  • contract_group (Optional[ContractGroup]) – Return trades for this contract group. If None (default), include all contract groups

  • start_date (Optional[datetime64]) – Include trades with date greater than or equal to this timestamp.

  • end_date (Optional[datetime64]) – Include trades with date less than or equal to this timestamp.

Return type

DataFrame

equity(timestamp)[source]

Returns equity in this account in Account currency. Will cause calculation if Account has not previously calculated up to this date

Return type

float

position(contract_group, timestamp)[source]

Returns netted position for a contract_group at a given date in number of contracts or shares.

Return type

float

positions(contract_group, timestamp)[source]

Returns all non-zero positions in a contract group

Return type

MutableSequence[Tuple[Contract, float]]

roundtrip_trades(contract_group=None, start_date=None, end_date=None)[source]

Returns a list of round trip trades with the given symbol and with trade date between (and including) start date and end date if they are specified. If symbol is None trades for all symbols are returned

Return type

MutableSequence[RoundTripTrade]

symbols()[source]
Return type

MutableSequence[str]

trades(contract_group=None, start_date=None, end_date=None)[source]

Returns a list of trades with the given symbol and with trade date between (and including) start date and end date if they are specified. If symbol is None trades for all symbols are returned

Return type

MutableSequence[Trade]

class pyqstrat.account.ContractPNL(contract, account_timestamps, price_function, strategy_context)[source]

Bases: object

Computes pnl for a single contract over time given trades and market data >>> from pyqstrat.pq_types import MarketOrder >>> ContractGroup.clear() >>> Contract.clear() >>> aapl_cg = ContractGroup.create(‘AAPL’) >>> aapl_contract = Contract.create(‘AAPL’, contract_group=aapl_cg) >>> timestamps = np.arange(np.datetime64(‘2018-01-01’), np.datetime64(‘2018-01-04’)) >>> def get_price(contract, timestamps, idx, strategy_context): … assert contract.symbol == ‘AAPL’, f’unknown contract: {contract}’ … return idx + 10.1

>>> contract_pnl = ContractPNL(aapl_contract, timestamps, get_price, SimpleNamespace())
>>> trade_5 = Trade(aapl_contract, MarketOrder(aapl_contract, timestamps[1], 20), timestamps[2], 10, 16.2)
>>> trade_6 = Trade(aapl_contract, MarketOrder(aapl_contract, timestamps[1], -20), timestamps[2], -10, 16.5)
>>> trade_7 = Trade(aapl_contract, MarketOrder(aapl_contract, timestamps[1], -20), timestamps[2], -10, 16.5)
>>> contract_pnl._add_trades([trade_5, trade_6])
>>> contract_pnl._add_trades([trade_7])
>>> df = contract_pnl.df()
>>> assert (len(df == 1))
>>> row = df.iloc[0]
>>> assert row.to_dict() == {'symbol': 'AAPL',
... 'timestamp': pd.Timestamp('2018-01-03 00:00:00'),
... 'position': -10,
... 'price': 12.1,
... 'unrealized': 44.0,
... 'realized': 3.000000000000007,
... 'commission': 0.0,
... 'fee': 0.0,
... 'net_pnl': 47.00000000000001}
calc_net_pnl(timestamp)[source]
Return type

None

df()[source]

Returns a pandas dataframe with pnl data

Return type

DataFrame

net_pnl(timestamp)[source]
Return type

float

pnl(timestamp)[source]
Return type

Tuple[float, float, float, float, float, float, float]

position(timestamp)[source]
Return type

float

pyqstrat.account.calc_trade_pnl(open_qtys, open_prices, new_qtys, new_prices, multiplier)[source]
>>> print(calc_trade_pnl(
...          open_qtys = np.array([], dtype = float), open_prices = np.array([], dtype = float),
...            new_qtys = np.array([-8, 9, -4]), new_prices = np.array([10, 11, 6]), multiplier = 100))
(array([-3.]), array([6.]), -3.0, 6.0, -1300.0)
>>> print(calc_trade_pnl(open_qtys = np.array([], dtype = float), open_prices = np.array([], dtype = float), new_qtys = np.array([3, 10, -5]),
...          new_prices = np.array([51, 50, 45]), multiplier = 100))
(array([8.]), array([50.]), 8.0, 50.0, -2800.0)
>>> print(calc_trade_pnl(open_qtys = np.array([]), open_prices = np.array([]),
...                new_qtys = np.array([-58, -5, -5, 6, -8, 5, 5, -5, 19, 7, 5, -5, 39]),
...                new_prices = np.array([2080, 2075.25, 2070.75, 2076, 2066.75, 2069.25, 2074.75, 2069.75, 2087.25, 2097.25, 2106, 2088.25, 2085.25]),
...                multiplier = 50))
(array([], dtype=float64), array([], dtype=float64), 0.0, 0, -33762.5)
Return type

Tuple[ndarray, ndarray, float, float, float]

pyqstrat.account.find_index_before(sorted_dict, key)[source]

Find index of the first key in a sorted dict that is less than or equal to the key passed in. If the key is less than the first key in the dict, return -1

Return type

int

pyqstrat.account.find_last_non_nan_index(array)[source]
Return type

int

pyqstrat.account.leading_nan_to_zero(df, columns)[source]
Return type

DataFrame

pyqstrat.account.test_account()[source]

pyqstrat.strategy module

class pyqstrat.strategy.Strategy(timestamps, contract_groups, price_function, starting_equity=1000000.0, pnl_calc_time=901, trade_lag=0, run_final_calc=True, strategy_context=None)[source]

Bases: object

__init__(timestamps, contract_groups, price_function, starting_equity=1000000.0, pnl_calc_time=901, trade_lag=0, run_final_calc=True, strategy_context=None)[source]
Parameters
  • timestamps (np.array of np.datetime64) – The “heartbeat” of the strategy. We will evaluate trading rules and simulate the market at these times.

  • contract_groups (Sequence[ContractGroup]) – The contract groups we will potentially trade.

  • price_function (Callable[[Contract, ndarray, int, SimpleNamespace], float]) – A function that returns the price of a contract at a given timestamp

  • starting_equity (float) – Starting equity in Strategy currency. Default 1.e6

  • pnl_calc_time (int) – Time of day used to calculate PNL. Default 15 * 60 (3 pm)

  • trade_lag (int) – Number of bars you want between the order and the trade. For example, if you think it will take 5 seconds to place your order in the market, and your bar size is 1 second, set this to 5. Set this to 0 if you want to execute your trade at the same time as you place the order, for example, if you have daily bars. Default 0.

  • run_final_calc (bool) – If set, calculates unrealized pnl and net pnl as well as realized pnl when strategy is done. If you don’t need unrealized pnl, turn this off for faster run time. Default True

  • strategy_context (Optional[SimpleNamespace]) – A storage class where you can store key / value pairs relevant to this strategy. For example, you may have a pre-computed table of correlations that you use in the indicator or trade rule functions. If not set, the __init__ function will create an empty member strategy_context object that you can access.

add_indicator(name, indicator, contract_groups=None, depends_on=None)[source]
Parameters
  • name (str) – Name of the indicator

  • indicator (Union[ndarray, Series, Callable[[ContractGroup, ndarray, SimpleNamespace, SimpleNamespace], ndarray]]) – A function that takes strategy timestamps and other indicators and returns a numpy array containing indicator values. The return array must have the same length as the timestamps object. Can also be a numpy array or a pandas Series in which case we just store the values.

  • contract_groups (Optional[Sequence[ContractGroup]]) – Contract groups that this indicator applies to. If not set, it applies to all contract groups. Default None.

  • depends_on (Optional[Sequence[str]]) – Names of other indicators that we need to compute this indicator. Default None.

Return type

None

add_market_sim(market_sim_function)[source]

Add a market simulator. A market simulator is a function that takes orders as input and returns trades.

Return type

None

add_rule(name, rule_function, signal_name, sig_true_values=None, position_filter=None)[source]
Add a trading rule. Trading rules are guaranteed to run in the order in which you add them. For example, if you set trade_lag to 0,

and want to exit positions and re-enter new ones in the same bar, make sure you add the exit rule before you add the entry rule to the strategy.

Parameters
  • name (str) – Name of the trading rule

  • rule_function (Callable[[ContractGroup, int, ndarray, SimpleNamespace, ndarray, Account, SimpleNamespace], Sequence[Order]]) – A trading rule function that returns a list of Orders

  • signal_name (str) – The strategy will call the trading rule function when the signal with this name matches sig_true_values

  • sig_true_values (Optional[Sequence]) – If the signal value at a bar is equal to one of these values, the Strategy will call the trading rule function. Default [TRUE]

  • position_filter (Optional[str]) – Can be “zero”, “nonzero” or None. Zero rules are only triggered when the corresponding contract positions are 0 Nonzero rules are only triggered when the corresponding contract positions are non-zero. If not set, we don’t look at position before triggering the rule. Default None

Return type

None

add_signal(name, signal_function, contract_groups=None, depends_on_indicators=None, depends_on_signals=None)[source]
Parameters
  • name (str) – Name of the signal

  • signal_function (function) – A function that takes timestamps and a dictionary of indicator value arrays and returns a numpy array containing signal values. The return array must have the same length as the input timestamps

  • contract_groups (list of ContractGroup, optional) – Contract groups that this signal applies to. If not set, it applies to all contract groups. Default None.

  • depends_on_indicators (list of str, optional) – Names of indicators that we need to compute this signal. Default None.

  • depends_on_signals (list of str, optional) – Names of other signals that we need to compute this signal. Default None.

Return type

None

df_data(contract_groups=None, add_pnl=True, start_date=None, end_date=None)[source]

Add indicators and signals to end of market data and return as a pandas dataframe.

Parameters
  • contract_groups (list of:obj:ContractGroup, optional) – list of contract groups to include. All if set to None (default)

  • add_pnl (bool) – If True (default), include P&L columns in dataframe

  • start_date (Union[str, datetime64, None]) – string or numpy datetime64. Default None

  • end_date (Union[str, datetime64, None]) – string or numpy datetime64: Default None

Return type

DataFrame

df_orders(contract_group=None, start_date=None, end_date=None)[source]

Returns a dataframe with data from orders with the given contract group and with order date between (and including) start date and end date if they are specified. If contract_group is None orders for all contract_groups are returned

Return type

DataFrame

df_pnl(contract_group=None)[source]

Returns a dataframe with P&L columns. If contract group is set to None (default), sums up P&L across all contract groups

Return type

DataFrame

df_returns(contract_group=None, sampling_frequency='D')[source]

Return a dataframe of returns and equity indexed by date.

Parameters
  • contract_group (Optional[ContractGroup]) – The contract group to get returns for. If set to None (default), we return the sum of PNL for all contract groups

  • sampling_frequency (str) – Downsampling frequency. Default is None. See pandas frequency strings for possible values

Return type

DataFrame

df_roundtrip_trades(contract_group=None, start_date=None, end_date=None)[source]

Returns a dataframe of round trip trades with the given contract group and with trade date between (and including) start date and end date if they are specified. If contract_group is None trades for all contract_groups are returned

Return type

DataFrame

df_trades(contract_group=None, start_date=None, end_date=None)[source]

Returns a dataframe with data from trades with the given contract group and with trade date between (and including) start date and end date if they are specified. If contract_group is None trades for all contract_groups are returned

Return type

DataFrame

evaluate_returns(contract_group=None, plot=True, display_summary=True, float_precision=4, return_metrics=False)[source]

Returns a dictionary of common return metrics.

Parameters
  • contract_group (ContractGroup, optional) – Contract group to evaluate or None (default) for all contract groups

  • plot (bool) – If set to True, display plots of equity, drawdowns and returns. Default False

  • float_precision (float, optional) – Number of significant figures to show in returns. Default 4

  • return_metrics (bool, optional) – If set, we return the computed metrics as a dictionary

Return type

Optional[Mapping]

orders(contract_group=None, start_date=None, end_date=None)[source]

Returns a list of orders with the given contract group and with order date between (and including) start date and end date if they are specified. If contract_group is None orders for all contract_groups are returned

Return type

Sequence[Order]

plot(contract_groups=None, primary_indicators=None, primary_indicators_dual_axis=None, secondary_indicators=None, secondary_indicators_dual_axis=None, indicator_properties=None, signals=None, signal_properties=None, pnl_columns=None, title=None, figsize=(20, 15), _date_range=None, date_format=None, sampling_frequency=None, trade_marker_properties=None, hspace=0.15)[source]

Plot indicators, signals, trades, position, pnl

Parameters
  • contract_groups (Optional[Sequence[ContractGroup]]) – Contract groups to plot or None (default) for all contract groups.

  • indicators (primary) – List of indicators to plot in the main indicator section. Default None (plot everything)

  • indicators – List of indicators to plot in the secondary indicator section. Default None (don’t plot anything)

  • indicator_properties (Optional[Mapping[str, Mapping[str, Any]]]) – If set, we use the line color, line type indicated for the given indicators

  • signals (Optional[Sequence[str]]) – Signals to plot. Default None (plot everything).

  • plot_equity – If set, we plot the equity curve. Default is True

  • title (Optional[str]) – Title of plot. Default None

  • figsize (Tuple[int, int]) – Figure size. Default (20, 15)

  • date_range – Used to restrict the date range of the graph. Default None

  • date_format (Optional[str]) – Date format for tick labels on x axis. If set to None (default), will be selected based on date range. See matplotlib date format strings

  • sampling_frequency (Optional[str]) – Downsampling frequency. The graph may get too busy if you have too many bars of data, so you may want to downsample before plotting. See pandas frequency strings for possible values. Default None.

  • trade_marker_properties (Optional[Mapping[str, Mapping[str, Any]]]) – A dictionary of order reason code -> marker shape, marker size, marker color for plotting trades with different reason codes. By default we use the dictionary from the ReasonCode class

  • hspace (float) – Height (vertical) space between subplots. Default is 0.15

Return type

None

plot_returns(contract_group=None)[source]

Display plots of equity, drawdowns and returns for the given contract group or for all contract groups if contract_group is None (default)

Return type

Optional[Tuple[Figure, Axes]]

roundtrip_trades(contract_group=None, start_date=None, end_date=None)[source]

Returns a list of trades with the given contract group and with trade date between (and including) start date and end date if they are specified. If contract_group is None trades for all contract_groups are returned

Return type

Sequence[RoundTripTrade]

run()[source]
Return type

None

run_indicators(indicator_names=None, contract_groups=None, clear_all=False)[source]

Calculate values of the indicators specified and store them.

Parameters
  • indicator_names (Optional[Sequence[str]]) – List of indicator names. If None (default) run all indicators

  • contract_groups (Optional[Sequence[ContractGroup]]) – Contract group to run this indicator for. If None (default), we run it for all contract groups.

  • clear_all (bool) – If set, clears all indicator values before running. Default False.

Return type

None

run_rules(rule_names=None, contract_groups=None, start_date=None, end_date=None)[source]

Run trading rules.

Parameters
  • rule_names (Optional[Sequence[str]]) – List of rule names. If None (default) run all rules

  • contract_groups (Optional[Sequence[ContractGroup]]) – Contract groups to run this rule for. If None (default), we run it for all contract groups.

  • start_date (Optional[datetime64]) – Run rules starting from this date. Default None

  • end_date (Optional[datetime64]) – Don’t run rules after this date. Default None

Return type

None

run_signals(signal_names=None, contract_groups=None, clear_all=False)[source]

Calculate values of the signals specified and store them.

Parameters
  • signal_names (Optional[Sequence[str]]) – List of signal names. If None (default) run all signals

  • contract_groups (Optional[Sequence[ContractGroup]]) – Contract groups to run this signal for. If None (default), we run it for all contract groups.

  • clear_all (bool) – If set, clears all signal values before running. Default False.

Return type

None

trades(contract_group=None, start_date=None, end_date=None)[source]

Returns a list of trades with the given contract group and with trade date between (and including) start date and end date if they are specified. If contract_group is None trades for all contract_groups are returned

Return type

Sequence[Trade]

pyqstrat.strategy.test_strategy()[source]
Return type

Strategy

pyqstrat.strategy.test_strategy_2()[source]

Test of a dummy strategy

pyqstrat.portfolio module

class pyqstrat.portfolio.Portfolio(name='main')[source]

Bases: object

A portfolio contains one or more strategies that run concurrently so you can test running strategies that are uncorrelated together.

__init__(name='main')[source]

Args: name: String used for displaying this portfolio

add_strategy(name, strategy)[source]
Parameters
  • name (str) – Name of the strategy

  • strategy (Strategy) – Strategy instance

Return type

None

df_returns(sampling_frequency='D', strategy_names=None)[source]

Return dataframe containing equity and returns with a date index. Equity and returns are combined from all strategies passed in.

Parameters
  • sampling_frequency (str) – Date frequency for rows. Default ‘D’ for daily so we will have one row per day

  • strategy_names (Optional[Sequence[str]]) – By default this is set to None and we use all strategies.

Return type

DataFrame

evaluate_returns(sampling_frequency='D', strategy_names=None, plot=True, float_precision=4)[source]

Returns a dictionary of common return metrics.

Parameters
  • sampling_frequency (str) – Date frequency. Default ‘D’ for daily so we downsample to daily returns before computing metrics

  • strategy_names (Optional[Sequence[str]]) – By default this is set to None and we use all strategies.

  • plot (bool) – If set to True, display plots of equity, drawdowns and returns. Default False

  • float_precision (int) – Number of significant figures to show in returns. Default 4

Return type

Mapping

plot(sampling_frequency='D', strategy_names=None)[source]

Display plots of equity, drawdowns and returns

Parameters
  • sampling_frequency (str) – Date frequency. Default ‘D’ for daily so we downsample to daily returns before computing metrics

  • strategy_names (Optional[Sequence[str]]) – A list of strategy names. By default this is set to None and we use all strategies.

Return type

None

run(strategy_names=None, start_date=None, end_date=None)[source]

Run indicators, signals and rules.

Parameters
  • strategy_names (Optional[Sequence[str]]) – A list of strategy names. By default this is set to None and we use all strategies.

  • start_date (Optional[datetime64]) – Run rules starting from this date. Sometimes we have a few strategies in a portfolio that need different lead times before they are ready to trade so you can set this so they are all ready by this date. Default None

  • end_date (Optional[datetime64]) – Don’t run rules after this date. Default None

Return type

None

run_indicators(strategy_names=None)[source]

Compute indicators for the strategies specified

Parameters

strategy_names (Optional[Sequence[str]]) – By default this is set to None and we use all strategies.

Return type

None

run_rules(strategy_names=None, start_date=None, end_date=None)[source]

Run rules for the strategies specified. Must be called after run_indicators and run_signals. See run function for argument descriptions

Return type

None

run_signals(strategy_names=None)[source]

Compute signals for the strategies specified. Must be called after run_indicators

Parameters

strategy_names (Optional[Sequence[str]]) – By default this is set to None and we use all strategies.

Return type

None

pyqstrat.optimize module

class pyqstrat.optimize.Experiment(suggestion, cost, other_costs)[source]

Bases: object

An Experiment stores a suggestion and its result

__init__(suggestion, cost, other_costs)[source]
Parameters
  • suggestion (Mapping[str, Any]) – A dictionary of variable name -> value

  • cost (float) – A float representing output of the function we are testing with this suggestion as input.

  • other_costs (Mapping[str, float]) – A dictionary of other results we want to store and look at later.

valid()[source]

Returns True if all suggestions and costs are finite, i.e not NaN or +/- Infinity

Return type

bool

class pyqstrat.optimize.Optimizer(name, generator, cost_func, max_processes=None)[source]

Bases: object

Optimizer is used to optimize parameters for a strategy.

__init__(name, generator, cost_func, max_processes=None)[source]
Parameters
  • name (str) – Display title for plotting, etc.

  • generator (Generator[Mapping[str, Any], Tuple[float, Mapping[str, float]], None]) – A generator (see Python Generators) that takes no inputs and yields a dictionary with parameter name -> parameter value.

  • cost_func (Callable[[Mapping[str, Any]], Tuple[float, Mapping[str, float]]]) – A function that takes a dictionary of parameter name -> parameter value as input and outputs cost for that set of parameters.

  • max_processes (Optional[int]) – If not set, the Optimizer will look at the number of CPU cores on your machine to figure out how many processes to run.

df_experiments(sort_column='cost', ascending=True)[source]

Returns a dataframe containing experiment data, sorted by sort_column (default “cost”)

Return type

DataFrame

experiment_list(sort_order='lowest_cost')[source]

Returns the list of experiments we have run

Parameters

sort_order (str) – Can be set to lowest_cost, highest_cost or sequence. If set to sequence, experiments are returned in the sequence in which they were run

Return type

Sequence[Experiment]

plot_2d(x, y='all', plot_type='line', figsize=(15, 8), marker='X', marker_size=50, marker_color='r', xlim=None, hspace=None)[source]

Creates a 2D plot of the optimization output for plotting 1 parameter and costs

Parameters
  • x (str) – Name of the parameter to plot on the x axis, corresponding to the same name in the generator.

  • y (str) – Can be one of: “cost” The name of another cost variable corresponding to the output from the cost function “all”, which creates a subplot for cost plus all other costs

  • plot_type (str) – line or scatter (default line)

  • figsize (Tuple[float, float]) – Figure size

  • marker (str) – Adds a marker to each point in x, y to show the actual data used for interpolation. You can set this to None to turn markers off.

  • hspace (Optional[float]) – Vertical space between subplots

Return type

None

plot_3d(x, y, z='all', filter_func=None, plot_type='surface', figsize=(15, 15), interpolation='linear', cmap='viridis', marker='X', marker_size=50, marker_color='r', xlim=None, ylim=None, hspace=None)[source]

Creates a 3D plot of the optimization output for plotting 2 parameters and costs.

Parameters
  • x (str) – Name of the parameter to plot on the x axis, corresponding to the same name in the generator.

  • y (str) – Name of the parameter to plot on the y axis, corresponding to the same name in the generator.

  • z (str) – Can be one of: “cost” The name of another cost variable corresponding to the output from the cost function “all”, which creates a subplot for cost plus all other costs

  • filter_func (Optional[Callable[[DataFrame], DataFrame]]) – A function that can be used to reduce the dataset before plotting. For example, you may want to filter on a dimension beyond x, y, z to pick a single value from that dimension

  • plot_type (str) – surface or contour (default surface)

  • figsize (Tuple[float, float]) – Figure size

  • interpolation (str) – Can be ‘linear’, ‘nearest’ or ‘cubic’ for plotting z points between the ones passed in. See scipy.interpolate.griddata for details

  • cmap (str) – Colormap to use (default viridis). See matplotlib colormap for details

  • marker (str) – Adds a marker to each point in x, y, z to show the actual data used for interpolation. You can set this to None to turn markers off.

  • hspace (Optional[float]) – Vertical space between subplots

Return type

None

run(raise_on_error=False)[source]

Run the optimizer.

Parameters

raise_on_error (bool) – If set to True, even if we are running a multiprocess optimization, any Exceptions will bubble up and stop the Optimizer. This can be useful for debugging to see stack traces for Exceptions.

Return type

None

pyqstrat.optimize.flatten_keys(experiments)[source]

Utility function so we can find all keys for other costs in all experiments even if the first experiment does not have all of them

Return type

List[str]

pyqstrat.optimize.test_optimize()[source]

pyqstrat.plot module

class pyqstrat.plot.BarPlotAttributes(color='red')[source]

Bases: DisplayAttributes

color: str = 'red'
class pyqstrat.plot.BoxPlotAttributes(proportional_widths=True, show_means=True, show_all=True, show_outliers=False, notched=False)[source]

Bases: DisplayAttributes

proportional_widths

if set to True, the width each box in the boxplot will be proportional to the number of items in its corresponding array

Type

bool

show_means

Whether to display a marker where the mean is for each array

Type

bool

show_outliers

Whether to show markers for outliers that are outside the whiskers. Box is at Q1 = 25%, Q3 = 75% quantiles, whiskers are at Q1 - 1.5 * (Q3 - Q1), Q3 + 1.5 * (Q3 - Q1)

Type

bool

notched

Whether to show notches indicating the confidence interval around the median

Type

bool

notched: bool = False
proportional_widths: bool = True
show_all: bool = True
show_means: bool = True
show_outliers: bool = False
class pyqstrat.plot.BucketedValues(name, bucket_names, bucket_values, display_attributes=None)[source]

Bases: PlotData

Data in a subplot where we summarize properties of a numpy array.

For example, drawing a boxplot with percentiles. x axis is a categorical

__init__(name, bucket_names, bucket_values, display_attributes=None)[source]
Parameters
  • name (str) – name used for this data in a plot legend

  • bucket_names (Sequence[str]) – list of strings used on x axis labels

  • bucket_values (Sequence[ndarray]) – list of numpy arrays that are summarized in this plot

display_attributes: DisplayAttributes
name: str
class pyqstrat.plot.CandleStickPlotAttributes(colorup='darkgreen', colordown='#F2583E')[source]

Bases: DisplayAttributes

colordown: str = '#F2583E'
colorup: str = 'darkgreen'
class pyqstrat.plot.ContourPlotAttributes(marker='X', marker_size=50, marker_color='red', interpolation='linear', cmap=<matplotlib.colors.LinearSegmentedColormap object>, min_level=nan, max_level=nan)[source]

Bases: DisplayAttributes

cmap: Colormap = <matplotlib.colors.LinearSegmentedColormap object>
interpolation: str = 'linear'
marker: str = 'X'
marker_color: str = 'red'
marker_size: int = 50
max_level: float = nan
min_level: float = nan
class pyqstrat.plot.DateFormatter(timestamps, fmt)[source]

Bases: Formatter

Formats timestamps on plot axes. See matplotlib Formatter

class pyqstrat.plot.DateLine(date, name=None, line_type='dashed', color=None)[source]

Bases: object

Draw a vertical line on a plot with a datetime x-axis

class pyqstrat.plot.DisplayAttributes[source]

Bases: object

class pyqstrat.plot.FilledLinePlotAttributes(positive_color='blue', negative_color='red')[source]

Bases: DisplayAttributes

colorup: Color for bars where close >= open. Default “darkgreen” colordown: Color for bars where open < close. Default “#F2583E”

negative_color: str = 'red'
positive_color: str = 'blue'
class pyqstrat.plot.HorizontalLine(y, name=None, line_type='dashed', color=None)[source]

Bases: object

Draws a horizontal line on a subplot

class pyqstrat.plot.LinePlotAttributes(line_type='solid', line_width=None, color=None, marker=None, marker_size=None, marker_color=None)[source]

Bases: DisplayAttributes

color: Optional[str] = None
line_type: Optional[str] = 'solid'
line_width: Optional[int] = None
marker: Optional[str] = None
marker_color: Optional[str] = None
marker_size: Optional[int] = None
class pyqstrat.plot.Plot(subplot_list, title=None, figsize=(15, 8), date_range=None, date_format=None, sampling_frequency=None, show_grid=True, show_date_gaps=True, hspace=0.15)[source]

Bases: object

Top level plot containing a list of subplots to draw

__init__(subplot_list, title=None, figsize=(15, 8), date_range=None, date_format=None, sampling_frequency=None, show_grid=True, show_date_gaps=True, hspace=0.15)[source]
Parameters
  • subplot_list (Sequence[Subplot]) – List of Subplot objects to draw

  • title (Optional[str]) – Title for this plot. Default None

  • figsize (Tuple[float, float]) – Figure size. Default (15, 8)

  • date_range (Union[Tuple[str, str], Tuple[Optional[datetime64], Optional[datetime64]], None]) – Tuple of strings or numpy datetime64 limiting timestamps to draw. e.g. (“2018-01-01 14:00”, “2018-01-05”). Default None

  • date_format (Optional[str]) – Date format to use for x-axis

  • sampling_frequency (Optional[str]) – Set this to downsample subplots that have a datetime x axis. For example, if you have minute bar data, you might want to subsample to hours if the plot is too crowded. See pandas time frequency strings for possible values. Default None

  • show_grid (bool) – If set to True, show a grid on the subplots. Default True

  • show_date_gaps (bool) – If set to True, then when there is a gap between timestamps will draw a dashed vertical line. For example, you may have minute bars and a gap between end of trading day and beginning of next day. Even if set to True, this will turn itself off if there are too many gaps to avoid clutter. Default True

  • hspace (Optional[float]) – Height (vertical) space between subplots. Default 0.15

draw(check_data_size=True)[source]

Draw the subplots.

Parameters

check_data_size (bool) – If set to True, will not plot if there are > 100K points to avoid locking up your computer for a long time. Default True

Return type

Optional[Tuple[Figure, Axes]]

class pyqstrat.plot.PlotData[source]

Bases: object

display_attributes: DisplayAttributes
name: str
class pyqstrat.plot.ScatterPlotAttributes(marker='X', marker_size=50, marker_color='red')[source]

Bases: DisplayAttributes

marker: str = 'X'
marker_color: str = 'red'
marker_size: int = 50
class pyqstrat.plot.Subplot(data_list, secondary_y=None, title=None, xlabel=None, ylabel=None, zlabel=None, date_lines=None, horizontal_lines=None, vertical_lines=None, xlim=None, ylim=None, height_ratio=1.0, display_legend=True, legend_loc='best', log_y=False, y_tick_format=None)[source]

Bases: object

A top level plot contains a list of subplots, each of which contain a list of data objects to draw

__init__(data_list, secondary_y=None, title=None, xlabel=None, ylabel=None, zlabel=None, date_lines=None, horizontal_lines=None, vertical_lines=None, xlim=None, ylim=None, height_ratio=1.0, display_legend=True, legend_loc='best', log_y=False, y_tick_format=None)[source]
Parameters
  • data_list (Union[PlotData, Sequence[PlotData]]) – A list of objects to draw. Each element can contain XYData, XYZData, TimeSeries, TradeBarSeries, BucketedValues or TradeSet

  • secondary_y (Optional[Sequence[str]]) – A list of objects to draw on the secondary y axis

  • title (Optional[str]) – Title to show for this subplot. Default None

  • zlabel (Optional[str]) – Only applicable to 3d subplots. Default None

  • date_lines (Optional[Sequence[DateLine]]) – A list of DateLine objects to draw as vertical lines. Only applicable when x axis is datetime. Default None

  • horizontal_lines (Optional[Sequence[HorizontalLine]]) – A list of HorizontalLine objects to draw on the plot. Default None

  • vertical_lines (Optional[Sequence[VerticalLine]]) – A list of VerticalLine objects to draw on the plot

  • xlim (Union[Tuple[float, float], Tuple[datetime64, datetime64], None]) – x limits for the plot as a tuple of numpy datetime objects when x-axis is datetime, or tuple of floats. Default None

  • ylim (Union[Tuple[float, float], Tuple[datetime64, datetime64], None]) – y limits for the plot. Tuple of floats. Default None

  • height_ratio (float) – If you have more than one subplot on a plot, use height ratio to determine how high each subplot should be. For example, if you set height_ratio = 0.75 for the first subplot and 0.25 for the second, the first will be 3 times taller than the second one. Default 1.0

  • display_legend (bool) – Whether to show a legend on the plot. Default True

  • legend_loc (str) – Location for the legend. Default ‘best’

  • log_y (bool) – Whether the y axis should be logarithmic. Default False

  • y_tick_format (Optional[str]) – Format string to use for y axis labels. For example, you can decide to use fixed notation instead of scientific notation or change number of decimal places shown. Default None

get_all_timestamps(date_range)[source]
Return type

ndarray

class pyqstrat.plot.SurfacePlotAttributes(marker='X', marker_size=50, marker_color='red', interpolation='linear', cmap=<matplotlib.colors.LinearSegmentedColormap object>)[source]

Bases: DisplayAttributes

marker

Adds a marker to each point in x, y, z to show the actual data used for interpolation. You can set this to None to turn markers off.

Type

str

interpolation

Can be ‘linear’, ‘nearest’ or ‘cubic’ for plotting z points between the ones passed in. See scipy.interpolate.griddata for details

Type

str

cmap

Colormap to use (default matplotlib.cm.RdBu_r). See matplotlib colormap for details

Type

matplotlib.colors.Colormap

cmap: Colormap = <matplotlib.colors.LinearSegmentedColormap object>
interpolation: str = 'linear'
marker: str = 'X'
marker_color: str = 'red'
marker_size: int = 50
class pyqstrat.plot.TimePlotData[source]

Bases: PlotData

abstract reindex(timestamps, fill)[source]
Return type

None

timestamps: ndarray
class pyqstrat.plot.TimeSeries(name, timestamps, values, display_attributes=None)[source]

Bases: TimePlotData

Data in a subplot where x is an array of numpy datetimes and y is a numpy array of floats

__init__(name, timestamps, values, display_attributes=None)[source]
Parameters

name (str) – Name to show in plot legend

display_attributes: DisplayAttributes
name: str
reindex(timestamps, fill)[source]

Reindex this series given a new array of timestamps, forward filling holes if fill is set to True

Return type

None

timestamps: ndarray
class pyqstrat.plot.TradeBarSeries(name, timestamps, o, h, l, c, v=None, vwap=None, display_attributes=None)[source]

Bases: TimePlotData

Data in a subplot that contains open, high, low, close, volume bars. volume is optional.

__init__(name, timestamps, o, h, l, c, v=None, vwap=None, display_attributes=None)[source]
Parameters

name (str) – Name to show in a legend

df()[source]
Return type

DataFrame

display_attributes: DisplayAttributes
name: str
reindex(all_timestamps, fill)[source]
Return type

None

timestamps: ndarray
class pyqstrat.plot.TradeSet(name, trades, display_attributes=None)[source]

Bases: TimePlotData

Data for subplot that contains a set of trades along with marker properties for these trades

__init__(name, trades, display_attributes=None)[source]
Parameters
  • name (str) – String to display in a subplot legend

  • trades (Sequence[Trade]) – List of Trade objects to plot

display_attributes: DisplayAttributes
name: str
reindex(all_timestamps, fill)[source]
Return type

None

timestamps: ndarray
class pyqstrat.plot.VerticalLine(x, name=None, line_type='dashed', color=None)[source]

Bases: object

Draws a vertical line on a subplot where x axis is not a date-time axis

class pyqstrat.plot.XYData(name, x, y, display_attributes=None)[source]

Bases: PlotData

Data in a subplot that has x and y values that are both arrays of floats

display_attributes: DisplayAttributes
name: str
class pyqstrat.plot.XYZData(name, x, y, z, display_attributes=None)[source]

Bases: PlotData

Data in a subplot that has x, y and z values that are all floats

__init__(name, x, y, z, display_attributes=None)[source]
Parameters

name (str) – Name to show in plot legend

display_attributes: DisplayAttributes
name: str
pyqstrat.plot.draw_3d_plot(ax, x, y, z, plot_type='contour', marker='X', marker_size=50, marker_color='red', interpolation='linear', cmap=<matplotlib.colors.LinearSegmentedColormap object>, min_level=nan, max_level=nan)[source]

Draw a 3d plot. See XYZData class for explanation of arguments

>>> points = np.random.rand(1000, 2)
>>> x = np.random.rand(10)
>>> y = np.random.rand(10)
>>> z = x ** 2 + y ** 2
>>> if has_display():
...    fig, ax = plt.subplots()
...    draw_3d_plot(ax, x = x, y = y, z = z, plot_type = 'contour', interpolation = 'linear');
Return type

None

pyqstrat.plot.draw_boxplot(ax, names, values, proportional_widths=True, notched=False, show_outliers=True, show_means=True, show_all=True)[source]

Draw a boxplot. See BucketedValues class for explanation of arguments

Return type

None

pyqstrat.plot.draw_candlestick(ax, index, o, h, l, c, v, vwap, colorup='darkgreen', colordown='#F2583E')[source]

Draw candlesticks given parrallel numpy arrays of o, h, l, c, v values. v is optional. See TradeBarSeries class __init__ for argument descriptions.

Return type

None

pyqstrat.plot.draw_date_line(ax, plot_timestamps, date, linestyle, color)[source]

Draw vertical line on a subplot with datetime x axis

Return type

Line2D

pyqstrat.plot.draw_horizontal_line(ax, y, linestyle, color)[source]

Draw horizontal line on a subplot

Return type

Line2D

pyqstrat.plot.draw_poly(ax, left, bottom, top, right, facecolor, edgecolor, zorder)[source]

Draw a set of polygrams given parrallel numpy arrays of left, bottom, top, right points

Return type

None

pyqstrat.plot.draw_vertical_line(ax, x, linestyle, color)[source]

Draw vertical line on a subplot

Return type

Line2D

pyqstrat.plot.get_date_formatter(plot_timestamps, date_format)[source]

Create an appropriate DateFormatter for x axis labels. If date_format is set to None, figures out an appropriate date format based on the range of timestamps passed in

Return type

DateFormatter

pyqstrat.plot.test_plot()[source]
Return type

None

pyqstrat.plot.trade_sets_by_reason_code(trades, marker_props={'backtest end': {'color': 'green', 'size': 50, 'symbol': '*'}, 'enter long': {'color': 'blue', 'size': 50, 'symbol': 'P'}, 'enter short': {'color': 'red', 'size': 50, 'symbol': 'P'}, 'exit long': {'color': 'blue', 'size': 50, 'symbol': 'X'}, 'exit short': {'color': 'red', 'size': 50, 'symbol': 'X'}, 'none': {'color': 'green', 'size': 50, 'symbol': 'o'}, 'roll future': {'color': 'green', 'size': 50, 'symbol': '>'}}, remove_missing_properties=True)[source]

Returns a list of TradeSet objects. Each TradeSet contains trades with a different reason code. The markers for each TradeSet are set by looking up marker properties for each reason code using the marker_props argument:

Parameters
  • trades (List[Trade]) – We look up reason codes using the reason code on the corresponding orders

  • marker_props (Mapping[str, Mapping]) – Dictionary from reason code string -> dictionary of marker properties. See ReasonCode.MARKER_PROPERTIES for example. Default ReasonCode.MARKER_PROPERTIES

  • remove_missing_properties (bool) – If set, we remove any reason codes that dont’ have marker properties set. Default True

Return type

List[TradeSet]

pyqstrat.evaluator module

class pyqstrat.evaluator.Evaluator(initial_metrics)[source]

Bases: object

You add functions to the evaluator that are dependent on the outputs of other functions. The evaluator will call these functions in the right order so dependencies are computed first before the functions that need their output. You can retrieve the output of a metric using the metric member function

>>> evaluator = Evaluator(initial_metrics={'x': np.array([1, 2, 3]), 'y': np.array([3, 4, 5])})
>>> evaluator.add_metric('z', lambda x, y: sum(x, y), dependencies=['x', 'y'])
>>> evaluator.compute()
>>> evaluator.metric('z')
array([ 9, 10, 11])
__init__(initial_metrics)[source]

Inits Evaluator with a dictionary of initial metrics that are used to compute subsequent metrics

Parameters

initial_metrics (Dict[str, Any]) – a dictionary of string name -> metric. metric can be any object including a scalar, an array or a tuple

add_metric(name, func, dependencies)[source]
Return type

None

compute(metric_names=None)[source]

Compute metrics using the internal dependency graph

Parameters

metric_names (Optional[Sequence[str]]) – an array of metric names. If not passed in, evaluator will compute and store all metrics

Return type

None

compute_metric(metric_name)[source]

Compute and store a single metric:

Parameters

metric_name (str) – string representing the metric to compute

Return type

None

metric(metric_name)[source]

Return the value of a single metric given its name

Return type

Any

metrics()[source]

Return a dictionary of metric name -> metric value

Return type

Mapping[str, Any]

pyqstrat.evaluator.compute_amean(returns, periods_per_year)[source]

Computes arithmetic mean of a return array, ignoring NaNs

Parameters
  • returns (ndarray) – Represents returns at any frequency

  • periods_per_year (int) – Frequency of the returns, e.g. 252 for daily returns

>>> compute_amean(np.array([0.003, 0.004, np.nan]), 252)
0.882
Return type

float

pyqstrat.evaluator.compute_annual_returns(timestamps, returns, periods_per_year)[source]

Takes the output of compute_bucketed_returns and returns geometric mean of returns by year

Return type

Tuple[ndarray, ndarray]

Returns

A tuple with the first element being an array of years (integer) and the second element an array of annualized returns for those years

pyqstrat.evaluator.compute_bucketed_returns(timestamps, returns)[source]

Bucket returns by year

Return type

Tuple[Sequence[int], Sequence[ndarray]]

Returns

A tuple with the first element being a list of years and the second a list of

numpy arrays containing returns for each corresponding year

pyqstrat.evaluator.compute_calmar(returns_3yr, periods_per_year, mdd_pct_3yr)[source]

Compute Calmar ratio, which is the annualized return divided by max drawdown over the last 3 years

Return type

float

pyqstrat.evaluator.compute_dates_3yr(timestamps)[source]

Given an array of numpy datetimes, return those that are within 3 years of the last date in the array

Return type

ndarray

pyqstrat.evaluator.compute_equity(timestamps, starting_equity, returns)[source]

Given starting equity, timestamps and returns, create a numpy array of equity at each date

Return type

ndarray

pyqstrat.evaluator.compute_gmean(timestamps, returns, periods_per_year)[source]

Compute geometric mean of an array of returns

Parameters
  • returns (ndarray) – a numpy array of returns, can contain nans

  • periods_per_year (float) – Used for annualizing returns

>>> round(compute_gmean(np.array(['2015-01-01', '2015-03-01', '2015-05-01'], dtype='M8[D]'), np.array([0.001, 0.002, 0.003]), 252.), 6)
0.018362
Return type

float

pyqstrat.evaluator.compute_k_ratio(equity, periods_per_year, halflife_years=None)[source]

Compute k-ratio (2013 or original versions by Lars Kestner). See https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2230949 We also implement a modification that allows higher weighting for more recent returns.

Parameters
  • equity (ndarray) – a numpy array of the equity in your account

  • periods_per_year (int) – 252 for daily values

  • halflife_years (Optional[float]) – If set, we use weighted linear regression to give less weight to older returns. In this case, we compute the original k-ratio which does not use periods per year or number of observations If not set, we compute the 2013 version of the k-ratio which weights k-ratio by sqrt(periods_per_year) / nobs

Return type

float

Returns

weighted or unweighted k-ratio

>>> np.random.seed(0)
>>> t = np.arange(1000)
>>> ret = np.random.normal(loc = 0.0025, scale = 0.01, size = len(t))
>>> equity = (1 + ret).cumprod()
>>> assert(math.isclose(compute_k_ratio(equity, 252, None), 3.888, abs_tol=0.001))
>>> assert(math.isclose(compute_k_ratio(equity, 252, 0.5), 602.140, abs_tol=0.001))
pyqstrat.evaluator.compute_mar(returns, periods_per_year, mdd_pct)[source]

Compute MAR ratio, which is annualized return divided by biggest drawdown since inception.

Return type

float

pyqstrat.evaluator.compute_maxdd_date(rolling_dd_dates, rolling_dd)[source]

Compute date of max drawdown given numpy array of timestamps, and corresponding rolling dd percentages

Return type

datetime64

pyqstrat.evaluator.compute_maxdd_date_3yr(rolling_dd_3yr_timestamps, rolling_dd_3yr)[source]

Compute max drawdown date over the last 3 years

Return type

datetime64

pyqstrat.evaluator.compute_maxdd_pct(rolling_dd)[source]

Compute max drawdown percentage given a numpy array of rolling drawdowns, ignoring NaNs

Return type

float

pyqstrat.evaluator.compute_maxdd_pct_3yr(rolling_dd_3yr)[source]

Compute max drawdown percentage over the last 3 years

Return type

float

pyqstrat.evaluator.compute_maxdd_start(rolling_dd_dates, rolling_dd, mdd_date)[source]

Compute date when max drawdown starts, given numpy array of timestamps corresponding rolling dd percentages and date of the max draw down

Return type

datetime64

pyqstrat.evaluator.compute_maxdd_start_3yr(rolling_dd_3yr_timestamps, rolling_dd_3yr, mdd_date_3yr)[source]

Comput max drawdown start date over the last 3 years

Return type

datetime64

pyqstrat.evaluator.compute_num_periods(timestamps, periods_per_year)[source]
Given an array of timestamps, we compute how many periods there are between the first and last element, where the length

of a period is defined by periods_per_year. For example, if there are 6 periods per year, then each period would be approx. 2 months long.

Parameters
  • timestamps (np.ndarray of np.datetime64) – a numpy array of returns, can contain nans

  • periods_per_year (float) – number of periods between first and last return

>>> assert(compute_num_periods(np.array(['2015-01-01', '2015-03-01', '2015-05-01'], dtype='M8[D]'), 6) == 2)
Return type

float

pyqstrat.evaluator.compute_periods_per_year(timestamps)[source]
Computes trading periods per year for an array of numpy datetime64’s.

e.g. if most of the timestamps are separated by 1 day, will return 252.

Parameters

timestamps (ndarray) – a numpy array of datetime64’s

>>> compute_periods_per_year(np.array(['2018-01-01', '2018-01-02', '2018-01-03', '2018-01-09'], dtype='M8[D]'))
252.0
>>> round(compute_periods_per_year(np.array(['2018-01-01 10:00', '2018-01-01 10:05', '2018-01-01 10:10'], dtype='M8[m]')), 2)
72576.05
Return type

float

pyqstrat.evaluator.compute_return_metrics(timestamps, rets, starting_equity, leading_non_finite_to_zeros=False, subsequent_non_finite_to_zeros=True)[source]

Compute a set of common metrics using returns (for example, of an instrument or a portfolio)

Parameters
  • timestamps (np.array of datetime64) – Timestamps for the returns

  • rets (nd.array of float) – The returns, use 0.01 for 1%

  • starting_equity (float) – Starting equity value in your portfolio

  • leading_non_finite_to_zeros (bool, optional) – If set, we replace leading nan, inf, -inf returns with zeros. For example, you may need a warmup period for moving averages. Default False

  • subsequent_non_finite_to_zeros (bool, optional) – If set, we replace any nans that follow the first non nan value with zeros. There may be periods where you have no prices but removing these returns would result in incorrect annualization. Default True

Return type

Evaluator

Returns

An Evaluator object containing computed metrics off the returns passed in. If needed, you can add your own metrics to this object based on the values of existing metrics and recompute the Evaluator. Otherwise, you can just use the output of the evaluator using the metrics function.

>>> timestamps = np.array(['2015-01-01', '2015-03-01', '2015-05-01', '2015-09-01'], dtype='M8[D]')
>>> rets = np.array([0.01, 0.02, np.nan, -0.015])
>>> starting_equity = 1.e6
>>> ev = compute_return_metrics(timestamps, rets, starting_equity)
>>> metrics = ev.metrics()
>>> assert(round(metrics['gmean'], 6) == 0.021061)
>>> assert(round(metrics['sharpe'], 6) == 0.599382)
>>> assert(all(metrics['returns_3yr'] == np.array([0.01, 0.02, 0, -0.015])))
pyqstrat.evaluator.compute_returns_3yr(timestamps, returns)[source]

Given an array of numpy datetimes and an array of returns, return those that are within 3 years of the last date in the datetime array

Return type

ndarray

pyqstrat.evaluator.compute_rolling_dd(timestamps, equity)[source]

Compute numpy array of rolling drawdown percentage

Parameters
  • timestamps (ndarray) – numpy array of datetime64

  • equity (ndarray) – numpy array of equity

Return type

Tuple[ndarray, ndarray]

pyqstrat.evaluator.compute_rolling_dd_3yr(timestamps, equity)[source]

Compute rolling drawdowns over the last 3 years

Return type

Tuple[ndarray, ndarray]

pyqstrat.evaluator.compute_sharpe(returns, amean, periods_per_year)[source]

Note that this does not take into risk free returns so it’s really a sharpe0, i.e. assumes risk free returns are 0

Parameters
  • returns (ndarray) – a numpy array of returns

  • amean (float) – arithmetic mean of returns

  • periods_per_year (float) – number of trading periods per year

>>> round(compute_sharpe(np.array([0.001, -0.001, 0.002]), 0.001, 252), 6)
0.050508
Return type

float

pyqstrat.evaluator.compute_sortino(returns, amean, periods_per_year)[source]

Note that this assumes target return is 0.

Parameters
  • returns (ndarray) – a numpy array of returns

  • amean (float) – arithmetic mean of returns

  • periods_per_year (float) – number of trading periods per year

>>> print(round(compute_sortino(np.array([0.001, -0.001, 0.002]), 0.001, 252), 6))
0.133631
Return type

float

pyqstrat.evaluator.compute_std(returns)[source]

Computes standard deviation of an array of returns, ignoring nans

Return type

float

pyqstrat.evaluator.display_return_metrics(metrics, float_precision=3, show=True)[source]

Creates a dataframe making it convenient to view the output of the metrics obtained using the compute_return_metrics function.

Parameters

float_precision (int) – Change if you want to display floats with more or less significant figures than the default, 3 significant figures.

Return type

DataFrame

Returns

A one row dataframe with formatted metrics.

pyqstrat.evaluator.handle_non_finite_returns(timestamps, rets, leading_non_finite_to_zeros, subsequent_non_finite_to_zeros)[source]
>>> np.set_printoptions(formatter={'float': '{: .6g}'.format})
>>> timestamps = np.arange(np.datetime64('2019-01-01'), np.datetime64('2019-01-07'))
>>> rets = np.array([np.nan, np.nan, 3, 4, np.nan, 5])
>>> handle_non_finite_returns(timestamps, rets, leading_non_finite_to_zeros = False, subsequent_non_finite_to_zeros = True)
(array(['2019-01-03', '2019-01-04', '2019-01-05', '2019-01-06'], dtype='datetime64[D]'), array([ 3,  4,  0,  5]))
>>> handle_non_finite_returns(timestamps, rets, leading_non_finite_to_zeros = True, subsequent_non_finite_to_zeros = False)
(array(['2019-01-01', '2019-01-02', '2019-01-03', '2019-01-04', '2019-01-06'], dtype='datetime64[D]'), array([ 0,  0,  3,  4,  5]))
>>> handle_non_finite_returns(timestamps, rets, leading_non_finite_to_zeros = False, subsequent_non_finite_to_zeros = False)
(array(['2019-01-01', '2019-01-02', '2019-01-03', '2019-01-04', '2019-01-06'], dtype='datetime64[D]'), array([ 0,  0,  3,  4,  5]))
>>> rets = np.array([1, 2, 3, 4, 4.5,  5])
>>> handle_non_finite_returns(timestamps, rets, leading_non_finite_to_zeros = False, subsequent_non_finite_to_zeros = True)
(array(['2019-01-01', '2019-01-02', '2019-01-03', '2019-01-04', '2019-01-05', '2019-01-06'],
    dtype='datetime64[D]'), array([ 1,  2,  3,  4,  4.5,  5]))
Return type

Tuple[ndarray, ndarray]

pyqstrat.evaluator.plot_return_metrics(metrics, title=None, disp_attribs=None, drawdown_lines=True, zero_line=True, show_date_gaps=True)[source]

Plot equity, rolling drawdowns and and a boxplot of annual returns given the output of compute_return_metrics.

Return type

Optional[Tuple[Figure, Axes]]

pyqstrat.evaluator.test_evaluator()[source]
Return type

None

pyqstrat.pyqstrat_cpp module

pyqstrat.pyqstrat_cpp.black_scholes_price(call: numpy.ndarray[bool], S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute Euroepean option price :param call: True for a call option, False for a put :type call: bool :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option price

Return type

float

pyqstrat.pyqstrat_cpp.cdf(x: numpy.ndarray[numpy.float64]) object

Cumulative density function of normal distribution :param x: random variable :type x: float

Returns

cdf of the random variable

Return type

float

pyqstrat.pyqstrat_cpp.d1(S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

d1 from Black Scholes :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1% :type q: float

Return type

float

pyqstrat.pyqstrat_cpp.d2(S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

d2 from Black Scholes :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1% :type q: float

Return type

float

pyqstrat.pyqstrat_cpp.delta(call: numpy.ndarray[bool], S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute European option delta :param call: True for a call option, False for a put :type call: bool :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option delta

Return type

float

pyqstrat.pyqstrat_cpp.gamma(S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute European option gamma. :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option gamma

Return type

float

pyqstrat.pyqstrat_cpp.implied_vol(call: numpy.ndarray[bool], price: numpy.ndarray[numpy.float64], S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute implied volatility for a European option. :param call: True for a call option, False for a put :type call: bool :param price: The option premium :type price: float :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Implied volatility. For 1% we return 0.01

Return type

float

pyqstrat.pyqstrat_cpp.rho(call: numpy.ndarray[bool], S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute European option rho. This is Black Scholes formula rho divided by 100 so we get rho per 1% change in interest rate :param call: True for a European call option, False for a put :type call: bool :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option theta

Return type

float

pyqstrat.pyqstrat_cpp.theta(call: numpy.ndarray[bool], F: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute European option theta per day. This is Black Scholes formula theta divided by 365 to give us the customary theta per day :param call: True for a call option, False for a put :type call: bool :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option theta

Return type

float

pyqstrat.pyqstrat_cpp.vega(S: float, K: float, t: float, r: float, sigma: float, q: float = 0.0) float

Compute European option vega. This is Black Scholes formula vega divided by 100 so we get rho per 1% change in interest rate :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option vega

Return type

float

pyqstrat.pyqstrat_io module

pyqstrat.pyqstrat_cpp.black_scholes_price(call: numpy.ndarray[bool], S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute Euroepean option price :param call: True for a call option, False for a put :type call: bool :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option price

Return type

float

pyqstrat.pyqstrat_cpp.cdf(x: numpy.ndarray[numpy.float64]) object

Cumulative density function of normal distribution :param x: random variable :type x: float

Returns

cdf of the random variable

Return type

float

pyqstrat.pyqstrat_cpp.d1(S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

d1 from Black Scholes :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1% :type q: float

Return type

float

pyqstrat.pyqstrat_cpp.d2(S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

d2 from Black Scholes :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1% :type q: float

Return type

float

pyqstrat.pyqstrat_cpp.delta(call: numpy.ndarray[bool], S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute European option delta :param call: True for a call option, False for a put :type call: bool :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option delta

Return type

float

pyqstrat.pyqstrat_cpp.gamma(S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute European option gamma. :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option gamma

Return type

float

pyqstrat.pyqstrat_cpp.implied_vol(call: numpy.ndarray[bool], price: numpy.ndarray[numpy.float64], S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute implied volatility for a European option. :param call: True for a call option, False for a put :type call: bool :param price: The option premium :type price: float :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Implied volatility. For 1% we return 0.01

Return type

float

pyqstrat.pyqstrat_cpp.rho(call: numpy.ndarray[bool], S: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute European option rho. This is Black Scholes formula rho divided by 100 so we get rho per 1% change in interest rate :param call: True for a European call option, False for a put :type call: bool :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option theta

Return type

float

pyqstrat.pyqstrat_cpp.theta(call: numpy.ndarray[bool], F: numpy.ndarray[numpy.float64], K: numpy.ndarray[numpy.float64], t: numpy.ndarray[numpy.float64], r: numpy.ndarray[numpy.float64], sigma: numpy.ndarray[numpy.float64], q: numpy.ndarray[numpy.float64] = 0.0) object

Compute European option theta per day. This is Black Scholes formula theta divided by 365 to give us the customary theta per day :param call: True for a call option, False for a put :type call: bool :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option theta

Return type

float

pyqstrat.pyqstrat_cpp.vega(S: float, K: float, t: float, r: float, sigma: float, q: float = 0.0) float

Compute European option vega. This is Black Scholes formula vega divided by 100 so we get rho per 1% change in interest rate :param S: Spot price. For a future discount the future price using exp(-rt) :type S: float :param K: Strike :type K: float :param t: Time to maturity in years :type t: float :param r: Continuously compounded interest rate. Use 0.01 for 1% :type r: float :param sigma: Annualized volatility. Use 0.01 for 1% :type sigma: float :param q: Annualized dividend yield. Use 0.01 for 1%. Default 0 :type q: float, optional

Returns

Option vega

Return type

float

pyqstrat.trade_bars module

class pyqstrat.trade_bars.TradeBars(timestamps, c, o=None, h=None, l=None, v=None, vwap=None)[source]

Bases: object

Used to store OHLCV bars. You must at least supply timestamps and close prices. All other fields are optional.

timestamp

A numpy datetime array with the datetime for each bar. Must be monotonically increasing.

c

A numpy float array with close prices for the bar.

o

A numpy float array with open prices . Default None

h

A numpy float array with high prices. Default None

l

A numpy float array with high prices. Default None

v

A numpy integer array with volume for the bar. Default None

vwap

A numpy float array with the volume weighted average price for the bar. Default None

__init__(timestamps, c, o=None, h=None, l=None, v=None, vwap=None)[source]

Zeroes in o, h, l, c are set to nan

add_timestamps(timestamps)[source]

Adds new timestamps to a market data object.

Parameters

timestamps (np.array of np.datetime64) – New timestamps to add. Does not have to be sorted or unique

>>> timestamps = np.array(['2018-01-05', '2018-01-09', '2018-01-10'], dtype = 'M8[ns]')
>>> c = np.array([8.1, 8.2, 8.3])
>>> o = np.array([9, 10, 11])
>>> trade_bar = TradeBars(timestamps, c, o)
>>> new_timestamps = np.array(['2018-01-07', '2018-01-09'], dtype = 'M8[ns]')
>>> trade_bar.add_timestamps(new_timestamps)
>>> print(trade_bar.timestamps)
['2018-01-05T00:00:00.000000000' '2018-01-07T00:00:00.000000000'
 '2018-01-09T00:00:00.000000000' '2018-01-10T00:00:00.000000000']
>>> np.set_printoptions(formatter = {'float': lambda x: f'{x:.4f}'})  # After numpy 1.13 positive floats don't have a leading space for sign
>>> print(trade_bar.o, trade_bar.c)
[9.0000 nan 10.0000 11.0000] [8.1000 nan 8.2000 8.3000]
Return type

None

describe(warn_std=10, time_distribution_frequency='15 min', print_time_distribution=False)[source]

Describe the bars. Shows an overview, errors and warnings for the bar data. This is a good function to use before running any backtests on a set of bar data.

Parameters
  • warn_std (int) – See warning function

  • time_distribution_frequency (str) – See time_distribution function

  • print_time_distribution (bool) – Whether to print the time distribution in addition to plotting it.

Return type

None

df(start_date=None, end_date=None)[source]
Return type

DataFrame

errors(display=True)[source]

Returns a dataframe indicating any highs that are lower than opens, closes, lows or lows that are higher than other columns Also includes any ohlcv values that are negative

Return type

Optional[DataFrame]

freq_str()[source]
Return type

str

has_ohlc()[source]

Returns True if we have all ohlc columns and none are empty

Return type

bool

overview(display=True)[source]

Returns a dataframe showing basic information about the data, including count, number and percent missing, min, max

Parameters

display (bool) – Whether to print out the warning dataframe as well as returning it

Return type

DataFrame

plot(figsize=(15, 8), date_range=None, sampling_frequency=None, title='Price / Volume')[source]

Plot a candlestick or line plot depending on whether we have ohlc data or just close prices

Parameters
  • figsize (Tuple[int, int]) – Size of the figure (default (15,8))

  • date_range (Union[Tuple[str, str], Tuple[Optional[datetime64], Optional[datetime64]], None]) – A tuple of strings or numpy datetimes for plotting a smaller sample of the data, e.g. (“2018-01-01”, “2018-01-06”)

  • sampling_frequency (Optional[str]) – Downsample before plotting. See pandas frequency strings for possible values.

  • title (str) – Title of the graph, default “Price / Volume”

Return type

None

resample(sampling_frequency)[source]

Downsample the trade bars data into a new bar frequency

Parameters

sampling_frequency (str) – See sampling frequency in pandas

Return type

Optional[TradeBars]

time_distribution(frequency='15 minutes', display=True, plot=True, figsize=None)[source]

Return a dataframe with the time distribution of the bars

Parameters
  • frequency (str) – The width of each bin (default “15 minutes”). You can use hours or days as well.

  • display (bool) – Whether to display the data in addition to returning it.

  • plot (bool) – Whether to plot the data in addition to returning it.

  • figsize (Optional[Tuple[int, int]]) – If plot is set, optional figure size for the plot (default (20,8))

Return type

DataFrame

valid_row(i)[source]

Return True if the row with index i has no nans in it.

Return type

bool

warnings(warn_std=10, display=True)[source]

Returns a dataframe indicating any values where the bar over bar change is more than warn_std standard deviations.

Parameters
  • warn_std (int) – Number of standard deviations to use as a threshold (default 10)

  • display (bool) – Whether to print out the warning dataframe as well as returning it

Return type

DataFrame

pyqstrat.trade_bars.roll_futures(fut_prices, date_func, condition_func, expiries=None, return_full_df=False)[source]

Construct a continuous futures dataframe with one row per datetime given rolling logic

Parameters
  • fut_prices (DataFrame) – A dataframe containing the columns ‘date’, ‘series’, and any other market data, for example, ohlcv data. Date can contain time for sub-daily bars. The series column must contain a different string name for each futures series, e.g. SEP2018, DEC2018, etc.

  • date_func (Callable[[DataFrame], ndarray]) – A function that takes the future prices as an input and returns a numpy array of booleans True indicates that the future should be rolled on this date if the condition specified in condition_func is met. This function can assume that we have all the columns in the original market data object plus the same columns suffixed with _next for the potential series to roll over to.

  • condition_func (Callable[[DataFrame], ndarray]) – A function that takes the future prices as input and returns a numpy array of booleans. True indicates that we should try to roll the future at that row.

  • expiries (Optional[DataFrame]) – An optional dataframe with 2 columns, ‘series’ and ‘expiry’. This should have one row per future series indicating that future’s expiry date. If you don’t pass this in, the function will assume that the expiry column is present in the original dataframe.

  • return_full_df (bool) – If set, will return the datframe without removing extra timestamps so you can use your own logic for rolling, including the _next columns and the roll flag

Return type

DataFrame

Returns

A pandas DataFrame with one row per date, which contains the columns in the original md DataFrame and the same columns suffixed with _next

representing the series we want to roll to. There is also a column called roll_flag which is set to True whenever the date and roll condition functions are met.

>>> fut_prices = pd.DataFrame({'timestamp': np.concatenate((np.arange(np.datetime64('2018-03-11'), np.datetime64('2018-03-16')),
...                                           np.arange(np.datetime64('2018-03-11'), np.datetime64('2018-03-16')))),
...                   'c': [10, 10.1, 10.2, 10.3, 10.4] + [10.35, 10.45, 10.55, 10.65, 10.75],
...                   'v': [200, 200, 150, 100, 100] + [100, 50, 200, 250, 300],
...                   'series': ['MAR2018'] * 5 + ['JUN2018'] * 5})[['timestamp','series', 'c', 'v']]
>>> expiries = pd.Series(np.array(['2018-03-15', '2018-06-15'], dtype = 'M8[D]'), index = ['MAR2018', 'JUN2018'], name = "expiry")
>>> date_func = lambda fut_prices: fut_prices.expiry - fut_prices.timestamp <= np.timedelta64(3, 'D')
>>> condition_func = lambda fut_prices: fut_prices.v_next > fut_prices.v
>>> df = roll_futures(fut_prices, date_func, condition_func, expiries)
>>> assert np.max(df[df.series == 'MAR2018'].timestamp.values == np.datetime64('2018-03-14'))
>>> assert(np.max(df[df.series == 'JUN2018'].timestamp.values) == np.datetime64('2018-03-15'))
pyqstrat.trade_bars.sort_trade_bars(columns)[source]

Given a list of column names, sort them in olhcv order

Return type

Sequence[str]

pyqstrat.trade_bars.test_trade_bars()[source]
Return type

None

pyqstrat.interactive_plot module

class pyqstrat.interactive_plot.InteractivePlot(data, labels=None, transform_func=<pyqstrat.interactive_plot.SimpleTransform object>, create_selection_widgets_func=<function create_selection_dropdowns>, dim_filter_func=<function simple_dimension_filter>, data_filter_func=<function simple_data_filter>, stat_func=<pyqstrat.interactive_plot.MeanWithCI object>, plot_func=<pyqstrat.interactive_plot.LineGraphWithDetailDisplay object>, display_form_func=<function display_form>, debug=False)[source]

Bases: object

Creates a multidimensional interactive plot off a dataframe.

__init__(data, labels=None, transform_func=<pyqstrat.interactive_plot.SimpleTransform object>, create_selection_widgets_func=<function create_selection_dropdowns>, dim_filter_func=<function simple_dimension_filter>, data_filter_func=<function simple_data_filter>, stat_func=<pyqstrat.interactive_plot.MeanWithCI object>, plot_func=<pyqstrat.interactive_plot.LineGraphWithDetailDisplay object>, display_form_func=<function display_form>, debug=False)[source]
Parameters
  • data (DataFrame) – The pandas dataframe to use for plotting

  • labels (Optional[Dict[str, str]]) – A dict where column names from the dataframe are mapped to user friendly labels. For any column names not found as keys in this dict, we use the column name as the label. Default None

  • dim_filter_func (Callable[[DataFrame, str, List[Tuple[str, Any]]], ndarray]) – A function that generates the values of a dimension based on other dimensions. For example, if the user chooses “Put Option” in a put/call dropdown, the valid strikes could change in a Strike dropdown that follows. Default simple_dimension_filter

  • data_filter_func (Callable[[DataFrame, List[Tuple[str, Any]]], DataFrame]) – A function that filters the data to plot. For example, if the user chooses “Put Option” in a put/call dropdown, we could filter the dataframe to only include put options. Default simple_data_filter

  • stat_func (Callable[[DataFrame, str, str, str], List[Tuple[str, DataFrame, Dict[Any, DataFrame]]]]) – Once we have filtered the data, we may need to plot some statistics, such as mean and confidence intervals. In this function, we compute these statistics. Default MeanWithCI()

  • plot_func (Callable[[str, str, List[Tuple[str, DataFrame, Dict[Any, DataFrame]]]], List[Widget]]) – A function that plots the data. This could also display detail data used to compute the statistics associated with each data point.

  • display_form_func (Callable[[Sequence[Widget], bool], None]) – A function that displays the form given a list of plotly widgets (including the graph widget)

  • debug – Dont clear forms if this is true so we can see print output

create_pivot(xcol, ycol, zcol, dimensions)[source]

Create the initial pivot :type xcol: str :param xcol: Column name to use as the x axis in the DataFrame :type ycol: str :param ycol: Column name to use as the y axis in the DataFrame :type zcol: str :param zcol: Column name to use for z-values. Each zvalue can be used for a different trace within this plot. For example, a column

called “option_type” could contain the values “American”, “European”, “Bermudan” and we could plot the data for each type in a separate trace

Parameters

dimensions (Dict[str, Any]) – The column names used for filter dimensions. For example, we may want to filter by days to expiration and put/call The key the column name and the value is the initial value for that column. For example, in a dropdown for Put/Call we may want “Put” to be the initial value set in the dropdown. Set to None if you don’t care what initial value is chosen.

Return type

None

update(owner_idx=-1)[source]

Redraw the form using the values of all widgets above and including the one with index owner_idx. If owner_idx is -1, we redraw everything.

Return type

None

class pyqstrat.interactive_plot.LineConfig(color=None, thickness=nan, secondary_y=False, marker_mode='lines+markers', show_detail=True)[source]

Bases: object

__init__(color=None, thickness=nan, secondary_y=False, marker_mode='lines+markers', show_detail=True)
color: Optional[str] = None
marker_mode: str = 'lines+markers'
secondary_y: bool = False
show_detail: bool = True
thickness: float = nan
class pyqstrat.interactive_plot.LineGraphWithDetailDisplay(display_detail_func=<pyqstrat.interactive_plot.SimpleDetailTable object>, line_configs={}, title=None, hovertemplate=None, debug=False)[source]

Bases: object

Draws line graphs and also includes a detail pane. When you click on a point on the line graph, the detail pane shows the data used to compute that point.

__call__(xaxis_title, yaxis_title, line_data)[source]

Draw the plot and also set it up so if you click on a point, we display the data used to compute that point. :type line_data: List[Tuple[str, DataFrame, Dict[Any, DataFrame]]] :param line_data: The zvalue, plot data, and detail data for each line to draw. The plot data must have

x as the first column and y as the second column

Return:

A list of widgets to draw. In this case, a figure widget and a output widget which contains the detail display

Return type

List[Widget]

__init__(display_detail_func=<pyqstrat.interactive_plot.SimpleDetailTable object>, line_configs={}, title=None, hovertemplate=None, debug=False)[source]
Parameters
  • display_detail_func (Callable[[Widget, DataFrame, bool], None]) – A function that displays the data on the detail pane. Default SimpleDetailTable

  • line_configs (Dict[str, LineConfig]) – Configuration of each line. The key in this dict is the zvalue for that line. Default {}

  • title (Optional[str]) – Title of the graph. Default None

  • hovertemplate (Optional[str]) – What to display when we hover over a point on the graph. See plotly hovertemplate

class pyqstrat.interactive_plot.MeanWithCI(mean_func=<function nanmean>, ci_level=0)[source]

Bases: object

Computes mean (or median) and optionally confidence intervals for plotting

__call__(filtered_data, xcol, ycol, zcol)[source]

For each unique value of x and z, compute mean (and optionally ci) of y. :rtype: List[Tuple[str, DataFrame, Dict[Any, DataFrame]]] :returns: x, y data for plotting lines of the mean of y versus x for each z and the data used to compute the mean

__init__(mean_func=<function nanmean>, ci_level=0)[source]
Parameters
  • mean – The function to compute ci for

  • ci_level (int) – Set to 0 for no confidence intervals, or the level you want. For example, set to 95 to compute a 95% confidence interval. Default 0

class pyqstrat.interactive_plot.SimpleDetailTable(colnames=None, float_format='{:.4g}', min_rows=100, copy_to_clipboard=True)[source]

Bases: object

Displays a pandas DataFrame under a plot that contains the data used to compute a statistic of y for each x, y pair

__call__(detail_widget, data, debug=False)[source]
Parameters
  • detail_widget (Widget) – The widget to display the data in

  • data (DataFrame) – The dataframe to display

Return type

None

__init__(colnames=None, float_format='{:.4g}', min_rows=100, copy_to_clipboard=True)[source]
Parameters
  • colnames (Optional[List[str]]) – List of column names to display. If None we display all columns. Default None

  • float_format (str) – Format for each floating point column. Default {:.4g}

  • min_rows (int) – Do not truncate the display of the table before this many rows. Default 100

  • copy_to_clipboard (bool) – If set, we copy the dataframe to the clipboard. On linux, you must install xclip for this to work

class pyqstrat.interactive_plot.SimpleTransform(transforms=None)[source]

Bases: object

Initial transformation of data. For example, you might add columns that are quantiles of other columns

__call__(data)[source]

Call self as a function.

Return type

DataFrame

__init__(transforms=None)[source]
class pyqstrat.interactive_plot.TestInteractivePlot(methodName='runTest')[source]

Bases: TestCase

test_interactive_plot()[source]
transform(data)[source]
Return type

DataFrame

pyqstrat.interactive_plot.create_selection_dropdowns(dims, labels, update_form_func)[source]

Create a list of selection widgets

Return type

Dict[str, Any]

pyqstrat.interactive_plot.display_form(form_widgets, debug=False)[source]
Return type

None

pyqstrat.interactive_plot.foo(name, old, new)[source]
pyqstrat.interactive_plot.on_widgets_updated(change, update_form_func, selection_widgets)[source]

Callback called by plotly when widgets are updated by the user.

Return type

None

pyqstrat.interactive_plot.percentile_buckets(a, n=10)[source]
>>> np.random.seed(0)
>>> a = np.random.uniform(size=10000)
>>> assert np.allclose(np.unique(percentile_buckets(a)), np.arange(0.05, 1, 0.1), atol=0.01)
Return type

ndarray

pyqstrat.interactive_plot.simple_data_filter(data, selected_values)[source]

Filters a dataframe based on the selected values

Return type

DataFrame

pyqstrat.interactive_plot.simple_dimension_filter(data, dim_name, selected_values)[source]

Produces a list to put into a dropdown for selecting a dimension value

Return type

ndarray

pyqstrat.markets module

class pyqstrat.markets.EminiFuture[source]

Bases: object

calendar = <pyqstrat.holiday_calendars.Calendar object>
static get_current_symbol(curr_date)[source]
>>> assert(EminiFuture.get_current_symbol(datetime.date(2019, 3, 14)) == 'ESH9')
>>> assert(EminiFuture.get_current_symbol(datetime.date(2019, 3, 15)) == 'ESM9')
>>> assert(EminiFuture.get_current_symbol(datetime.date(2020, 3, 14)) == 'ESH0')
Return type

str

static get_expiry(fut_symbol)[source]
>>> assert(EminiFuture.get_expiry('ESH8') == np.datetime64('2018-03-16T08:30'))
Return type

datetime64

static get_next_symbol(curr_future_symbol)[source]
>>> assert(EminiFuture.get_next_symbol('ESZ8') == 'ESH9')
Return type

str

static get_previous_symbol(curr_future_symbol)[source]
>>> assert(EminiFuture.get_previous_symbol('ESH9') == 'ESZ8')
Return type

str

class pyqstrat.markets.EminiOption[source]

Bases: object

calendar = <pyqstrat.holiday_calendars.Calendar object>
static decode_symbol(name)[source]
>>> EminiOption.decode_symbol('E1AF8')
(MO, 2018, 1, 1)
Return type

Tuple[weekday, int, int, int]

static get_expiry(symbol)[source]
>>> EminiOption.get_expiry('EW2Z5')
numpy.datetime64('2015-12-11T15:00')
>>> EminiOption.get_expiry('E3AF7')
numpy.datetime64('2017-01-17T15:00')
>>> EminiOption.get_expiry('EWF0')
numpy.datetime64('2020-01-31T15:00')
Return type

datetime64

pyqstrat.markets.future_code_to_month(future_code)[source]

Given a future code such as “X”, return the month abbreviation, such as “nov”

Parameters

future_code (str) – the one letter future code

>>> future_code_to_month('X')
'nov'
Return type

str

pyqstrat.markets.future_code_to_month_number(future_code)[source]

Given a future code such as “X”, return the month number (from 1 - 12)

Parameters

future_code (str) – the one letter future code

>>> future_code_to_month_number('X')
11
Return type

int

pyqstrat.markets.get_future_code(month)[source]

Given a month number such as 3 for March, return the future code for it, e.g. H >>> get_future_code(3) ‘H’

Return type

str

Module contents