Home › FX Trading School › Advanced
The Kelly Criterion in Depth: Sizing Bets Without Blowing Up
The Kelly criterion is a formula for working out how much of your capital to risk on a trade so that your account grows as fast as mathematically possible over the long run — and understanding it properly means understanding exactly why almost nobody should use it at full strength. This is Module 19 of the PipTax FX Trading School, and it builds directly on Module 18's fixed-fractional risk sizing and Module 14's expectancy calculations — if those concepts are hazy, it's worth revisiting them before this one, because Kelly is really just expectancy pushed to its logical, aggressive extreme.
What the Kelly Criterion Actually Calculates
The Kelly formula answers one narrow question: given a repeatable edge, what fraction of capital maximises the geometric growth rate of your account? Not the average profit per trade — the compounded growth rate, which is what actually matters over hundreds of trades.
The classic formula for a simple win/loss bet is:
f\* = W − [(1 − W) / R]
Where:
- f\* = fraction of capital to risk
- W = probability of winning (your historical win rate as a decimal)
- R = win/loss ratio (average winning trade size ÷ average losing trade size)
Example: if you win 45% of trades and your average win is 1.8x your average loss, then:
f\* = 0.45 − (0.55 / 1.8) = 0.45 − 0.306 = 0.144, or roughly 14.4% of capital per trade.
That number should already look alarming to anyone used to the standard 1–2% risk guidance taught earlier in this course — and that gap is the entire subject of this lesson.
Why Full Kelly Blows Up Real Accounts
Kelly's maths is correct, but it depends on two assumptions that rarely hold for retail FX traders:
- You know W and R precisely. In reality, these are estimated from a finite sample of trades, often fewer than 100. A small sample overstates edge more often than it understates it — survivorship and recency bias both push win rates up in your own memory.
- Your edge is stationary. Kelly assumes the same probability and payoff repeat indefinitely. Market regimes change; a strategy's win rate on trending EUR/USD conditions is not the same as its win rate in a choppy, range-bound month.
Even with a genuinely accurate edge, full Kelly produces enormous volatility of growth. It's a mathematical fact that a full-Kelly strategy will, at some point, suffer a drawdown of 50% or more — that's not a sign something's wrong, it's what the optimal-growth path looks like. Very few traders can psychologically or financially survive that, and most will abandon the system exactly at the bottom, converting a temporary drawdown into a permanent loss.
Overestimate your edge even slightly — say your real win rate is 40%, not 45% — and the "optimal" fraction becomes a fast route to ruin.
Fractional Kelly: The Practical Compromise
Because full Kelly is too aggressive for anyone trading with real, imperfect estimates, most practitioners use fractional Kelly — deliberately risking a fraction of the calculated f\*.
| Kelly fraction | Approx. risk in example above | Growth trade-off | |---|---|---| | Full Kelly (1.0×) | ~14.4% per trade | Maximum growth, maximum volatility | | Half Kelly (0.5×) | ~7.2% per trade | ~75% of max growth, roughly half the volatility | | Quarter Kelly (0.25×) | ~3.6% per trade | ~44% of max growth, much smoother equity curve |
The half-Kelly result is the classic finding worth remembering: cutting your bet size in half sacrifices only around a quarter of the theoretical growth rate but removes a large chunk of the drawdown risk. That asymmetry is why quarter- to half-Kelly is the common professional compromise, not full Kelly.
For most retail FX accounts, even quarter-Kelly numbers from a small sample will look large next to the 0.5–1% fixed-risk guidance taught in Module 18. Treat any Kelly output as an upper ceiling on position size, not a target to reach.
Building Trading Costs Into the Kelly Inputs
The Kelly formula doesn't know about spread, commission, or swap — you have to bake those into W and R yourself, or the output will be optimistic.
- Spread and commission reduce every winning trade slightly and add to every losing trade's cost, which pulls R down.
- Swap/rollover matters for anything held overnight and compounds over time in carry-heavy pairs.
- Slippage on stops during volatile releases effectively increases your average loss size beyond the theoretical stop distance.
Two brokers with different cost structures — say Pepperstone on a raw-spread MetaTrader account with commission, versus IG's own platform with a spread-inclusive pricing model — will produce genuinely different R values for the exact same strategy, because the cost drag differs. Before trusting any Kelly-derived position size, run your actual trade costs through the [cost tool](/audit.html) and check current broker terms on the [brokers page](/brokers/index.html) — the fraction that looked safe on paper can shrink once realistic costs are subtracted from R.
A Practical Kelly-Aware Workflow
Rather than calculating f\* and trading it directly, use Kelly as a diagnostic check inside your existing risk process:
1. Log at least 50–100 real or backtested trades for a specific strategy and pair — small samples give unreliable W and R. 2. Calculate R net of realistic costs, using live spread/commission figures rather than assumed round numbers. 3. Compute full Kelly f\* as a reference ceiling, not a target. 4. Apply a quarter- or half-Kelly multiplier, and compare that number against your existing fixed-fractional risk rule. 5. Use whichever is smaller. If quarter-Kelly still exceeds your normal 1% risk rule, that's a signal your edge estimate is probably too optimistic, not a licence to size up. 6. Re-check periodically. As your trade sample grows or market conditions shift, W and R will drift — recalculate rather than fixing a Kelly fraction forever.
This keeps the discipline of fixed-fractional sizing from Module 18 while using Kelly as an evidence-based sanity check on whether you're under- or over-risking relative to a measured edge.
Kelly's Limits and Where It Fits in a Portfolio
The Kelly criterion was built for a single, repeated, independent bet — real FX trading rarely matches that cleanly:
- Correlated positions (e.g. several USD-pair trades open at once) mean your true portfolio risk is higher than the sum of individually-Kelly-sized trades suggests.
- Multi-strategy portfolios need a portfolio-level Kelly extension, which is materially more complex and sensitive to estimated correlations between strategies.
- Regime dependency means a Kelly fraction calculated in a trending market can be badly wrong applied to a ranging one.
None of this makes the Kelly criterion useless — it's a genuinely useful lens for understanding why oversized positions destroy accounts even with a real edge, and why professional sizing is conservative by design, not by accident. Used as a ceiling check alongside fixed-fractional risk, cost-adjusted expectancy, and realistic sample sizes, it adds real rigour to Module 19's capital allocation topic. Used at full strength on optimistic inputs, it's one of the more reliable ways to turn a working strategy into a blown account. For the full course sequence and prerequisite modules, see the [FX Trading School index](/school/index.html), and for methodology on how PipTax calculates cost impact on strategies like this, see the [methodology page](/methodology.html).
Key takeaways
- The Kelly criterion gives the position size that maximises long-run geometric growth of capital, not the size that maximises profit per trade.
- Full Kelly requires an accurate win rate and payoff ratio — inputs traders almost always overestimate, which is why full Kelly is dangerous in practice.
- Fractional Kelly (e.g. quarter or half Kelly) trades some growth for a large cut in volatility and drawdown, and is what most professional sizing models actually use.
- Kelly sizing sits on top of a fixed-risk framework: you still need stop-loss placement and cost-aware position sizing to know your true risk per trade.
- Spreads, commissions and swaps eat into the payoff ratio in the Kelly formula, so cost differences between brokers change your optimal size — check them with the cost tool before assuming a Kelly fraction is safe.
- Kelly assumes independent, stationary edge; regime shifts and correlated trades break that assumption, so treat any Kelly output as a ceiling, not a target.
Frequently asked questions
- What is the Kelly criterion in simple terms?
- It's a formula that calculates the fraction of your capital to risk on a bet so that, over many repeated bets with the same edge, your capital grows fastest in the long run. It needs two inputs: your win rate and your average win-to-loss ratio.
- Is the Kelly criterion safe to use for forex trading?
- Full Kelly is not safe for most retail traders because it assumes you know your true win rate and payoff ratio precisely, and any overestimate leads to oversized positions and severe drawdowns. Fractional Kelly (a quarter or half of the calculated size) is the more realistic, safer approach.
- How do you calculate Kelly percentage for a trading strategy?
- The basic formula is f* = W − [(1 − W) / R], where W is your historical win rate (as a decimal) and R is your average win size divided by your average loss size. The result, f*, is the fraction of capital to risk — but only if your sample size is large and representative.
- Why is full Kelly considered dangerous?
- Full Kelly maximises growth but also maximises volatility of that growth — drawdowns of 50% or more are mathematically normal along a full-Kelly path even with a genuine edge. Any error in your win-rate or payoff estimate (which is common with small trade samples) makes the real risk of ruin far higher than the model suggests.
- Does the Kelly criterion account for trading costs like spread and commission?
- Not directly — you have to build costs into your payoff ratio yourself. Spread, commission and swap reduce your average win and slightly increase your effective loss, so recalculating R with realistic costs (via a tool like PipTax's cost tool) will lower your Kelly fraction versus a cost-free estimate.
- What's a practical alternative to full Kelly for position sizing?
- Most practical traders use fixed fractional risk (e.g. risking 0.5–1% of capital per trade) combined with a Kelly calculation used only as a sanity check or upper bound, often applying a quarter- or half-Kelly multiplier for extra safety margin.