Different backtest results between v0.3.3 and v0.6.4 even with identical code

[iimka]

I previously used version 0.3.3 and recently updated to 0.6.4. Even with identical code, the backtest results differ between the two versions – including for simple strategies like buy-and-hold. The code is as follows:

from backtesting import Backtest, Strategy
from backtesting.lib import crossover

from backtesting.test import SMA, GOOG

class BuyAndHold (Strategy):
    def init(self):
        pass
    def next(self):
        if (not self.position.is_long
        ):
            self.buy()

bt = Backtest(GOOG, BuyAndHold,
              cash=10000, commission=.002,
              exclusive_orders=True)

output = bt.run()
bt.plot()

print(output)

​​Results from 0.3.3:​​

​​Results from 0.6.4:​​

When testing other strategies, the trades, outcomes, and charts also show discrepancies between versions. I want to understand if there are configuration changes, and which version’s results should be considered more reliable for actual reference? Personally, I prefer the 0.3.3 results as I’ve consistently used them for trading signals. I’m concerned future updates might make it impossible to continue relying on the 0.3.3 outcomes.

[kernc]

Several core changes, as listed in #1221 (comment). And the change log!

In your case, it's finalize_trades=False as the new default, because that was confusing users.