The Most Dangerous Mistake You Can Make In Trading. - page 2
To actually answer the question instead of the tangent: the evidence I'd trust is a live losing streak longer than what a Monte Carlo shuffle of your own historical trades says should happen at a high confidence level, say the 95th percentile. If your backtest's worst modeled streak at your risk level is 8 losers and you're at 14, that's not a rough patch, that's outside what the strategy's own history says is plausible.
Calendar time or trade count alone isn't enough, since a slow month can still be completely normal. What I wouldn't trust is a gut feeling that something "feels different" without a number behind it, that's exactly the trigger that leads to premature tweaking and overfitting to recent price action.
Determining Expert Advisor parameters
Account and Behavioral Analysis
Introduction to MetaTrader 5