The primary cause of skilled forecasters underperforming in prediction markets is rarely inaccurate forecasting — it's inadequate capital allocation. Even a sound probability assessment becomes worthless if an unlucky run depletes your entire stake. This guide outlines the methodology that safeguards against such catastrophe.
The Kelly Criterion: The Mathematical Foundation
Kelly Criterion determines the theoretically ideal percentage of your stake to allocate to each wager: f = (bp - q) / b
- b = net odds received (e.g., if YES costs 0.40, b = 1.5)
- p = your probability estimate
- q = 1 - p
- Result: optimal fraction of bankroll for this position
In practice: use half-Kelly. Whilst Kelly delivers mathematical optimality under conditions of certainty, our probability estimates carry inherent uncertainty, making half-Kelly the superior choice for risk-adjusted performance.
Hard Rules: Never Break These
- Maximum 5% of bankroll per single position — no exceptions regardless of conviction
- Maximum 25% of bankroll in any single correlated cluster — e.g., all US election markets
- Stop-loss: if you lose 25% of your starting bankroll in a month, stop trading for the rest of the month
- Never add to a losing position to "average down" — reevaluate the fundamental thesis first
Drawdown Recovery
Temporary losses occur regularly, even among traders with genuine advantage. Following a 20% decline in account value, halve your position sizes until you return to your previous peak. This approach ensures that temporary setbacks do not spiral into irreversible losses.
FAQ
- How much starting capital do I need for serious prediction market trading?
- $500-1,000 allows sufficient capital to build a balanced portfolio spanning 10-20 trades using half-Kelly allocation. Below $100, sizing constraints prevent effective application of disciplined methods.
- What should I do after a winning streak?
- Exercise greater caution, not complacency. Successful runs breed confidence bias. Maintain your systematic allocation framework independent of recent results.