Get more than one statistic when optimize

[edofe99]

When I optimize a strategy I can get only a chosen statistic to see for every combinations, for example:

stats, heatmap = bt.optimize(dc_period = range(3,30,3),close_after = range(3,30,3),maximize = 'Return [%]',return_heatmap=True)

If I print(heatmap) I get:

dc_period  close_after
3          3             0.91800
           6             1.65825
           9             1.37100
           12            1.11650
           15            2.13875
                           ...  
27         15            0.46700
           18            1.14275
           21            0.91300
           24            1.07525
           27            1.50150
Name: Return [%], Length: 81, dtype: float64

So I see the two parameters and corresponding return of that combination, but say I also want to see winrate and average trade % of every combination? How can I do that?

Thank's.

[coder145]

To the best of my knowledge, there isn't a way it can be done directly in the library, but you could always rerun the parameters you are interested in to obtain your statistic of choice.

[chrisgft]

You can run this yourself a different route and get the same results as well as the stats for every run.

from itertools import product

def run_backtest(params):
    ema_period, keltner_atr_multiplier = params
    bt = Backtest(df_training, Test, cash=CASH, commission=0.0005, margin=0.02, trade_on_close=True, hedging=True)
    stats = bt.run(
        ema_period=ema_period,
        keltner_atr_multiplier=keltner_atr_multiplier
    )
    return stats

# Define ranges for each parameter
ema_period_range = range(30, 50, 10)
keltner_atr_multiplier_range = range(3, 6, 1)

# Generate all possible combinations of parameters
params_list = list(product(ema_period_range, keltner_atr_multiplier_range))

# Use ThreadPoolExecutor to run backtests
with concurrent.futures.ThreadPoolExecutor() as executor:
    results = list(executor.map(run_backtest, params_list))

max_result = None
for result in results:
    if (max_result == None):
        max_result = result['Equity Final [$]']
    else:
        max_result = max(result['Equity Final [$]'], max_result)```

[s-kust]

This repo - https://github.com/s-kust/python-backtesting-template - may be helpful.

See Optimization of Strategy Parameters in readme.md and optimize_params.py script.

This solution is more flexible than the original bt.optimize and also provides you with an Excel file containing the results.

[mateusz-brodowicz]

I would suggest looking into Optuna (https://github.com/optuna/optuna). You can define custom objective functions, and you're not limited to scalar outputs. You can pass your backtest result inside the optimize function, and define the objective function as any combination of stats object's attributes.