Drawdown is peak-to-trough decline — how far the equity curve falls from its previous high before it recovers. Annual returns are what you brag about; drawdowns are what determines whether you stay in the trade long enough for the annual returns to compound.
Max drawdown is the honest metric. A strategy with a 1.4 Sharpe and a 35% max drawdown is meaningfully different from a strategy with a 1.4 Sharpe and an 8% max drawdown. The first one looks great in a deck and gets abandoned by every investor who watches their account fall by a third. The second compounds.
The behavioral problem is that humans don't tolerate drawdowns intellectually. They tolerate them emotionally, and the emotion threshold is lower than the math threshold. A system that's mathematically guaranteed to recover from a 25% drawdown will lose its operator at 18% because the operator stops trusting the system. The drawdown metric has to be designed around the operator, not just the math.
In my Polymarket stack, position sizing is constrained more by acceptable rolling drawdown than by Kelly-optimal sizing. The math allows bigger bets than I let the system take, because the math doesn't account for the human watching the equity curve. Smaller positions, smaller drawdowns, longer survival, more compounding.
The most expensive mistake in trading isn't a single bad trade. It's the operator quitting on a strategy three days before the recovery. Drawdown discipline is what prevents that — sized so the operator can hold through the worst case the math allows.