I don’t think this approach scales well, retry with 100K bets and I think your model would approach 0% profit.
The only way to beat the book is to predict events more accurately than them and bet where there’s margin of profit.