I added trend-following bots to my fleet to decorrelate it. The correlation matrix said it worked. The P&L said it didn't. Here's the number that settles the argument. I run a small fleet of trading bots in paper mode. Most of them are variations on the same idea — mean-reversion on crypto: buy the dip, scale in, exit in green. Running five bots that all do the same thing isn't a fleet, it's one bet with extra logging. So I deliberately added bots designed to make money when the dip-buye...