Kelly Criterion Explained for Retail Traders
Each faint trail is one simulated future for this strategy. The bold line is the median — half of runs finished above it, half below. The shaded band spans the lucky 10th to unlucky 90th percentile, so an outcome inside the band was reasonable to expect; outside, less so.
- ✓ The Kelly Criterion is a formula for the mathematically optimal bet size given edge and odds
- ✓ Full Kelly maximises long-run growth but produces severe drawdowns most traders cannot tolerate
- ✓ Half Kelly (0.5× the Kelly output) is the practical standard — 75% of the growth with half the variance
- ✓ Kelly requires accurate estimates of win rate and reward:risk — errors in inputs cause over-betting
- ✓ Use it as an upper bound for position sizing, not a precise prescription
traders the most important position sizing formula ever derived. The Kelly Criterion later popularised by gamblers, then by traders including Ed Thorp — calculates exactly how much of your capital to risk on each bet to maximise the long-run growth rate of your account.
It is also, in its pure form, too aggressive for most retail traders to use directly. Understanding why is as important as understanding the formula itself.
The Kelly formula
For a binary outcome (win a fixed amount or lose a fixed amount), Kelly is:
Where: b = net odds received (your reward:risk ratio), p = probability of winning (win rate as decimal), q = probability of losing (1 − p).
For trading, this simplifies to:
Where W = win rate and R = reward:risk ratio. Example: 55% win rate, 1.5:1 reward:risk: f* = 0.55 − (0.45 / 1.5) = 0.55 − 0.30 = 0.25 Kelly says to risk 25% of your capital per trade. That is not a typo. At this edge, the mathematically growth-maximising position size is a quarter of your account per trade.
Why full Kelly is impractical
The Kelly Criterion maximises long-run geometric growth — but the path to that growth involves drawdowns that most humans cannot tolerate in practice.
At full Kelly with the above example (25% per trade):
5 consecutive losses = 76% account drawdown
10 consecutive losses = 94% drawdown (effectively zero)
Even though recovery is mathematically guaranteed given enough trades, the psychological experience of a 76% drawdown causes virtually all traders to abandon the strategy
Kelly also assumes exact knowledge of your edge. If your true win rate is 52% but you believe it's 55%, you are overbetting. Overbetting relative to true Kelly has asymmetric consequences — it destroys capital faster than underbetting. And in trading, unlike in a casino, the true odds are never precisely known.
Half Kelly: the practical standard
The professional standard in quantitative trading and sports betting is Half Kelly: bet half the Kelly-optimal fraction. This produces approximately 75% of the maximum geometric growth rate while cutting variance (and maximum drawdowns) roughly in half.
| Kelly Fraction | Growth Rate (% of max) | Typical Max Drawdown |
|---|---|---|
| Full Kelly (1.0×) | 100% | Severe (40–80%) |
| Half Kelly (0.5×) | ~75% | Moderate (20–40%) |
| Quarter Kelly (0.25×) | ~56% | Manageable (10–20%) |
| Fixed 1% risk | Varies by edge | Usually <15% |
For a retail trader with the 55%/1.5R example: full Kelly = 25%, half Kelly = 12.5%, quarter Kelly = 6.25%. Most professional guidance for retail traders recommends staying between quarter Kelly and half Kelly — typically in the 2–5% risk-per-trade range for most strategies.
Kelly as an upper bound, not a prescription
The most useful way to think about Kelly for retail trading is as a ceiling. If Kelly says risk 20% per trade, the answer is not to risk 20% — it is to understand that your strategy has real edge, and to choose a risk fraction that is comfortably below Kelly based on your psychological risk tolerance and the uncertainty in your edge estimates.
If Kelly says risk 1% per trade, that is a warning: your edge is thin, and even small errors in your win rate or R:R estimate could push your true Kelly fraction to zero (meaning the strategy has no edge worth betting on).
Run your parameters through the [Expectancy Calculator →] to see your edge clearly before applying Kelly. A negative expectancy strategy has a negative Kelly fraction — meaning the mathematically optimal bet size is zero.
FAQ
What if Kelly gives me a very large number like 30%?
This means your edge estimate is either very strong or very optimistic. Use it as a ceiling and apply quarter Kelly or less — the larger the Kelly fraction, the more important it is to be conservative, because estimation errors in edge have larger consequences at higher bet sizes.
Does Kelly work for all trading instruments?
The formula applies to any strategy with a measurable win rate and average win/loss ratio. The main limitation is that real trades are not perfectly binary — partial closes, slippage, and variable R:R ratios make the true Kelly fraction harder to calculate precisely. Use it as an approximation, not a precise figure.
Is Kelly the same as fixed fractional position sizing?
Fixed fractional sizing (risking a fixed % of balance per trade) IS Kelly if the fixed % equals the Kelly output. In practice, most traders use fixed fractional with a conservative fraction rather than computing Kelly explicitly.
What is overbetting and why is it dangerous?
Overbetting means risking more than the Kelly-optimal fraction. It reduces long-run growth rate and increases drawdowns. Severe overbetting (>2× Kelly) produces negative long-run growth even for strategies with positive expectancy — the losses on bad sequences outweigh the gains on good ones.