[pzhon]

The number of other players matters a lot.

Let's suppose you are considering a short cut, something that will cost you 100,000 hands of grinding if it fails, but save you 20,000 hands if it succeeds. If you are competing with thousands of winning players, succeeding might increase your chance of winning from 0.1% to 2%, so it would be worth it if you only succeed 1 time in 20. However, if there are only 5 competing players, success might be worth 60% while failure might cost 15%, and you would need to succeed 20% of the time for it to be worth it.

Anyway, against a large field, there will be plenty of people using the Kelly criterion. You should probably use something much more aggressive while you are in contention. You should sacrifice expected growth rate for variance that might make you first.