StructureX

Home › Comparisons

Kelly criterion vs fixed fractional: what if your win rate is wrong?

Fixed fractional risks a small, set percent you choose, like 1%. The Kelly criterion works the percent out from your win rate and payoff, and it is usually huge. Use a small fixed percent. If you use Kelly at all, use a fraction of it, because your win rate is a guess.

Comparison · 6-minute read · By Paspir · Updated 2026-11-16

The short answer

Both methods risk a percent of your current balance on each trade. The difference is who picks the percent.

That last part is the catch. You never know your true win rate. You only have a guess from past trades. Kelly takes that guess at face value, and when the guess is too high, Kelly bets far too much.

Side by side

Fixed fractionalKelly criterion
Who picks the percentYouA formula
Typical size0.5% to 2%Often 10% or more
Needs your true win rateNoYes, exactly
If your guess is too highSmall extra damageCan wreck the account
Swings along the waySmallVery deep, even when right
Easy to useVeryEasy math, hard inputs

The formula, in one line

Kelly % = W − (1 − W) ÷ R

W is your chance of winning. R is your average win divided by your average loss. Our Kelly criterion page works through it in more detail, and the Kelly criterion calculator does the sum.

You think 60%, it's really 50% (hypothetical)

Your last few dozen trades won 60% of the time, and wins and losses were the same size (R = 1). Kelly says:

0.60 − 0.40 ÷ 1 = 20% of the account on every trade.

But the 60% was luck. Your true win rate is 50%. With R = 1, true Kelly is 0.50 − 0.50 = 0%. There's no edge at all, and the right bet is nothing.

Here are 20 trades, 10 wins and 10 losses, in this order: W L W W L L L L W L W W L W L L W W W L. Three traders start with $10,000:

After tradeFull Kelly (20%)Half Kelly (10%)Fixed 1%
4$13,824$11,979$10,200
8$5,662$7,859$9,798
12$7,828$9,415$9,994
16$4,809$7,550$9,794
20$6,648$9,044$9,990
Result−$3,352−$956−$10
Worst drop from a high65%37%4%

Check the worst drops. Full Kelly peaked at $13,824 after trade 4 and fell to $4,809 after trade 16. That's a drop of about 65%. Half Kelly went from $11,979 to $7,550, about 37%. The 1% trader went from $10,200 to $9,794, about 4%.

Notice something odd: half the trades won, yet full Kelly lost a third of the account. A 20% gain followed by a 20% loss leaves you at 96%, not 100%. Big bets lose ground even when wins and losses balance.

Even when the guess is right, it hurts

Now say your win rate really is 60%. Same order, but 12 of the 20 trades win (W L W W L L L L W L W W L W W W W L W W).

Full Kelly ends ahead, but its worst drop along the way is still about 61%. Half Kelly's is about 35%. Fixed 1% is about 4%.

Very few people can sit through a 61% drop without changing the plan halfway. That's why traders who use Kelly at all tend to use half or a quarter of it.

What a wrong guess does

Being too small costs you a little. Being too big costs you a lot. Since every win rate is a guess, it pays to lean small. That's the case for fixed fractional, and for reading a big Kelly number as "this may have an edge", not "bet this much".

Pick fixed fractional if… use a slice of Kelly if…

Pick fixed fractional if…A slice of Kelly can fit if…
You have fewer than a few hundred logged tradesYou have a long, honest record of your results
You'd quit after a 30% dropYou use a quarter or half, not full Kelly
Your setups change oftenYour setup and costs stay the same
You want one simple ruleYou cap it, say at 2%, whatever it says

Fixed fractional vs fixed dollar shows a losing streak at a fixed percent, trade by trade. The risk of ruin page shows how bet size drives the chance of losing it all.

Option sellers face the same problem: a high win rate tempts big size. The options strategy comparison chart shows the worst case of each strategy, which is where sizing starts.

Questions

Is the Kelly criterion good for trading?

The idea is useful: bet more when the edge is bigger. Full Kelly in practice is too big, because your win rate and payoff are guesses. Most traders who use it take a quarter or half.

What is half Kelly?

Half of the percent the Kelly formula gives. It gives up some growth for much smaller drops, and it's far more forgiving when your win rate guess is too high.

Why did full Kelly lose money with half the trades winning?

Big percentage swings don't cancel out. Up 20% then down 20% leaves 96% of what you had. The bigger the bet, the more ground each pair of trades loses.

How many trades do I need to trust my win rate?

More than most people think. A few dozen trades can easily show 60% when the truth is 50%. Hundreds of trades with the same setup give a much better estimate.

Follow along

XTradingViewFacebook