I once figured I can use ML to predict stock market movements and I went about buying 5 years of data from CBOE and started working on modeling. At one point I managed to get fairly accurate predictions of movements and I thought I had struck a lottery.
Turns out my training data had tomorrow’s movement indicator in it which I accidentally added. Despite that dumb mistake it was only accurate up to 80% or so. That’s when I realized Im better off with a random number generator.