Backtester

Backtester is a Python application for testing trading strategies against historical data. Backtester is heavily reliant on the Backtrader framework.

Architecture

The package structure is shown below:

The class structure is shown below:

The diagrams have been generated using pyreverse:

pyreverse -o png .

Data

The data used for the shown results is OHLCV with the following format:

date	open	high	low	close	volume	ticker
2000-01-21	554.73761	554.73761	508.50955	554.73761	26	ZNT

This corresponds to the following line in config.properties:

cols=Date,Open,High,Low,Close,Volume,Ticker

Data was obtained from Premium Data at 04/05/2018. The data covers 2,628 stocks listed on the ASX between 1992-05-29 and 2018-05-04.

By default the sample .csv files in the data subdirectory will be used when the application is run. To use a wider set of data update the path variable in config.properties to point to a folder with additional data.

The following data format is also supported:

Date,Low,Open,Volume,High,Close,Adjusted Close

Results

The table below summarises the results for the strategies.

Strategy	Asset	Start Date	End Date	Positions	Position Size	CAGR	Sharpe Ratio	Skew	Kurtosis	Maximum Drawdown	Maximum Monthly Return	Recovery Factor
Crossover	ASX	29/05/1992	04/05/2018	4	25%	4.55%	0.33	2.83	73.55	-41.13%	24.91%	5.27
Crossover	ASX	29/05/1992	04/05/2018	10	10%	2.42%	0.21	9.65	510.96	-53.72%	61.62%	1.59
CrossoverLongOnly	ASX	17/04/2015	02/09/2022	4	25%	-2.82%	-0.51	-0.78	17.18	-24.37%	4.98%	-0.78
CrossoverPlus	ASX	29/05/1992	04/05/2018	4	25%	5.94%	0.35	1.22	32.38	-62.78%	46.81%	5.53
CrossoverPlus	ASX	29/05/1992	04/05/2018	10	10%	3.26%	0.27	2.23	96.23	-64.27%	25.9%	2.02
Pump	ASX	29/05/1992	04/05/2018	4	25%	6.31%	0.42	0.6	120.1	-24.64%	15.05%	15.79
Pump	ASX	29/05/1992	04/05/2018	10	10%	5.58%	0.63	-0.43	94.91	-13.29%	-10.28%	23.26
HolyGrail	ASX	29/05/1992	04/05/2018	4	25%	1.38%	0.17	0.29	9.26	-43.14%	39.44%	0.99
HolyGrail	ASX	29/05/1992	04/05/2018	10	10%	0.31%	0.08	0.02	9.19	-30.73%	21.73%	0.27
Benchmark	XJO	29/05/1992	04/05/2018	1	100%	5.04%	0.4	-0.34	5.23	-53.93%	8.18%	4.78

The terms are defined:

• Strategy: The name of the strategy to be tested. Corresponds to a strategy implementation in the strategies folder.
• Asset: The universe of assets.
• Start Date: The earliest date in the dataset.
• End Date: The latest date in the dataset.
• Positions: The number of positions that can be held by the strategy.
• Position Size: The percentage of the total portfolio to be used when entering a position.
• CAGR: Compound Annual Growth Rate.
• Sharpe Ratio: A measure of the excess returns compared to the volatility. Ratios above 1 are generally considered good.
• Skew: A measure of the distortion of symmetrical distribution or asymmetry in the results. A positive skew indicates the tail of a distribution curve is longer on the right side, and vice-versa for a negative skew.
• Kurtosis: A measure of extreme values in either tail. A high kurtosis implies an occasional extreme return.
• Maximum Drawdown: The maximum observed loss from peak to trough of a portfolio.
• Maximum Monthly Return: The largest return in any month.
• Recovery Factor: The absolute value of the returns divided by the Maximum Drawdown. Indicates a strategies ability to overcome a drawdown.

Assumptions

The following assumptions or decisions have been made:

1. The brokerage commission fee structure is as per Nabtrade. This is assumed to have been available historically.
2. There is only 1 open position per asset.
3. Orders from trading do not impact the market price.
4. No rebalancing is done for the positions. Entry and exit conditions are defined in the individual strategies.

Strategies

Sample results for each strategy are shown below. Where results for multiple strategies are shown above, the graphs below are for runs with 4 long positions of 25% the portfolio value each.

Crossover

An implementation of a 50-day and 200-day moving average crossover strategy provided in Crossover. This strategy will buy the individual equity when long and sell it when short.

CrossoverPlus

An implementation of a strategy using 50-day and 200-day moving averages, the Percent Price Oscillator (PPO) indicator, and the Relative Strength Index (RSI) indicator is available in CrossoverPlus.

CrossoverLongOnly

An implementation of a 50-day and 200-day moving average crossover strategy provided in CrossoverLongOnly. This implementation is long only, as instead of selling it buys an inversely correlated equity. This strategy configured to buy VAS when long and buy BBOZ when short. All cash available to the broker is used in each trade.

Pump

An implementation of a strategy attempting to detect "pump and dump" schemes based on large volume, small price changes, and local price maxima is available in Pump. This implementation is long only.

HolyGrail

An implementation of a strategy attempting to detect "pump and dump" schemes based on large volume, small price changes, and local price maxima is available in HolyGrail. This implementation is long only. This strategy when run for all data (2,377 stocks with 250 discarded due to insufficient data quantity for the indicators) returns a CAGR of 1.42% with 2,171 trades. This underperforms the XJO benchmark for this period by 3.6%.

Benchmark

The benchmark strategy buys and holds the asset in data/benchmark (XJO in the examples above). This strategy is available in Benchmark.

Getting Started

Pre-requisites

The following pre-requisites are required:

1. Anaconda 22.9.0.

Once these are installed:

1. Open the Anaconda Prompt.
2. Run where conda and note the folder paths (excluding the file name component).
3. Add these folder paths to the PATH environment variable.
4. Open a new Command Prompt.
5. Run conda --version to verify that it is accessible.
6. Python should automatically be added to the PATH but run python --version to confirm.

Installation

To install backtester:

1. Create the virtual environment:

   conda env create -f environment.yml

2. Activate the virtual environment:

   conda activate backtester

3. Configure your IDE to use the new environment.

To uninstall backtester:

1. Remove the virtual environment:

   conda env remove -n backtester

Updating Dependencies

1. Update the conda packages to the latest compatible versions:

   conda update --all

2. Create an updated environment.yml based on the current environment:

   conda env export > environment.yml

Run

1. To run the backtester:

   backtester.py -strategy $strategy -verbose True

Configuration

The config.properties file contains configuration for the backtester application. A description of each property is shown below.

Property	Description	Type
data/path	The folder directory containing OLHCV data.	Str
data/cols	The column headers for the OLHCV data.	Str
data/constituents	The name of the file containing ASX300 data as at 22/10/21. This data is used if global_options/small_cap_only is set to True.	Str
data/benchmark	The name of the comparison benchmark.	Str
data/bulk	True if all tickers are to be processed, and False if they will be set individually in data/tickers.	Bool
data/tickers	A comma separated list of tickers to be tested.	Str
data/tickers_for_exclusion	A comma separated list of tickers to be excluded from testing.	Str
data/start_date	The overriding start date (%d/%m/%Y) of the strategy.	Str
data/end_date	The overriding end date (%d/%m/%Y) of the strategy.	Str
data/date_format	The data format of the OLHCV data.	Str
global_options/strategy	The strategy to be tested.	Str
global_options/position_limit	The maximum number of positions that can be held. Must be consistent with global_options/position_size.	Int
global_options/position_size	The size of each position. Must be consistent with global_options/position_limit.	Int
global_options/plot_enabled	True if a plot is to be created, and False otherwise.	Bool
global_options/plot_tickers	True if the individual tickers are to be plotted, and False otherwise. The global_options/plot_enabled property should also be set to True if this is set to True.	Bool
global_options/plot_volume	True if the individual ticker plots are to include volume, and False otherwise.	Bool
global_options/plot_benchmark	True if the benchmark plot is to be created, and False otherwise.	Bool
global_options/reports	True if a quantstats report is to be created for the strategy, and False otherwise.	Bool
global_options/vectorised	True if the backtrader vectorised option is to be used, and False otherwise.	Bool
global_options/cheat_on_close	True if the backtrader cheat on close functionality is to be used, and False otherwise. If True, you will eliminate slippage for orders at the cost of less realistic results.	Bool
global_options/small_cap_only	True if small cap stocks (as defined by not being a member of data/constituents) are to be used exclusively, and False otherwise.	Bool
global_options/no_penny_stocks	True if penny stocks (close < $1) are to be excluded, and False otherwise.	Bool
global_options/use_adjusted_close	True if the adjusted close column in the OLHCV data is to be used instead of close, and False otherwise.	Bool
crossover_strategy_options/crossover_strategy_sma1	The period of the fast moving average.	Int
crossover_strategy_options/crossover_strategy_sma2	The period of the slow moving average.	Int
crossover_strategy_long_only_options/crossover_strategy_long_only_sma1	The period of the fast moving average.	Int
crossover_strategy_long_only_options/crossover_strategy_long_only_sma2	The period of the slow moving average.	Int
crossover_plus_strategy_options/crossover_plus_strategy_sma1	The period of the fast moving average.	Int
crossover_plus_strategy_options/crossover_plus_strategy_sma2	The period of the slow moving average.	Int
crossover_plus_strategy_options/RSI_crossover_high	The upper bound for the RSI.	Int
crossover_plus_strategy_options/RSI_crossover_low	The lower bound for the RSI.	Int
crossover_plus_strategy_options/RSI_period	The period for the RSI.	Int
crossover_plus_strategy_options/optimise	True if multiple versions of the strategy are to be run, and False otherwise.	Bool
crossover_plus_strategy_options/sma1_low	The lower bound of sma1 for optimisation.	Int
crossover_plus_strategy_options/sma1_high	The upper bound of sma1 for optimisation.	Int
crossover_plus_strategy_options/sma1_step	The step size of sma1 for optimisation.	Int
crossover_plus_strategy_options/sma2_low	The lower bound of sma2 for optimisation.	Int
crossover_plus_strategy_options/sma2_high	The upper bound of sma2 for optimisation.	Int
crossover_plus_strategy_options/sma2_step	The step size of sma2 for optimisation.	Int
crossover_plus_strategy_options/RSI_crossover_low_low	The lower bound of RSI_crossover_low for optimisation.	Int
crossover_plus_strategy_options/RSI_crossover_low_high	The upper bound of RSI_crossover_low for optimisation.	Int
crossover_plus_strategy_options/RSI_crossover_low_step	The step size of RSI_crossover_low for optimisation.	Int
crossover_plus_strategy_options/RSI_crossover_high_low	The lower bound of RSI_crossover_high for optimisation.	Int
crossover_plus_strategy_options/RSI_crossover_high_high	The upper bound of RSI_crossover_high for optimisation.	Int
crossover_plus_strategy_options/RSI_period_low	The lower bound of RSI_period for optimisation.	Int
crossover_plus_strategy_options/RSI_period_high	The upper bound of RSI_period for optimisation.	Int
crossover_plus_strategy_options/RSI_period_step	The step size of RSI_period for optimisation.	Int
holygrail_strategy_options/adx_period	The period used to calculate the ADX.	Int
holygrail_strategy_options/ema_long_period	The period of the long EMA.	Int
holygrail_strategy_options/ema_short_period	The period of the short EMA.	Int
holygrail_strategy_options/lag_days	The number of days we can wait after being tagged for a position but not hitting the entry condition. If the days go over, we abandon the possibility for a trade.	Int
holygrail_strategy_options/bounce_off_min	The bounce from min we require as part of tagging a potential future position.	Float
holygrail_strategy_options/bounce_off_max	The bounce from max we require as part of tagging a potential future position.	Float
holygrail_strategy_options/volume_period	The period used to calculate the average volume.	Int
holygrail_strategy_options/minimum_volume	The minimum volume to consider a position in a particular ticker.	Int
pump_strategy_options/volume_factor	The volume factor that must be exceeded to potentially open a long position.	Float
pump_strategy_options/buy_timeout	The number of days required between closing and opening a position in the same ticker.	Int
pump_strategy_options/sell_timeout	The number of days before we close the position.	Int
pump_strategy_options/price_comparison_lower_bound	The lower bound for the price comparison to potentially open a long position.	Float
pump_strategy_options/price_comparison_upper_bound	The upper bound for the price comparison to potentially open a long position.	Float
pump_strategy_options/price_max_period	The period for the maximum price indicator.	Int
pump_strategy_options/volume_average_period	The period for the average volume indicator.	Int
pump_strategy_options/profit_factor	The profit factor required before closing a position.	Float
broker/cash	The total cash available for trading.	Float