Kelly gets quoted like a secret formula. For a simple win/lose bet it says to risk f = p - q/b, where p is the win probability, q is 1 - p and b is the payoff ratio. A 55% system at 1:1 gives 0.55 - 0.45 = 0.10. Ten percent of the account per trade.
Now the contrarian bit. Nobody should do that. The formula assumes the win rate and payoff are known exactly. A scalper has an estimate from a few hundred trades, with error bars wide enough to drive a truck through. If the true win rate is 52% rather than 55%, full Kelly is 4%, so betting 10% means more than double the ideal size, and over-betting costs growth, not just comfort.
In the textbook model, full Kelly also carries roughly a 50% chance of cutting the account in half at some point along the way. Half Kelly gives up some growth but trims those drawdowns a lot, which is why practitioners tend to use a fraction.
For the 55% example, half Kelly would still be 5% per trade, which sounds wild next to the 0.5% to 2% many scalpers stick with.
So Kelly works better as a ceiling than a target.
Do you use anything Kelly-like to size, or is it all fixed percentages?
full Kelly is a terrible idea for a scalper
-
Shadow Trader
- Posts: 5438
- Joined: Wed Sep 09, 2026 5:19 pm