Because I am a betting game player, I was surprised by the task given in the question, so I calculated the reward rate of the training data set and found that the reward rate is greater than 1. This is not in line with reality. In reality, the reward rate of betting companies is less than 1.upd:Sorry I got the formula wrong. In fact, the reward rate of the training data set is around 0.99. This is how the rebate rate is calculated. If the home team win odds are x, the draw odds are y, the away team wins the odds is z, and the return rate is 1/(1/x+1/y+1/z).