Kelly Criterion for Traders: How Much to Risk Per Trade

Most traders pick a risk percentage and stick with it — 1%, 2%, whatever feels right. And that's fine as a starting point. But there's a mathematical framework that tells you the theoretically optimal amount to risk given your actual edge.

It's called the Kelly Criterion. Developed by John Kelly at Bell Labs in 1956, it was originally designed for telephone signal noise problems. Gamblers and investors figured out pretty quickly that it applied to them too.

What the Kelly Criterion Is

The Kelly Criterion answers one question: given a known edge, what fraction of your capital should you bet to maximize long-term growth?

Bet too little and you're leaving money on the table. Bet too much and variance destroys you even when you have an edge. Kelly finds the sweet spot.

A Worked Example

Let's say your trading system has the following stats over the last 100 trades:

• Win rate: 45% (W = 0.45)
• Average win: $300
• Average loss: $150
• Win/loss ratio: $300 ÷ $150 = 2.0 (R = 2.0)

Kelly calculation:

K% = 0.45 − (1 − 0.45) / 2.0
K% = 0.45 − 0.55 / 2.0
K% = 0.45 − 0.275
K% = 0.175 = 17.5%

Kelly says to risk 17.5% of your capital per trade.

If your account is $25,000, that's $4,375 of risk on a single trade.

That should make you uncomfortable. Good — it should. Which leads us to the most important lesson about Kelly.

Why You Should Never Use Full Kelly

Full Kelly maximizes long-run geometric growth. It also produces gut-wrenching drawdowns in the short run.

At full Kelly, expected drawdowns of 30-50% are completely normal — even with a positive edge. The math guarantees growth eventually, but "eventually" might mean sitting through a 40% drawdown for six months first.

Almost no human being can handle that psychologically. When your account is down 40%, you abandon the system — which is exactly the worst time to do it.

There's also a bigger problem: your edge estimate is almost certainly imprecise. Kelly is brutally sensitive to errors in your win rate and win/loss ratio estimates. If you think your win rate is 45% but it's actually 40%, full Kelly might be telling you to bet twice what you should.

The Half Kelly Solution

The practical answer that most professional gamblers and fund managers use is Half Kelly: take whatever Kelly tells you and cut it in half.

In our example: 17.5% ÷ 2 = 8.75%

Half Kelly gives you roughly 75% of the growth rate of full Kelly, but with dramatically smaller drawdowns. It's also much more forgiving of estimation errors in your edge.

For most retail traders, I'd go further and cap it at 2-3% regardless of what Kelly says. Here's why:

The position sizing guide covers why 1-2% risk per trade is the standard answer for most retail accounts — and why violating that ceiling is what blows up most traders, regardless of what Kelly suggests.

When Kelly Says Don't Trade

The Kelly formula produces a negative number when there's no edge. That's not an error — it's Kelly telling you this trade will destroy capital over time regardless of how you size it.

Example: 40% win rate, 1:1 R:R

K% = 0.40 − (0.60 / 1.0) = 0.40 − 0.60 = −0.20

Negative Kelly = negative edge = do not trade this system.

This is one of the most useful things Kelly tells you. Before asking "how much should I bet?", ask "does Kelly even give me a positive number?" If it doesn't, work on the strategy first. (And if you're not sure why a 40% win rate at 1:1 is a losing system, see why 1:1 R:R is a losing game.)

What Kelly Actually Requires

For Kelly to work, you need reliable estimates of two inputs:

1. Win rate — You need at least 50-100 trades of data before your win rate estimate is meaningful. Early in a strategy's life, the sample size is too small. A 55% win rate over 20 trades might be luck; over 200 trades, it's edge.

2. Win/loss ratio — Your average win divided by your average loss. Note: this is actual averages, not planned R:R. If you plan for 2:1 but cut winners early and let losers run, your actual ratio might be 1.2:1.

Most traders don't have clean enough statistics to use Kelly properly. That's not a reason to not use it — it's a reason to start tracking your trades meticulously right now. The win rate vs risk/reward guide covers exactly which stats to track and why expectancy (a related metric) is worth calculating in parallel.

Kelly in Practice: A Realistic Framework

Here's how I'd actually apply this:

• New strategy: Use 1% risk until you have 50+ trades of data
• Proven strategy: Calculate Kelly, use Quarter Kelly, cap at 2%
• High-conviction setup (rare): Consider up to 3% if Kelly supports it
• Negative Kelly: Don't trade the system — it has no edge

The value of Kelly isn't necessarily the exact number it spits out. It's the framework it forces you into: quantifying your edge, tracking your results, and making position sizing a function of math rather than gut feel.

Kelly vs Fixed Fractional

Fixed fractional (always risk X% per trade) is simpler and arguably more practical for retail traders. The advantages of Kelly are theoretical — they show up over thousands of trades with stable edges.

The real value of the Kelly formula is what it teaches you:

• Your edge must be quantified, not assumed
• There is a mathematically correct amount to bet — and it's determined by win rate and win/loss ratio, not by how you feel about the trade
• Negative Kelly = no edge = don't trade it

Use it as a sanity check and a framework for thinking. Don't use it as a precise instruction to bet 17% of your account.