N agents each choose between 0 or 1; winners are in the minority. Agents use strategies based on the history of past M outcomes. Over time, agents coordinate despite conflicting incentives — σ²/N (volatility per agent) falls below random chance when M is optimal.