webertonmaia/weekend-backtrader

Implementation of the weekend trader strategy with backtrader

Weekend Backtrader

Implementation of the Weekend Trader strategy with Backtrader.

Strategy

From Nick Radge's book "Unholy Grails" and Chat with Traders podcast: https://chatwithtraders.com/ep-178-nick-radge/

The strategy is a modification of Turtle Trading: equity long only momentum strategy. The original strategy trades on small cap industrials: australian, outside ASX 100 but inside ASX 500.

Basic ideas:

• every trend starts with a breakout
• riding winners, cutting losers

Rules:

• only long stocks
• breakout period: 100 days (instead of 20 days)
• weekend trend trader: 20 week breakout (break 20 week high)
• confirmation filter: Rate of Change (indicator) above a level (acceleration) i.e. -> 30 (strong breakout) percentage change price on price for the last 20 weeks
• regime filter: broader market is trending up. Underlying index trends up. Buy signal if index above SMA(200).
• trailing stop loss: 20% trailing stop -> 1% total loss in ptf per trade (0.5 * 0.2 = 0.01)
• 20 pos max, 5% capital each
• when regime filter goes from uptrend to downtrend, use 10% instead of 20% stop loss

Stats reported by Nick Radge:

• avg trading loss (single position loss) 11.97% -> 0.6% of total ptf
• avg win: 21% -> 2.6 times the loss
• win rate: 44%
• win/loss ration: 2.6

Getting Started

1. (Optional) Create virtualenv or conda env

2. Install requirements

   pip install -r requirements.txt
   

3. Download data feeds:

   python download_data_feed.py 
   

   The symbols downloaded can be capped with MAX_DOWNLOADS const.

4. Execute Backtest

   python main.py
   

Caveats

• POC, possibly there is some bug
• the backtest only runs from to (see main.py's _load_datafeed)
• the feed data is not cleaned
• The data feed has survivorship bias!
• Some symbols are ignored because there is not enough data for our timeframe
• The strategy is not profitable as it is, do not use it
• Yahoo financial download limit should be 2000 req / hour. download_data_feed.py will download around 378 symbols.

Current results

============ Summary ================
Sharpe Ratio:  0.6201373437021387
Annual returns
    2015: 0.0
    2016: 0.4622394790766664
    2017: 1.8888994250547002
    2018: -0.22068818768678122
    2019: 0.22746678809392806
Total returns
    rtot: 1.3964532412290513
    ravg: 0.001103915605714665
    rnorm: 0.32073279810641175
    rnorm100: 32.073279810641175
=====================================

See Also

• https://www.allordslist.com/
• https://www.asx100list.com/