To install on your system, use pip:
pip install qfinuwa
To pull market data ensure you have a text file with the API key and call API.fetch_stocks:
from qfinuwa import API
path_to_API = 'API_key.txt'
download_folder = './data'
API.fetch_stocks(['AAPL', 'GOOG', 'TSLA'], path_to_API, download_folder)A multi-indicator takes in a single signal (price of an arbitary stock) and outputs a transformation of that stock.
It is called MultiIndicator because the indicator will have multiple values (one for each stock)
# Example
class CustomIndicators(Indicators):
@Indicators.MultiIndicator
def bollinger_bands(self, stock: pd.DataFrame):
BOLLINGER_WIDTH = 2
WINDOW_SIZE = 100
mid_price = (stock['high'] + stock['low']) / 2
rolling_mid = mid_price.rolling(WINDOW_SIZE).mean()
rolling_std = mid_price.rolling(WINDOW_SIZE).std()
return {"upper_bollinger": rolling_mid + BOLLINGER_WIDTH*rolling_std,
"lower_bollinger": rolling_mid - BOLLINGER_WIDTH*rolling_std}Similar to MultiIndicator, SingleIndicator is implemented as a function that takes in stock data and returns an indicator or indicators.
It is called SingleIndicator because there is only a single signal.
# Example
class CustomIndicators(Indicators):
@Indicators.SingleIndicator
def etf(self, stock: dict):
apple = 0.2
tsla = 0.5
goog = 0.3
return {'etf': apple*stock['AAPL'] + tsla*stock['TSLA'] + goog*stock['GOOG']}You can manually test you indicators as follows:
stock_a = pd.from_csv('stockA.csv')
stock_b = pd.from_csv('stockA.csv')
# multi-indicator for stockA (returns dict of dict of pd.Series)
output_a = CustomIndicators.bollinger(stockA)
# multi-indicator for stockB (returns dict of dict of pd.Series)
output_b = CustomIndicators.bollinger(stockA)
# single-indicator for stockA + stockB (returns dict of pd.Series)
output = CustomIndicators.etf({'stockA': stock_a, 'stockB': stock_b})Each function you implement can be thought of as a hyperparameter "group" that bundles the indicator it returns (the keys to the dictionary the indicator function returns).
The backtester can change hyperparameters for you, but to do so you need to give each one a name, in the form of kwargs.
The kwargs must include a default value which will be used if you do not specify a value.
class CustomIndicators(Indicators):
@Indicators.MultiIndicator
def bollinger_bands(self, stock: pd.DataFrame, BOLLINGER_WIDTH = 2, WINDOW_SIZE=100):
mid_price = (stock['high'] + stock['low']) / 2
rolling_mid = mid_price.rolling(WINDOW_SIZE).mean()
rolling_std = mid_price.rolling(WINDOW_SIZE).std()
return {"upper_bollinger": rolling_mid + BOLLINGER_WIDTH*rolling_std,
"lower_bollinger": rolling_mid - BOLLINGER_WIDTH*rolling_std}
@Indicators.SingleIndicator
def etf(self, stock: dict, apple = 0.2, tsla= 0.5, goog=0.3):
return {'etf': apple*stock['AAPL'] + tsla*stock['TSLA'] + goog*stock['GOOG']}To define your strategy extend qfin.Strategy to inherit its functionalities. Implement your own on_data function.
Your on_data function will be expected to take 4 positional arguments.
self: reference to this objectprices: a dictionary of numpy arrays of historical pricesportfolio: object that manages positions
Similar to qfin.Indicators, you can define hyperparameters for your model in __init__.
# Example Strategy
class BasicBollinger(Strategy):
def __init__(self, quantity=5):
self.quantity = quantity
self.n_failed_orders = 0
def on_data(self, prices, indicators, portfolio):
# If current price is below lower Bollinger Band, enter a long position
for stock in portfolio.stocks:
if(prices['close'][stock][-1] < indicators['lower_bollinger'][stock][-1]):
order_success = portfolio.order(stock, quantity=self.quantity)
if not order_success:
self.n_failed_orders += 1
# If current price is above upper Bollinger Band, enter a short position
if(prices['close'][stock][-1] > indicators['upper_bollinger'][stock][-1]):
order_success = portfolio.order(stock, quantity=-self.quantity)
if not order_success:
self.n_failed_orders += 1
def on_finish(self):
# Added to results object - access using result.on_finish
return self.n_failed_ordersAdditionally, you can specify a function on_finish that will run on the completion of a run, if you want to save your own data. Whatever this function returns will can be accessed in the results (see SingleRunResults.on_finish).
The Backtester class asks for a custom strategy, custom indicators and data from the user. Once created, it can run multiple backtests without having to recalculate the indicators - when used in a Notebook environment the backtester object can persist and incrementally updated with new values.
See qfinuwa.Backtester docstrings for specifics.
from qfinuwa import Backtester
backtester = Backtester(CustomStrategy, CustomIndicators,
data=r'\data', days=90,
delta_limits=1000, fee=0.01)backtester.indicators.update_params(dict_of_updates)backtester.indicators.paramsbacktester.indicators.defaultsbacktester.indicators = NewIndicatorClassbacktester.strategy.update_params(dict_of_updates)backtester.strategy.paramsbacktester.strategy.defaultsbacktester.strategy = NewStrategyClass
