Kelly Criterion for Polymarket: Position Sizing That Prevents Ruin

The Kelly Formula

The Kelly Criterion calculates the optimal fraction of your bankroll to allocate to a single bet. The formula is:

Where:

• f* = the fraction of your bankroll to wager
• b = the net odds received on the bet (i.e. if you risk $1 and win $1.50, b = 1.5)
• p = your estimated probability that the bet wins
• q = 1 − p (the probability the bet loses)

On Polymarket, translating this is straightforward. If you believe an event has a 70% probability of occurring and the current YES share price is $0.60 (implying 60% market probability), your net odds on a YES position are: for every $0.60 you spend, you win $0.40 if correct and lose $0.60 if wrong. So b = 0.40 / 0.60 = 0.667, p = 0.70, q = 0.30.

Kelly fraction: (0.667 × 0.70 − 0.30) / 0.667 = (0.467 − 0.30) / 0.667 = 0.167 / 0.667 = approximately 25% of bankroll.

That is what full Kelly says. Wager 25% of your bankroll on this single trade. For a $10,000 bankroll, that is $2,500 — a number that should immediately concern you.

Why Full Kelly Is Dangerous

Full Kelly is mathematically optimal only when your probability estimates are perfectly accurate. In practice, no trader has perfect calibration. Overestimating your edge by just 10-20% — which is extremely common — transforms Kelly from a growth-maximiser into a ruin accelerator.

The mathematics of Kelly betting mean that under full Kelly:

• You have a 33% probability of halving your bankroll at some point before doubling it
• Variance is enormous — sustained runs of losses, even with a genuine edge, can feel indistinguishable from having no edge at all
• Emotional pressure at large position sizes leads to poor decisions that compound the mathematical problem
• Any systematic bias in your probability estimates amplifies catastrophically — you are betting large on your mistakes

Professional traders across all markets — sports betting, financial trading, poker — almost universally use a fraction of Kelly rather than full Kelly. The most common standard is quarter-Kelly.

Quarter-Kelly: The Practitioner Standard

Quarter-Kelly means multiplying your Kelly fraction by 0.25. In the example above, instead of wagering 25% of your bankroll, you wager 6.25%. The expected growth rate is slightly lower, but the probability of halving your bankroll drops from 33% to under 4%.

This is not just a conservative heuristic — it is the rational response to uncertainty in your own probability estimates. If your edge is real, you will still grow your bankroll significantly at quarter-Kelly. If your edge is smaller than you think (likely), quarter-Kelly keeps you in the game long enough to discover and correct the error.

In practice, most experienced Polymarket traders operate within a 2-15% per-position range, with the exact size determined by a combination of Kelly sizing and a hard cap per market. A good default rule: never put more than 10% of your total bankroll on any single Polymarket position, regardless of what Kelly suggests.

Worked Example: A Political Market

You are looking at a market: "Will Party A win the Senate seat in State X?" Current YES price: $0.44. Your analysis — incorporating polling, historical base rates, and expert forecasters — gives you an estimated probability of 55%.

Let's run the Kelly calculation:

• YES price: $0.44, so each share pays $0.56 profit if correct
• b = 0.56 / 0.44 = 1.27 (net odds)
• p = 0.55 (your estimated probability)
• q = 0.45
• Full Kelly: (1.27 × 0.55 − 0.45) / 1.27 = (0.699 − 0.45) / 1.27 = 0.249 / 1.27 ≈ 19.6%
• Quarter-Kelly: 19.6% × 0.25 = 4.9% of bankroll

On a $5,000 bankroll, quarter-Kelly says wager approximately $245. If you win, your bankroll grows by $245 × (0.56 / 0.44) ≈ $312 profit for a total of $5,312. If you lose, you lose $245 — painful but not destructive.

Now imagine you are wrong about your edge and the true probability is only 48%, not 55%. At quarter-Kelly based on 55%, you are still wagering a small enough fraction that the downside is manageable. At full Kelly, you would have been wagering 19.6% — nearly $1,000 — on a trade with a slight negative edge. The difference between full and quarter-Kelly is the difference between a recoverable mistake and a bankroll-threatening one.

Diversification Across Uncorrelated Markets

Kelly sizing assumes independence between bets. On Polymarket, this assumption breaks down quickly if you are not careful. Multiple political markets on the same election cycle are correlated — a wave result in one race affects all the others. Multiple crypto markets are correlated to underlying asset price moves.

The practical implication: running Kelly independently on each position ignores portfolio-level correlation, which means your actual risk is higher than the individual Kelly fractions suggest. To account for this:

• Cap total correlated exposure: Never have more than 20-25% of your bankroll in markets that move together (e.g. multiple markets on the same election).
• Diversify across categories: Spread exposure across political, crypto, sports, and economic markets. A black-swan event in one category will not wipe positions across all categories simultaneously.
• Apply portfolio-level Kelly: Some traders run a global Kelly calculation on their entire portfolio rather than position by position. The total at-risk fraction stays within a Kelly-optimal range regardless of how many positions are open.

Kelly for Different Strategy Types

The appropriate Kelly fraction varies by strategy type. Not all Polymarket strategies carry the same uncertainty in edge estimation:

• Arbitrage: Edge is mathematical and near-certain (YES + NO price below $1.00 guarantees profit). Kelly suggests large fractions, often up to 50-80% per opportunity. In practice, speed and liquidity constraints are the binding limit, not Kelly.
• Directional with strong research: High conviction backed by multiple data sources. Quarter-Kelly is appropriate — typically 5-12% of bankroll per position.
• Copy trading: You are delegating the edge estimation to the wallet you are copying. Use eighth-Kelly or a fixed small percentage until you have 50+ data points on the wallet's actual performance.
• Directional with thin research: Markets where you have a vague view but no quantified edge. Minimum position sizes only — think of these as exploration trades, not bankroll-allocation decisions.

Frequently Asked Questions

Should I ever bet more than Kelly suggests? No. Betting more than full Kelly produces lower expected growth AND higher variance — the worst of both worlds. Kelly is an upper bound. The only intelligent deviation is to bet less.

How do I estimate my edge on Polymarket? Track your probability estimates against outcomes over at least 50 trades. If you consistently estimate 65% probability on events that occur 55% of the time, your effective edge is smaller than your Kelly inputs assume and you should reduce your fraction accordingly.

Does Kelly apply to market making strategies? Yes, but differently. Market makers on Polymarket size their quote depth based on inventory risk and expected spread capture rather than directional Kelly. The underlying principle — risk proportional to edge — is the same.