Why Position Sizing Matters More Than Picking Winners
Here's a truth that separates profitable prediction market traders from everyone else: how much you bet matters more than what you bet on. You can have a 60% hit rate and still go broke with bad sizing. Conversely, disciplined position sizing can turn a modest edge into consistent, compounding returns.
Most traders on platforms like Polymarket focus almost exclusively on finding mispriced markets. That's important — you need an edge to win long-term. But the math of how you allocate capital across those edges is where the real money is made or lost.
The Kelly Criterion: A Starting Point
If you've spent any time in betting or trading circles, you've probably encountered the Kelly Criterion. Developed by John Kelly at Bell Labs in 1956, it tells you the optimal fraction of your bankroll to wager:
Kelly % = (bp - q) / b
Where b is the odds received (decimal odds minus 1), p is your probability of winning, and q is the probability of losing (1 - p).
For prediction markets, this translates neatly. If a market is trading at $0.40 (implying 40% probability) and you believe the true probability is 55%, Kelly tells you to bet roughly 25% of your bankroll. That's a huge bet — and therein lies the problem.
Why Full Kelly Is Dangerous
Full Kelly sizing assumes you know your edge exactly. In practice, you almost never do. Your probability estimates carry uncertainty, and prediction markets have unique complications:
- Resolution ambiguity — markets can resolve in unexpected ways based on rule interpretation
- Correlated outcomes — your portfolio of bets may be more correlated than you think (e.g., multiple political markets moving together)
- Liquidity constraints — large positions move the price against you, degrading your actual edge
- Estimation error — even a 5% miscalibration in your probability estimate can flip Kelly from profitable to destructive
Studies consistently show that fractional Kelly — betting a fraction of what full Kelly recommends — produces better real-world results. The theoretical sacrifice in growth rate is small, while the reduction in drawdowns and variance is enormous.
Fractional Kelly: The Practical Standard
Most professional bettors and quantitative traders use quarter-Kelly to half-Kelly sizing. Here's why the tradeoff is so favorable:
- Half-Kelly captures roughly 75% of the growth rate of full Kelly while cutting variance in half
- Quarter-Kelly still captures about 50% of optimal growth with dramatically smoother equity curves
- The risk of ruin drops from "possible" to "essentially zero" as you move to fractional sizing
For prediction market traders managing a portfolio of 10-30 active positions, half-Kelly is a solid baseline. If you're less confident in your probability estimates — honestly, most of us should be — quarter-Kelly provides excellent protection.
Going Beyond Kelly: Portfolio-Level Thinking
Kelly was designed for sequential, independent bets. Prediction markets are messier. You're running a portfolio of simultaneous positions with varying correlations, time horizons, and liquidity profiles. Here's how to adapt:
1. Account for correlation. If you hold YES on five different markets that all depend on the same political outcome, your real exposure is much larger than any single position suggests. Group correlated bets and apply Kelly to the group, not individual positions.
2. Weight by confidence. Not all edges are created equal. A 15% edge backed by strong data from multiple independent sources deserves more capital than a 15% edge based on a gut feeling. Scale your Kelly fraction by your confidence in the estimate itself.
3. Reserve capital for opportunities. Markets move. New edges appear. If you're fully allocated, you can't capitalize on a sudden mispricing. Keeping 20-30% of your bankroll in reserve isn't just conservative — it's strategic.
4. Adjust for time horizon. A market resolving in two days ties up capital very differently than one resolving in six months. Factor in the opportunity cost — that capital could be earning returns elsewhere during the wait.
A Practical Sizing Framework
Here's a framework that works well for active prediction market traders:
- Maximum single position: 5-8% of bankroll (hard cap, no exceptions)
- Maximum correlated cluster: 15% of bankroll across all positions sharing the same underlying driver
- Base sizing: Half-Kelly, scaled down further when edge estimate uncertainty is high
- Minimum edge threshold: Don't trade edges below 5-7% — transaction costs, slippage, and estimation error eat small edges alive
- Portfolio heat: Total capital at risk shouldn't exceed 60-70% of bankroll at any time
This framework won't maximize theoretical returns, but it will keep you in the game long enough for your edge to compound — which is the whole point.
Automating the Edge Detection
Position sizing only works when you have accurate edge estimates to size against. This is where most manual traders struggle — the process of comparing market prices against data-driven probability models is tedious and error-prone when done by hand.
Tools like EdgeScouts automate the edge detection process, scanning Polymarket continuously and comparing prices against models built on data from sportsbooks, weather services, options markets, and other sources. When you can see the estimated edge for each market in real time, applying a disciplined sizing framework becomes straightforward rather than guesswork.
The Bottom Line
Finding edges in prediction markets is only half the battle. Sizing those edges correctly is what turns a good trader into a profitable one. Start with fractional Kelly, think at the portfolio level, respect correlation, and keep reserves for new opportunities.
The math isn't complicated. The discipline to follow it consistently — especially when a high-conviction bet tempts you to go big — is what separates the winners from the washouts.
Ready to find edges worth sizing? Check out edgescouts.com to see where the prediction markets are mispriced right now.