I understand that random() is generally not supported, but what about random seed? Namely, my backtests keep getting different results when I use different seeds (my algo resets the seed multiple times). It works fine when I just use one seed.
Should I be expecting this kind of behavior?
Edit: answered my own question. Solution: use RandomState().
Re-edit: solution does not work.
Code in question:
if seed is not None:
np.subtract(returns, np.nanmean(returns), out=returns)
np.divide(returns, np.sqrt(np.nanvar(returns)+1e-12),
out=returns)
np.random.seed(seed)
nan_count = np.sum(np.isnan(returns))
returns[np.isnan(returns)] = np.random.randn(nan_count)
else:
returns[np.isnan(returns)] = 0
