The Kelly criterion gives the bet size that maximizes long-run compound growth. Above it you blow up; below it you leave growth on the table. Most professionals run half Kelly or less — estimation error punishes over-betting far more than under-betting.
Binary bet: f* = p − (1−p)/b, the edge divided by the odds. Continuous strategy (Gaussian approx): f* = μ/σ², expected excess return over variance — here f* is the fraction of capital as leverage (f*=2 means 2× exposure).
Half Kelly delivers ~75% of the growth at half the drawdown severity — the standard professional compromise for when μ and σ are estimates, not truths.
Kelly assumes you know the true edge. You don't — you have a backtest. If your estimated edge is 2× the real one, full Kelly means betting 2× the optimum, which reduces long-run growth below zero in bad cases. Fractional Kelly is insurance against your own estimation error.
Leverage. f*=1.8 says optimal exposure is 180% of capital. Whether you can implement that depends on financing costs and margin — subtract borrowing cost from μ before deciding.
Sizing garbage optimally still compounds garbage. Validate first with the Deflated Sharpe Ratio and check your backtest length.