Is it possible to optimize based on multiple variables?

[Gladi-iwnl]

Is it possible to optimize based on multiple variables? Or set the constraint to a value from the results?

Example:

stats = backtest.optimize(
    n1=range(10, 110, 10),
    n2=range(20, 210, 20),
    constraint=stats['Max. Drawdown [%]'] < 20,
    maximize='Return [%]]')

[kernc]

Or set the constraint to a value from the results?

Constraint is a function that takes parameter combination (e.g. specific values of n1 and n2) and either marks it admissible (acceptable, valid) or not. It operates on combinations of parameters (optimization variables) way before we obtain the backtest results. So, no.

I think what you want is maximize=, which can also be a callable that takes results and returns a value, the greater the better. Like:

def maximize(stats):
    if stats['Max. Drawdown [%]'] < -20:
        return -np.inf  # i.e. worst
    return stats['Return [%]']

stats = bt.optimize(..., maximize=maximize)

[Gladi-iwnl]

So if I use this maximize function, I would only get returns where the Max. Drawdown is higher than -30?

def maximize(stats):
    if stats['Max. Drawdown [%]'] > -30:
        return stats['Return [%]']

stats, heatmap = bt.optimize(
    p1=range(1,6,1),
    p2=range(1,11,1),
    p3=range(3,20,1),
    maximize=maximize,
    return_heatmap=True)

print(heatmap.sort_values().iloc[-20:])

Also, when I try to run this optimization and sort the heatmap I get a TypeError:

  File "/run/media/gladi/B204D33B04D300F1/Work/backtestingTryout/MAShiftSl.py", line 64, in <module>
    print(heatmap.sort_values().iloc[-20:])
  File "/home/gladi/.local/lib/python3.9/site-packages/pandas/core/series.py", line 3260, in sort_values
    sorted_index = nargsort(values_to_sort, kind, ascending, na_position)
  File "/home/gladi/.local/lib/python3.9/site-packages/pandas/core/sorting.py", line 379, in nargsort
    indexer = non_nan_idx[non_nans.argsort(kind=kind)]
TypeError: '<' not supported between instances of 'float' and 'str'

[kernc]

def maximize(stats):
    if stats['Max. Drawdown [%]'] > -30:
        return stats['Return [%]']

No, with maximize function as above, you have an else branch you're not handling. The later TypeError is probably a consequence of implicitly returned None value. Inspect heatmap to confirm.