HyperBacktest

An incredibly overengineered (and blazingly fast!) backtesting framework with a neat UI.

Screenshots:

Run:

First, you gotta create a virtual environment and install the dependencies

python3 -m venv .venv && source ./.venv/bin/activate && pip3 install -r requirements.txt

Then, you can run the UI on the port you desire:

python3 app.py 8080

You can also run the automatic backtester in the background:

python3 det_analyzer.py Currently it is focused on fetching past trades of hyperliquid wallets to determine copying which wallets could feasibly be profitable if done on your exchange(by default it uses Bybit, but this can be modified). The top wallets' data with associated original ratios is available in hl_trader_ratios.csv, for which the code for calculation will be released a bit later.

How it works:

First, openbooka.py can fetch past trades of:

• Bybit copy traders
• Binance copy traders
• OKX traders and copy traders
• Hyperliquid wallets
• Discord (FreqAI models' signals in their Discord server currently, but can probably be tweaked to be used for other servers - in any case, you will have to fill your own authorization key)
• Feather file (a table with columns open_time, close_time, ticker, leverage, margin) Then, it uses the functionality in downloader.py to efficiently download past data by filling the gaps in data required for a full backtest. The default timeframe for backtests is 1m. This high precision is crucial if highly leveraged trades are backtested, so better safe than sorry.

Replace the FLARESOLVERR_PATH variable with the path of your flaresolverr instance at the very top of openbooka.py. Take note that you will also need to replace headers/cookies for some of the websites in the code.

Afterwards, a vectorized (and thus blazingly fast) backtester based on polars framework in backtester.py is used to backtest the traders on the desired timeframe. The vectorized architecture allows it to do backtests involving millions of datapoints in mere minutes, whereas polars allows for using power of more than one core to speed it up further. After the backtest is done, it calculates a variety of different ratios:

• Sharpe
• Calmar
• Drawdown Capture Ratio
• Information Ratio
• Gittins Index
• Calmar Divergence (Calmar of the strategy - Calmar of the benchmark)

Based on these ratios it calculates some overall metrics:

• Performance
• Rating (based on both Performance and OfficialR, which is the original metric provided from the hl_trader_ratios.csv file, with the latter having much smaller weight than the former)

(Just as a note, while these metrics provide a pretty good grasp of how a trader performs, some manual analysis of trades needs to be conducted)

After all above is done, it saves the trader data to ./my_data/data/caches/caches.csv, which allows the results to be viewed within the UI.

P.S. This code is an adaptation from a larger project, so some features are missing. This is not to be used for production purposes(this part of the project has a much higher focus on development speed than code readability), so it is rather to be used just as an inspiration for some design choices, perhaps when it comes to backtest simulation or data downloading(the script here was also adopted for my EE). I am also not responsible for any investment decisions taken due to usage of this tool.