filib 0.3.0

Factor Investing Library

Factor Investing LIBrary is a lightweight algorithmic trading Python library built for easy testing of predictive factors and portfolio rebalance via Oanda. Inspired by and compatible with Quantopian Open Source.

Changelog »

Installation

Install with pip:

$ pip install filib

Usage

Proposed workflow contains three steps. Here’s an example:

1. Assemble

Begin with imports, then hypothesize and create predictive factors as class methods:

from filib.oanda import Oanda  # Currently only Oanda is available
from filib.helpers import *    # Optional functions, useful for factor generation


class MyFactors(Oanda):

    def momentum(self):  # HYPOTHESIS: there is persistence in an asset's performance
        factor = self.returns
        split = [-1, -.003, .003, 1]  # List of thresholds or int to split equally
        return factor, split  # Always return at least factor and follow this order

    def relative_strenght_index(self):  # H: signal oversold or overbought assets
        import pandas as pd
        import pandas_ta as ta  # Example Technical Analysis Library
        factor = pd.DataFrame({
            instrument: self.price_data[instrument].ta.rsi(length=14)
            for instrument, _ in self.price_data
        }) * -1  # Short low factor values and long high factor values
        split = [-100, -70, -30, 0]
        return factor, split

    def big_mac_index(self):  # H: simplified Purchasing Power Parity theory
        import quandl  # Example Financial, Economic and Alternative Data Library
        iso_codes = get_iso_codes(self.price_data)
        codes = [f'ECONOMIST/BIGMAC_{COUNTRY}.5' for COUNTRY in iso_codes]
        factor = quandl.get(codes).dropna(how='all', axis=1)
        factor.columns = [iso_codes[c.split('_')[1].split()[0]] for c in factor]
        factor.index = factor.index.tz_localize('UTC')  # Convert time zone to UTC
        return factor  # If not specified split = 3 by default

2. Research

Initialize parameters (during the first run you will be asked to provide credentials):

research = MyFactors(
    instruments = ['EUR_USD', 'GBP_USD', 'USD_JPY', 'AUD_USD', 'NZD_USD',
                   'USD_CAD', 'USD_CHF', 'USD_NOK', 'USD_SEK'],  # Define universe
    granularity = 'D',    # Time range between rebalancing and between each candle
    count = 500,          # Number of historical candles to return for analysis
    symbol = 'USD',       # Optional, specify symbol to arrange your price data
    periods = (1, 2, 3),  # Optional, specify periods to analyze factor decay
    split = 3,            # Number of quantiles to split your combined factor data
    long_short = True,    # Trade only top and bottom quantile of combined factor
    combination = 'sum_of_weights')  # Select factor combination method

Check the performance of factors combined together:

>>> research.performance()
...
MyFactors - INFO - Factor `MyFactors_combined` Analytics:

                   Min    Max   Mean   Size Returns (bps)
                factor factor factor factor            1D     2D     3D
factor_quantile
1.0             -1.131  0.009 -0.238   1299        -0.551 -1.292 -0.845
2.0             -0.206  0.331  0.004   1213        -1.411 -2.811 -3.009
3.0              0.000  1.125  0.247   1232        -0.663 -1.289 -3.189

                               1D     2D     3D
- Information Coefficient:  0.005 -0.001 -0.007
- Factor Rank Autocorrelation: 0.09

- Annualized Sharpe Ratio: -0.46
- Annualized Alpha (Beta): -0.011 (0.119)
- Win Rate: 48.32%
- Risk / Reward: 0.99
- Profit Factor: 0.92

- Start Date: 2018-08-07
- End Date: 2020-03-10
- Duration: 581 days 00:00:00 (1.6 years)
- Rebalance every: 1D

- Compound Annual Growth Rate: -1.39%
- Annualized Volatility: 2.94%
- Maximum Drawdown: -4.09%
- Maximum Drawdown Duration: 23 days 00:00:00
...

Alternatively set selection rules with a query string to evaluate. Available metrics are:

• ic: Information Coefficient based on Spearman’s Rank Correlation

• autocorr: Factor Rank Autocorrelation

• sharpe: Annualized Sharpe Ratio

• beta: Annualized Beta as exposure to trading universe

• alpha: Annualized Alpha as excess returns over trading universe

• win: Win Rate

• rr: Risk / Reward Ratio

• profit: Profit Factor = (sum of earnings) / (sum of losses)

• cagr: Compound Annual Growth Rate

Then analyze the performance of individual factors and select those that meet the rules:

>>> research.select(
...     rules = 'abs(ic) > .01 or profit > 1',  # Example query expression
...     swap = 'cagr')  # Align the signs of selected factors to specified metric
...
MyFactors - INFO - Factors with signs that meet the rules `abs(ic) > .01 or profit > 1`:

momentum                  -1.0
relative_strenght_index    1.0


MyFactors - INFO - Factor `MyFactors_selected` Analytics:

                   Min    Max   Mean   Size Returns (bps)
                factor factor factor factor            1D     2D     3D
factor_quantile
1.0             -1.000  0.026 -0.107   1815        -1.972 -3.022 -4.282
2.0             -0.152  1.000  0.095    757         2.189  3.523  3.425
3.0              0.000  0.880  0.222    551         1.257 -0.077  1.388

                               1D    2D     3D
- Information Coefficient:  0.017  0.01  0.016
- Factor Rank Autocorrelation: 0.04

- Annualized Sharpe Ratio: 0.30
- Annualized Alpha (Beta): 0.011 (0.025)
- Win Rate: 44.71%
- Risk / Reward: 0.92
- Profit Factor: 1.06

- Start Date: 2018-08-07
- End Date: 2020-03-10
- Duration: 581 days 00:00:00 (1.6 years)
- Rebalance every: 1D

- Compound Annual Growth Rate: 1.02%
- Annualized Volatility: 3.61%
- Maximum Drawdown: -4.34%
- Maximum Drawdown Duration: 371 days 00:00:00
...

3. Trade

Execute or schedule a script to rebalance your portfolio based on selected factors:

PLEASE USE AT YOUR OWN RISK - THIS CAN TRADE REAL MONEY - NO WARRANTY IS GIVEN

# strategy.py
from filib.oanda import Oanda


class SelectedFactors(Oanda):

    def momentum(self):
        factor = self.returns * -1.0  # Sign from the research
        split = [-1, -.003, .003, 1]
        return factor, split

    def relative_strenght_index(self):
        import pandas as pd
        import pandas_ta as ta
        factor = pd.DataFrame({
            instrument: self.price_data[instrument].ta.rsi(length=14)
            for instrument, _ in self.price_data
        }) * -1
        split = [-100, -70, -30, 0]
        return factor, split


if __name__ == "__main__":

    strategy = SelectedFactors(
        instruments = ['EUR_USD', 'GBP_USD', 'USD_JPY', 'AUD_USD', 'NZD_USD',
                       'USD_CAD', 'USD_CHF', 'USD_NOK', 'USD_SEK'],
        granularity = 'D',
        count = 250,
        symbol = 'USD',
        split = 3,
        accountID = '',  # Your Oanda's account ID for creating orders
        leverage = 7,    # Set the leverage for the portfolio positions
        long_short = True,
        combination = 'sum_of_weights')

    strategy.rebalance(live=True)  # Actually place orders

Check portfolio positions and generated orders in a log file or by dry run:

>>> strategy.rebalance()
SelectedFactors - INFO - Portfolio from `2020-03-11 00:00:00+00:00`:

CHF    -38.4%
SEK     -8.1%
EUR     -3.5%
GBP      0.0%
NOK      0.0%
NZD      0.0%
AUD     13.9%
CAD     15.7%
JPY     20.4%

- Account NAV: 10000.00 EUR
- Position Value: 0.00
- Needed Orders:

USD_CHF    30410
USD_SEK     6430
EUR_USD    -2430
AUD_USD    16860
USD_CAD   -12450
USD_JPY   -16140

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.