By Keel Research Team · Updated May 12, 2026
Optimal bet size from win probability and win/loss ratio, with fractional-Kelly safety guards. The growth-maximizing bet — and why most traders bet half or quarter Kelly instead of the full math.
From your strategy backtest. Example: 55 = 55% of trades close profitable.
Magnitude only — enter as a positive number.
Full Kelly maximizes growth but is psychologically punishing. Half-Kelly is the common practical default.
The Kelly criterion answers: given a known edge, what fraction of bankroll bet maximizes long-run logarithmic growth?
f* = (b × p − q) / b
p = probability of winning
q = 1 − p
b = avg_win / avg_loss (positive)The result f* is your bet size as a fraction of bankroll for each trade. Compounding at f* gives the highest expected long-run wealth — but the path there is extreme. Full Kelly drawdowns can hit 50%+ even when the underlying edge is positive; the variance of returns scales with f.
Why half or quarter Kelly: Kelly assumes you know p and b exactly. In real trading, you estimate them from limited samples. Overestimating edge by 20% leads full Kelly to ruin; half-Kelly is much more forgiving. Practical compromise: capture 75% of full-Kelly growth at 25% of full-Kelly variance.
Use this calculator with backtested inputs — win rate + average win + average loss from a Keel backtest are the right numbers to feed in, not gut-feel estimates.
Keel is a Strategy OS for AI-assisted systematic trading on Hyperliquid. Build, backtest, and run live strategies with realistic fees, slippage, and funding modeled. Free to start — connect a Hyperliquid wallet when you’re ready to go live.
The fraction of your bankroll that maximizes long-run logarithmic growth, given a known win probability and win/loss ratio. Formula: f* = (b × p − q) / b, where p is win prob, q = 1−p, and b is avg_win / avg_loss. Full Kelly is mathematically growth-optimal but in practice almost always overbet relative to estimation error in p and b.
Kelly assumes you know the true win probability and win/loss ratio. In trading, both are estimated with significant error. If you overestimate edge by even 10%, full Kelly leads to ruinous drawdowns. Half-Kelly captures most of the growth at much lower drawdown variance; quarter-Kelly is even safer. Most professional traders bet between quarter and half Kelly.
A negative result means the edge is negative — the strategy loses money in expectation. The Kelly math is suggesting you bet on the opposite side, but in practice it means: don't take this trade. Re-examine your win-rate and win/loss-ratio estimates before risking capital.
Fixed-percent risk (e.g. "always risk 1%") is bankroll-relative but doesn't account for the edge of the trade. Kelly is edge-aware — bigger bets when edge is bigger. In practice, Kelly is best used as a ceiling: never risk more than half-Kelly, and use fixed-percent for trades where you can't estimate edge cleanly.
Backtest the strategy on Keel. The backtest engine returns win rate, average win, average loss, total trades, and standard error — exactly the inputs Kelly needs. Run the same strategy across multiple time periods to estimate parameter stability before committing to live capital at the Kelly-suggested size.
Risk-based sizing — fixed % of account per trade. Complementary to Kelly.
How much your strategy could lose peak-to-trough. Sets the floor for Kelly-style aggression.
Documented strategies with real backtest numbers — find the win-rate + win/loss inputs Kelly needs.