Change an entry filter, then change an exit rule. The second backtest looks better. Which change deserves the credit?
That's one of the problems I'm working on in XTester Studio. A result needs to stay attached to the version that produced it and the question being tested. Otherwise, the next experiment starts with guesswork.
Here's how I'd set up that comparison: keep the original strategy as a baseline, test the entry filter on its own, then test the exit rule from that same baseline. Keep the historical period and trading-cost assumptions the same. Combining both changes can be a separate experiment.
There's a less obvious complication in this work: the simulator can change too. A recent fix in Studio corrected slippage on market orders placed at minute boundaries. If an older test used that behavior, rerun the baseline before comparing it with a new strategy version. A different result may come from the execution model.
Before keeping a new rule, I'd want to answer one question: did I change the strategy, the test conditions, or both?