Position sizing is what determines whether a trading system survives a losing streak. Most retail traders obsess over entries and exits; the actual variable that matters most is how much — what fraction of capital goes to each trade.
The math is unforgiving. A strategy that wins 60% of the time with even-money outcomes is profitable in expectation, but a string of five losses in a row will happen often enough to matter. If each loss costs 30% of capital, the system is dead before the wins compound. If each loss costs 2% of capital, the system absorbs the streak and keeps going.
Fractional Kelly is the academic answer. The practical answer adds gates on top: a hard cap on per-trade exposure regardless of what Kelly says, a portfolio-level variance budget that constrains the sum of open positions, correlation checks so two "independent" bets don't both blow up because they were correlated all along.
In my system, every trade hits four checks before sizing: jurisdiction, tax consequences, correlation across open positions, variance budget for the week. None of these are about whether the trade is good. They're about whether the next trade can survive if this one is wrong.
The hardest discipline is sizing down when you're confident. Conviction wants to bet big. Conviction is also the most overconfident state a trader is ever in. Sizing fractionally even at 95% model confidence is what separates the systems that compound from the systems that look brilliant for six months and then give it all back.
Bet small. Stay alive. Compound.