calculation of `Buy & Hold Return`

[zillionare]

this is how Buy & Hold Return is calculated:

        c = data.Close.values
        s.loc['Buy & Hold Return [%]'] = (c[-1] - c[0]) / c[0] * 100  # long-only return

so it's calced use day one and the day last.

Expected Behavior

Buy & Hold Return is used for compare with strategy gain. Therefore, I guess they should started at same time, since the strategy get enough data to work on.

Take cross ma strategy example, use 5 days for short line and 10 days for long line, the strategy can only works from 10th trade day. It's better to make comparision since this day

• Backtesting version: 0.3

[eervin123]

One thing that is nice about the backtesting.py framework is its generic nature. I have a number of strategies and many of them rely on the entire period for a representative benchmark.

Having said that, it would be nice if it were easier to specify a "warming period" for both the strategy and the benchmark. I haven't come up with a clean way to do that as of yet, though I haven't really dug into it.

As a feature, it would be kind of nice if there were a way to specify a warming_period as a number of candles to ignore for the plot and stats and thus the benchmark, etc.

eg.

bt = Backtest(df, smacross, warming_period=10)

Full disclosure, I've been away for a bit so this might have been implemented in a different fashion. I'm not sure.

[kernc]

a way to specify a warming_period as a number of candles to ignore for the plot and stats

Done automatically like this for backtest and plots:

backtesting.py/backtesting/backtesting.py

Lines 1146 to 1149 in 4f21a65

	# Skip first few candles where indicators are still "warming up"
	# +1 to have at least two entries available
	start = 1 + max((np.isnan(indicator.astype(float)).argmin(axis=-1).max()
	for _, indicator in indicator_attrs), default=0)

Therefore, you can extend the warming period by containing an indicator with sufficient leading np.nan values.

I like that it's done automatically. We should extended it more consistently into stats (Buy&Hold computation) — PR welcome.

[flexelem]

I am also having a similar issue because of non configurable warming up period explained in #477

[Orchid420]

Has the feature "warming_period" been added to make sure that the strategy tested and the buy and hold strategy are calculated on the same timeframe?