Kelly fraction = (win-rate × avg-win-multiple − loss-rate) / avg-win-multiple. For a 60% win-rate strategy with 1.5R winners and 1R losers, Kelly says size at (0.6×1.5 − 0.4) / 1.5 ≈ 33% of account per trade.
That number sounds absurd because it is. Kelly assumes you know your edge precisely. In real trading, your win rate and average R are estimates with wide confidence intervals; sizing at full Kelly on a wrong estimate ruins you quickly.
The practical use is fractional Kelly: size at ½-Kelly (cuts drawdown ~50%, only loses 25% of expected growth) or ¼-Kelly (very conservative, durable, common among professional shops).
Kelly is sensitive to inputs. A win-rate estimate off by 5 percentage points or an average-R estimate off by 0.2 changes the prescribed size by a factor of 2+. This is why most journals use a simpler 'fixed fractional' rule (e.g. risk 1% per trade) — it's robust to estimation error in a way Kelly isn't.