a trading book from 1997 tells you to almost never pick the most profitable parameters your optimizer finds. it's Cybernetic Trading Strategies by murray ruggiero, written when most people didn't own a machine. "you should almost never select the most profitable set of parameters. the rare exception to this rule would be if the most profitable pair is surrounded by almost equally profitable pairs." "next, you want to have the highest possible number of neighbouring parameters with similar performance." that's the whole idea. you're not looking for the peak. you're looking for the widest area where performance barely changes. then he lists what else has to hold before you accept a set: reasonable drawdowns. profit not coming from one or two trades. profits distributed evenly across the sample. an upward sloping equity curve. he wrote that nearly thirty years ago. still applies today in systematic strategy testing.