Prediction Market Bankroll Management: How to Size Positions and Survive
Prediction market bankroll management requires different math than sports betting. 5 worked examples cover Kelly sizing, capital lockup, and why 91% of traders lose.
Why prediction market bankroll management is a different game
Prediction market bankroll management is not sports betting bankroll management with different labels. The math changes in ways that catch most traders off guard.
In sports betting, capital recycles fast. You bet $500 on a Sunday NFL game and win or lose by Monday morning. Your bankroll is fully available for the next opportunity within hours. The Kelly Criterion works cleanly because the assumption of sequential bets with rapid settlement holds.
Prediction markets break that assumption. A Kalshi contract on "Will the Fed cut rates by June?" locks your capital for months. A Polymarket position on an election outcome might not resolve for a year. Meanwhile, new opportunities appear daily and you have no capital to deploy. This is the core problem of how prediction markets work from a bankroll perspective: your money is frozen while the market moves without you.
Three structural differences matter:
Capital lockup. Sportsbook capital turns over in hours. Prediction market capital can be locked for weeks or months. A $1,000 position in a quarterly contract earns zero additional return during that entire period.
Simultaneous positions. Active prediction market traders hold 10 to 30 open contracts at once. Each one consumes bankroll. Total portfolio exposure can quietly exceed 100% of your intended risk budget.
Fee asymmetry. Prediction market fees hit only on wins (Kalshi) or through spreads (Polymarket). This changes the Kelly math because your downside is the full contract cost, but your upside is reduced by platform fees.
Understanding these differences is the gap between the 9% of prediction market traders who profit and the 91% who do not.
Kelly Criterion adapted for binary contracts
The standard Kelly formula for a binary prediction market contract looks the same as for a sportsbook bet. You buy a contract at price c with true probability p. The decimal odds equivalent is d = 1/c.
Kelly % = (p x d - 1) / (d - 1)
But this raw number is wrong for prediction markets because it ignores fees. On Kalshi, a 7% winner fee reduces your net profit per contract. The adjustment is straightforward: replace the raw payout with the fee-adjusted payout.
Worked example: Kalshi contract at $0.62
You find a Kalshi contract trading at $0.62. Your model says the true probability is 70%. Without fees:
- Decimal odds: 1 / 0.62 = 1.613
- Raw Kelly: (0.70 x 1.613 - 1) / (1.613 - 1) = (1.129 - 1) / 0.613 = 0.129 / 0.613 = 21.0%
Now adjust for Kalshi's 7% winner fee. On a win, your profit is $0.38 per contract. After fees: $0.38 x 0.93 = $0.3534. Your effective payout per contract is $0.62 + $0.3534 = $0.9734. Effective decimal odds: $0.9734 / $0.62 = 1.570.
- Fee-adjusted Kelly: (0.70 x 1.570 - 1) / (1.570 - 1) = (1.099 - 1) / 0.570 = 0.099 / 0.570 = 17.4%
Fees cut your optimal position size from 21.0% to 17.4%. That is a 17% reduction in recommended allocation from a single fee. Ignore this adjustment and you are systematically over-betting every position. The Kelly Criterion calculator handles this math, and the fee calculator shows you the exact fee drag at any contract price.
Why half Kelly is mandatory for prediction markets
Full Kelly is already dangerous in sports betting. In prediction markets, it is reckless. Three reasons.
Probability estimation is harder. Sportsbook lines are benchmarked against sharp markets with millions in liquidity. You can de-vig a -110/-110 line and get a high-quality true probability estimate. Prediction market contracts on political, economic, or cultural events have no equivalent benchmark. Your 70% estimate on that Kalshi contract could easily be 65% or 75%. That 5-point error range makes full Kelly toxic.
Positions overlap. You might hold 15 contracts simultaneously. Full Kelly on each one ignores the fact that they share your bankroll. If 5 of those 15 contracts are correlated (say, all economic indicators that move together), a single macro shift can hit all 5 at once. Read the correlated positions guide for the portfolio math.
No quick recovery. In sports betting, a bad week can be recovered by the following week's bets. In prediction markets, if a bad month wipes 40% of your bankroll and most of your remaining capital is locked in open positions, you cannot resize. You are stuck riding out positions that are now oversized relative to your diminished bankroll.
The recommendation: half Kelly at most. For the Kalshi contract above, half Kelly says 8.7% of bankroll. On a $5,000 bankroll, that is $435. If your edge estimate is uncertain, drop to quarter Kelly: $218. These numbers feel small. They are supposed to. Survival is the first priority.
Capital allocation across multiple positions
The single-bet Kelly formula does not account for a portfolio of 20 open contracts. Most prediction market traders hold positions across multiple markets simultaneously, and the total allocation math gets ignored.
The exposure cap rule
Calculate Kelly (at half or quarter) for each individual position. Then sum all current allocations. If total exposure exceeds 25% of your bankroll, scale every position proportionally downward until total exposure hits 25%.
Worked example: five open positions
| Contract | Price | True Prob | Half Kelly | Allocation ($5,000 BR) |
|---|---|---|---|---|
| Fed rate cut | $0.62 | 70% | 8.7% | $435 |
| Senate race A | $0.45 | 55% | 5.2% | $260 |
| GDP growth > 3% | $0.30 | 42% | 4.8% | $240 |
| Supreme Court ruling | $0.72 | 80% | 6.1% | $305 |
| Tech earnings beat | $0.55 | 65% | 7.9% | $395 |
Total allocation: 32.7% of bankroll ($1,635). That exceeds the 25% cap. Scale factor: 25% / 32.7% = 0.764.
Adjusted allocations: $332, $199, $183, $233, $302. Total: $1,249 (25% of $5,000).
This seems conservative. It is. But consider what happens without the cap. If all five positions lose (unlikely but not impossible, especially if the Fed rate cut and GDP growth contracts are correlated), you lose 32.7% of your bankroll in a single resolution cycle. With the cap, max loss is 25%. After that drawdown, your $3,750 remaining bankroll still has enough capital to continue trading. Without the cap, $3,365 remaining after a five-contract wipeout leaves you in a deeper hole that requires a 49% gain just to recover.
The capital lockup problem
This is the silent killer of prediction market returns. Capital locked in open positions earns zero additional return until the contract settles.
Worked example: lockup vs. turnover
You have $5,000 and two options:
Option A: Buy $2,000 of a 5% edge contract on Kalshi that resolves in 90 days. Expected profit: $2,000 x 0.05 = $100 over 3 months.
Option B: Deploy $2,000 across weekly contracts with a 2% edge that resolve every 7 days. Over 90 days, you cycle through roughly 13 rounds. Expected profit: $2,000 x 0.02 x 13 = $520 over 3 months.
Option B generates 5.2x more expected profit despite having less than half the per-trade edge. This is the bankroll turnover concept applied to prediction markets. The turnover calculator models this tradeoff for any combination of edge and duration.
The practical implication: never deploy more than 40-50% of your bankroll into contracts that resolve beyond 30 days. Keep the other half available for short-duration opportunities. A 2% edge on a weekly contract compounds faster than a 6% edge on a quarterly contract.
| Contract Duration | Annual Turnover | 3% Edge Annual EV per $1,000 |
|---|---|---|
| Daily | 365x | $10,950 |
| Weekly | 52x | $1,560 |
| Monthly | 12x | $360 |
| Quarterly | 4x | $120 |
The numbers speak for themselves. Duration is a cost. Treat it that way.
Risk of ruin for prediction market traders
Risk of ruin is the probability your bankroll hits zero (or some unusable minimum) before the math converges in your favor. For prediction market traders, the numbers are worse than most expect. The full math, including the formulas and detailed ruin probability tables, is in the prediction market risk of ruin guide.
Why 91% of Polymarket traders lose money
The statistic is real. Roughly 91% of Polymarket wallets that have traded show net losses. The primary drivers are not bad market reads. They are bankroll mismanagement:
Over-concentration. Putting 30-50% of capital into a single high-conviction contract. Even a 75% probability contract loses 25% of the time. Do that four times and you have a 32% chance of losing at least once.
Ignoring fee drag. Trading thin edges without accounting for platform fees that flip the trade to -EV. A 3% edge on a $0.70 Kalshi contract becomes roughly 0.5% after fees. One bad trade wipes a dozen winners.
No position sizing system. Betting "what feels right" instead of running the Kelly formula. Emotional sizing overweights recent winners and chases losses.
Capital lockup death spiral. Deploying 90% of bankroll into long-dated contracts, then watching short-term opportunities pass by. When those long-dated contracts lose, there is no capital to recover with.
Risk of ruin scenarios
For a trader using half Kelly with a 3% average edge and a 25% exposure cap:
| Starting Bankroll | Ruin Threshold | Risk of Ruin |
|---|---|---|
| $5,000 | $500 | < 1% |
| $2,000 | $500 | ~4% |
| $1,000 | $500 | ~12% |
| $500 | $200 | ~25% |
Small bankrolls face serious risk even with correct sizing. If your bankroll is under $1,000, quarter Kelly is the only defensible strategy. The math does not care about your conviction level. The PM EV calculator will confirm whether a trade has positive expected value before you size it.
Monte Carlo simulation: position sizing and ruin over 500 trades
What does ruin actually look like across hundreds of trades? A Monte Carlo simulation of 10,000 traders, each making 500 trades at a 55% win rate on even-money contracts, makes the relationship between position sizing and survival concrete.
Setup: $5,000 starting bankroll, 55% win probability, $0.50 contracts, ruin threshold $500.
| Bet Fraction | Traders Ruined (of 10,000) | Survival Rate | Median Surviving Bankroll |
|---|---|---|---|
| 2% (Quarter Kelly) | 12 | 99.88% | $7,180 |
| 5% (Half Kelly) | 184 | 98.2% | $9,350 |
| 10% (Full Kelly) | 831 | 91.7% | $11,800 |
| 20% (2x Kelly) | 3,847 | 61.5% | Ruined |
The difference between 2% and 20% sizing is the difference between 99.88% survival and 61.5% survival. Same edge. Same markets. Only the position size changes. The median 2x Kelly trader is bankrupt after 500 trades despite having a real 5% edge on every single trade.
The simulation also reveals path dependency. A trader betting 10% who hits a 7-trade losing streak early (probability: 0.45^7 = 0.37%) drops from $5,000 to $2,325. Recovery from that drawdown requires a 115% gain, or 233 net winning trades at 10% sizing. Most traders at that point either quit or increase bet size to "make it back faster," which accelerates ruin.
For the complete ruin probability tables across different edge sizes and Kelly fractions, plus the formulas behind these numbers, read the full risk of ruin analysis.
When to sell early vs. hold to resolution
Prediction markets have an option sports bettors do not: you can sell a position before it resolves. This creates a bankroll management decision at every price movement.
The opportunity cost framework
You bought a contract at $0.45 and it has moved to $0.65. You still believe the true probability is 75%. Should you hold or sell?
Hold to resolution: Expected value = (0.75 x $0.55) - (0.25 x $0.45) = $0.4125 - $0.1125 = $0.30 per contract (after subtracting your $0.45 cost, net expected profit is $0.30). But Kalshi takes 7% of the $0.55 profit: net expected profit = (0.75 x $0.55 x 0.93) - (0.25 x $0.45) = $0.384 - $0.1125 = $0.271 net per contract, with resolution in 60 days.
Sell now at $0.65: Lock in $0.20 profit per contract immediately (no Kalshi settlement fee on secondary market sales). Redeploy that $0.65 into a new position with edge today.
If you can find a new trade with a 3% edge on weekly contracts, that $0.65 redeployed generates roughly $0.65 x 0.03 x 8 = $0.156 over the same 60 days. Add the locked-in $0.20 profit: total = $0.356.
Selling and redeploying beats holding ($0.356 vs. $0.271) in this scenario. The breakpoint depends on your redeployment edge and the remaining time to resolution. As a rule of thumb: if a contract has moved more than halfway to your target and resolution is more than 30 days away, selling and redeploying usually wins.
The prediction market bankroll management system
Bringing it all together into a repeatable process:
- Set your bankroll. Dedicate a specific amount to prediction market trading. This is money you can lose entirely without financial hardship.
- Run the Kelly calculator on every trade. Input your true probability estimate and the contract price. Use the PM EV calculator to confirm positive expected value after fees.
- Apply half Kelly. Cut the calculator output in half. If your probability estimate is uncertain, use quarter Kelly.
- Enforce the 25% exposure cap. Sum all open positions. If total exceeds 25% of current bankroll, scale everything down proportionally.
- Limit long-dated exposure to 40%. No more than 40% of bankroll in contracts resolving beyond 30 days. The rest stays liquid for high-turnover opportunities.
- Review weekly. Check total exposure, correlation between positions, and whether any contracts should be sold early and redeployed. Track your actual expected value per trade over time.
This system will not make you profitable on its own. You still need to find genuine edge through better probability estimates than the market price implies. But without this system, even accurate probability estimates lead to ruin through over-concentration, capital lockup, and fee erosion. That is the lesson hidden in the 91% loss rate. The prediction market strategy guide covers the edge-finding side. The position sizing guide goes deeper on portfolio construction.
The math is not complicated. The discipline to follow it when a high-conviction trade tempts you to oversize is what separates the 9% from the 91%.
Frequently asked questions
- How much of my bankroll should I risk on a single prediction market contract?
- Use half Kelly sizing, which typically produces allocations of 2-8% per position depending on your edge. Apply a 25% total exposure cap across all open positions. Never put more than 10% into a single contract regardless of conviction.
- Why do 91% of Polymarket traders lose money?
- The primary drivers are bankroll mismanagement: over-concentration in single positions, ignoring fee drag on thin edges, emotional sizing without a formula, and locking too much capital in long-dated contracts. The edge-finding is secondary to the sizing and risk management failures.
- How does capital lockup affect prediction market returns?
- Capital locked in long-dated contracts earns zero additional return until resolution. A 2% edge on weekly contracts generates roughly 5x more annual profit than a 5% edge on quarterly contracts because of higher turnover. Keep at least 50-60% of your bankroll available for short-duration opportunities.
- Should I use full Kelly or half Kelly for prediction markets?
- Half Kelly at most. Prediction market probability estimates are less precise than de-vigged sportsbook lines, positions overlap and correlate, and locked capital prevents quick recovery from drawdowns. Quarter Kelly is appropriate when your edge estimate is uncertain.
- When should I sell a prediction market position early instead of holding to resolution?
- Sell when the contract has moved more than halfway to your target price and resolution is more than 30 days away. The capital freed up can be redeployed into new +EV positions, often generating more total return than waiting for the original contract to settle.