.They are all trending systems look for trends , whichever logic finds the trend enters it first.curve fit over 10 years on 50k trades?
highly unlikely but still possible. My suggestion would be: 1. Still take the first trades in a forward test and use half position for those trades. 2. Rely on the backtested results and do not trade first trades and forget about the "why don't they work so well".
10 years = about 2500 days of trading. If the systems are correlated then the sample is much less than 50K. For the early periods of the trading day there might only be a few hundred uncorrelated signals.