Kelly Criterion for Trading: Position Size on a Prop Account

The short answer

The Kelly criterion sets position size from win rate and reward-to-risk: f = (bp - q) / b. On a prop firm account, the capital in that formula is the drawdown limit, not the account balance. This makes the right size far smaller than most traders expect.

• ›Formula: f = (bp - q) / b. Here p is win rate, q = 1 - p and b is average win divided by average loss.
• ›Example: 40% win rate at 1:3 gives Kelly 20%; half Kelly on a $2,000 drawdown is $200 per trade, 0.4% of a $50K account.
• ›Professionals run half or quarter Kelly, because full Kelly's variance ends a prop account before the long run arrives.

What the Kelly criterion is

Professional gamblers and fund managers size positions with the Kelly criterion. It is a formula that converts a measured edge into a mathematically optimal bet size.

Most retail traders have not used it, and many trade at several times the optimal size without knowing it. This is one reason sound strategies still fail evaluations.

K = W − [(1 − W) ÷ R]

Here W is the win rate and R the average reward-to-risk ratio. Example: a 40% win rate at 1:3 gives K = 0.40 − (0.60 ÷ 3) = 20%. This is the theoretical maximum fraction of capital the edge justifies risking, rather than a recommended size.

Why full Kelly is too aggressive

Full Kelly maximises long-run growth, but the variance is severe. A prop account's drawdown limit ends the account long before the long run arrives.

Professionals usually run half Kelly (roughly 75% of the growth for half the swings) or quarter Kelly.

Applying it to a $50K challenge

On a prop account the real capital is the drawdown rather than the balance. On a $50K account with a $2,000 max drawdown:

• ›Kelly (40% WR, 1:3) = 20% → half Kelly = 10%
• ›10% of the $2,000 drawdown = $200 risk per trade
• ›conservative quarter Kelly = $100 per trade

That is 0.2-0.4% of the nominal account size. The popular "risk 1%" rule gives $500 per trade, which is a quarter of the entire drawdown.

Four ordinary losers in a row would end the account, and that streak is close to certain at a 40% win rate. The evaluation tests position sizing at least as much as it tests entries.

Checking the numbers against real rules

Every program's drawdown and daily-loss numbers are in the programs table. The pass simulator runs a Monte-Carlo test of a trader's exact win rate, reward-to-risk and risk-per-trade against any challenge's real rules. It also shows how much the pass probability improves when the size is halved.

The formulas assume independent trades and an accurately measured edge. Traders who do not know their win rate and average R from at least a few hundred trades should measure that first. Kelly calculations on estimated numbers give a false sense of precision.