Position sizing is where most Polymarket traders lose money. Not from bad market calls — from sizing correctly-identified edges too aggressively. You can have a 70% edge and still go broke if you're putting 40% of your bankroll on each trade. A losing streak of three at that size cuts your account to 21.6% of its original value, and the math of recovery from there is brutal. The Kelly Criterion solves this problem, and understanding it is non-negotiable for anyone trading political markets with real capital.
What the Kelly Criterion Actually Says
The Kelly Criterion is the mathematically optimal formula for determining how much of your bankroll to wager on a bet with positive expected value. It maximizes the long-term geometric growth rate of your capital — meaning it compounds your edge as efficiently as possible without risking ruin.
The formula for a binary Polymarket bet is straightforward:
Kelly fraction = (p × b - q) / b
Where:
- p = your estimated true probability of the outcome
- q = 1 - p (probability of losing)
- b = the payout odds (calculated as (1 / market price) - 1)
In practice, on Polymarket: if a market prices an outcome at 60 cents (implied 60% probability) and you believe the true probability is 70%, the Kelly formula tells you exactly how much to bet.
Working through the math: the payout odds (b) equal (1/0.60) - 1 = 0.667. Kelly fraction = (0.70 × 0.667 - 0.30) / 0.667 = (0.467 - 0.30) / 0.667 = 0.167 / 0.667 = 25%.
Full Kelly says bet 25% of your bankroll. This is mathematically optimal but, as we'll see, practically insane for most traders.
Why Full Kelly Is Too Aggressive
The Kelly formula assumes you know the true probability with certainty. In political markets, you never do. Your estimate that a candidate has a 70% chance of winning is itself uncertain — it might be 65% or 75% or 60%. When your probability estimate contains error, full Kelly systematically overbets because it treats your uncertain estimate as if it were precise.
The consequences of overbetting are asymmetric and devastating. Betting slightly more than Kelly optimal reduces long-term growth only marginally in good scenarios, but it dramatically increases the variance of outcomes and the probability of severe drawdowns. Betting double Kelly — which sounds reasonable to anyone who hasn't done the math — actually produces zero expected growth. Beyond double Kelly, your expected bankroll shrinks over time even with a genuine edge.
This is counterintuitive. Most people assume that betting bigger on a winning strategy makes more money. It does, up to the Kelly fraction, and then it doesn't. The curve peaks at Kelly and declines, hitting zero growth at 2× Kelly and going negative beyond that.
Half-Kelly: The Standard for Political Traders
Virtually every serious prediction market trader uses half-Kelly (or less) as their practical sizing rule. Half-Kelly means taking the Kelly formula's output and dividing it by two. For the example above, that's 12.5% of bankroll instead of 25%.
Half-Kelly sacrifices about 25% of the maximum possible growth rate in exchange for dramatically lower variance and drawdown risk. In a market where your probability estimates carry meaningful uncertainty — which describes every political market — this trade is overwhelmingly positive.
The practical impact: at half-Kelly, a losing streak of three trades sized at 12.5% each reduces your bankroll to about 67% of its starting value. Painful, but recoverable. At full Kelly (25% each), the same losing streak takes you to 42% — a 58% drawdown that requires a 138% gain to recover.
For political markets specifically, half-Kelly has an additional advantage. Political markets are long-dated and illiquid compared to financial markets. If you're overexposed and need to exit, the spread costs and market impact of unwinding a large position can be several percent of the position size. Lower sizing keeps your positions small relative to market liquidity, ensuring you can exit when you need to.
Applying Kelly to Multi-Outcome Political Markets
Standard Kelly assumes a binary bet — win or lose. But many Polymarket political markets feature multiple outcomes. The 2028 Democratic Nominee market, for example, has dozens of candidates. Applying Kelly to multi-outcome markets requires a modified approach.
The simplest method is to treat each candidate's "Yes" share as an independent binary bet. If you believe AOC has a 25% true probability and she's trading at 19 cents, you calculate the Kelly fraction for a binary bet at those odds. The limitation is that positions across multiple candidates in the same market are correlated — if AOC goes up, the sum of everyone else must go down — and independent Kelly doesn't account for this.
More sophisticated traders use simultaneous Kelly, which optimizes across all positions in a multi-outcome market at once. The math is significantly more complex (it involves constrained optimization), but the intuition is simple: your total exposure to a single market should be bounded, and the allocation across candidates within that market should reflect your relative confidence in each one.
For most political traders, a practical rule suffices: your total position size across all outcomes in a single market should not exceed what Kelly would recommend for a single binary bet. This prevents overconcentration in any one event, regardless of how many angles you're playing within it.
Position Sizing Across Correlated Markets
Political markets are correlated in ways that create hidden portfolio risk. A position on House control, a position on individual House races, and a position on the generic ballot average are all expressions of similar underlying views. If the political environment shifts, all three move against you simultaneously.
Kelly sizing should account for these correlations by treating correlated positions as partial duplicates. If you have a half-Kelly position on Democrats winning the House and another half-Kelly position on a specific Democratic House candidate, your effective exposure to a pro-Democratic political environment is close to full Kelly — which is exactly the danger zone.
The fix is to think in terms of portfolio Kelly rather than individual trade Kelly. Group your positions by the underlying drivers: national political environment, specific state dynamics, individual candidate factors. Set a maximum exposure budget for each driver, and allocate within that budget.
Common Sizing Mistakes in Political Markets
Sizing based on conviction rather than edge. Conviction is emotional. Edge is mathematical. A trade where you're 90% confident might have a 2% edge (true probability 62% vs. market price 60%), while a trade where you're less confident might have a 10% edge (true probability 45% vs. market price 35%). Kelly sizes the second trade much larger because the edge is larger, regardless of your feelings about it.
Ignoring the vig. Polymarket's fee structure and the bid-ask spread are transaction costs that reduce your effective edge. A market priced at 60 cents where your true probability is 62% looks like a 2% edge — but after spread costs and fees, the effective edge might be 1% or less. Kelly on a 1% edge produces a tiny position, which is the correct answer even if it feels unsatisfying.
Increasing size after wins. Kelly naturally scales position sizes with your bankroll — bigger bankroll, bigger positions. But traders often accelerate this after a winning streak, sizing up faster than Kelly recommends because they feel confident. This is a recipe for giving back gains, because the overconfidence peaks exactly when a reversion is most likely.
Flat sizing on all trades. Putting the same dollar amount on every trade, regardless of edge size, is suboptimal. Kelly tells you to bet more on larger edges and less on smaller ones. Flat sizing either oversizes your small-edge trades (increasing drawdown risk) or undersizes your large-edge trades (leaving money on the table).
Frequently Asked Questions
What's the minimum edge to trade on Polymarket? Kelly produces a positive bet size for any positive expected value, no matter how small. But practical constraints — transaction costs, time, attention, and the uncertainty in your probability estimates — mean that edges below 3-5% are usually not worth trading. The Kelly fraction on a 2% edge is so small that the expected profit after costs doesn't justify the effort.
How do I estimate true probabilities for political markets? Combine base rates (historical patterns for similar events), polling data, and qualitative assessment of candidate and environmental factors. Your estimate doesn't need to be precise — it needs to be more accurate than the market price. Starting with the market price and adjusting based on information you believe the market hasn't fully incorporated is a practical approach.
Should I use Kelly for Polymarket sports markets too? The Kelly framework applies to any market with a quantifiable edge, including sports prediction markets. The key challenge is the same: accurately estimating your true probability relative to the market price. Sports markets tend to be more efficiently priced than political markets, so edges are smaller and Kelly fractions are correspondingly tiny.
The Bottom Line on Sizing
Kelly Criterion isn't just a formula — it's a framework for surviving long enough to let your edge compound. In political markets, where positions are long-dated, liquidity is variable, and probability estimates carry real uncertainty, half-Kelly or lower is the standard. The traders who blow up on Polymarket aren't the ones who pick the wrong candidates — they're the ones who pick the right candidates but size the position as if they were certain. Size for survival first. The compound growth takes care of itself.
Key Takeaways
- Kelly sizing maximizes long-term geometric growth without risking ruin — the formula for a binary bet is (p × b − q) / b.
- Full Kelly is too aggressive given uncertain probability estimates; half-Kelly (or lower) is the standard for political traders.
- Most blow-ups come from sizing, not bad picks — a correct 70% edge still ruins you at 40% of bankroll per trade.
- Adjust for correlation: stacked positions across linked markets behave like one oversized bet.
Related guides: why 84% of Polymarket traders lose money · building a correlated-market portfolio — or see what PolyBro is and join the PolyBro waitlist for an autonomous AI agent that researches any Polymarket market for you.