The asymmetry that makes prediction markets unforgiving
Prediction market traders make money in probability space. You are not predicting whether a candidate wins — you are predicting whether the price of "candidate wins" is wrong. That asymmetry has consequences that crush new traders who treat prediction markets like equities or sports betting.
- A 10¢ contract that pays $1.00 has a 9× upside. A 90¢ contract that pays $1.00 has a ~11% upside. The same edge in basis points produces wildly different optimal positions.
- Losses are bounded (you cannot lose more than you put in), but variance is brutal in low-priced contracts where a wrong call costs your entire stake.
- Liquidity dries up fast at the extremes, exactly when you most want to trade — and exactly when news arrives that proves you right or wrong.
Without explicit risk rules, even a strong forecaster gets ground down. This article walks through the five disciplines that separate consistently profitable prediction market traders from the people who blow up their accounts within six months: edge estimation, fractional Kelly sizing, per-market exposure caps, spread awareness, and pre-committed exits.
If you are new to the broader category, our introduction to prediction markets covers the foundational mechanics. For the bankroll layer above this discipline, see bankroll frameworks for information traders.
Step 1 — Estimate your edge honestly
Edge is the difference between your probability estimate and the market price.
edge = p_you - p_market
If you think a binary outcome is 65% likely and the market prices it at 0.55, your edge on the Yes side is +0.10. Don't trade without writing this number down. It is your single most important input, and writing it down forces you to confront the question "where did my probability come from?" before risking capital.
A few rules for honest edge estimation:
- Calibrate first, predict second. Track your past forecasts and look at hit rates per probability bucket. If your 70% calls only resolve 60% of the time, your edge is half what you think it is. The forecast calibration playbook has the full methodology.
- Decompose the question. Which sub-events drive the outcome? Are any already settled? "Will the Fed cut rates at the next meeting?" decomposes into "Will inflation come in below 3%?" + "Will employment data weaken?" + "What is the Fed reaction function?"
- Beware of expert overconfidence. Domain experts often hold tighter probability bands than warranted. Philip Tetlock's Superforecasting shows that the best forecasters update their beliefs more often and hold less extreme views than the average expert. Be more like the superforecasters and less like the cable news pundit.
- Beware of recency bias. A news headline that just dropped feels overwhelmingly important. The market price often has already partially absorbed it. Ask yourself: would I make this trade if the headline were a week old?
The honest test for your edge is whether you would bet your own money at your probability. If you would not, you do not have an edge.
Step 2 — Size with fractional Kelly
The Kelly criterion, introduced in John Kelly's 1956 paper, gives the optimal fraction of bankroll to wager when you have a known edge:
f* = edge / odds
In a binary market with price p, buying Yes gives odds (1 - p) / p. So:
f* = (p_you - p) / ((1 - p) / p) = (p_you - p) * p / (1 - p)
Pure Kelly is volatile. The math assumes your edge estimate is perfectly accurate, which it never is. Real-world traders use fractional Kelly — typically ¼ to ½ of full Kelly — to dampen drawdowns at the cost of growth.
| Strategy | Fraction of Kelly | Expected drawdown | Expected growth |
|---|---|---|---|
| Full Kelly | 1.0× | 50%+ | Highest |
| Half Kelly | 0.5× | 20–25% | 75% of full |
| Quarter Kelly | 0.25× | 10–12% | 50% of full |
| Fixed 1% | n/a | <5% | Slow but steady |
For most retail prediction market traders, quarter Kelly is the practical sweet spot. It keeps drawdowns survivable, lets you compound, and protects you against your own overconfidence in edge estimates. Sophisticated quant traders often use half Kelly with extensive backtesting; pure Kelly is the domain of cold-blooded probability machines that almost nobody actually is.
Step 3 — Cap individual exposure
Even if Kelly says wager 20% of your bankroll, that is almost always too much for a single market. There are three reasons:
- Markets resolve discretely — bad luck can hit several times in a row, and your bankroll has to survive the worst-case path.
- Liquidity walls can blow out your fill price on exit, especially in long-tail markets.
- Resolution disputes happen, and a disputed market that should resolve in your favor can sit frozen for weeks.
Set a hard ceiling — something like 5% of bankroll per market, 20% per category. Walk away from any thesis that wants more. This is the same principle as the position-sizing rules used by professional hedge funds, which limit exposure not just by Kelly but by absolute capital at risk.
The category cap is at least as important as the per-market cap. If you have 4% on a Bitcoin-up market, 4% on an Ethereum-up market, and 4% on an S&P-up market, you have not diversified — you have a 12% bet on "risk-on prevails". Cap the category, not just the market.
Step 4 — Watch the spread
Wide bid/ask spreads are a tax on every entry and exit. Before trading, check:
- What is the round-trip cost (slippage in + slippage out)?
- Does your edge survive that round trip with margin to spare?
- Is liquidity drying up as resolution approaches?
A 5% edge melts fast in a 6% spread. The microstructure layer matters as much as your probability estimate, which is why our market microstructure for forecasting post is required reading for serious traders.
A practical rule: do not enter a position unless the round-trip spread is less than half your expected edge. If you have a 5¢ edge and the spread will cost you 3¢ to enter and exit, you are not really capturing 5¢ — you are capturing 2¢, before any of the other costs (fees, gas, slippage on size). At quarter Kelly with a 2¢ edge, you should size very small — usually less than 1% of bankroll.
Step 5 — Pre-commit to exits
Decide before you open the position:
- What price would make you take profit?
- What event or new information would make you cut?
- Is there a time-stop ("if no news by X, halve the position")?
- What is the resolution date, and does it require holding all the way through?
Write it down. Pre-commitment beats post-commitment every time. The behavioral economics literature is clear: humans systematically hold losers too long and sell winners too quickly. The only protection is to make the decision when you are calm, not when you are watching the price move.
A trade exit framework I have seen work in production:
- Target price. A price level where the trade has substantially captured the edge.
- Stop level. A price level where the original thesis is broken.
- Time stop. A date beyond which the position is closed regardless of price.
- Event triggers. Specific news or data events that automatically close the position.
The discipline is not about predicting which trigger will fire; it is about removing in-the-moment decision-making. You decided when you were objective. Trust your past self.
The 60-second pre-trade checklist
Before pressing the buy button, run through this checklist. It takes less than 60 seconds and prevents 90% of the trades you would regret.
- What is my probability estimate? Why?
- What is the market price?
- What is my edge in basis points?
- What is the round-trip spread cost?
- What is my position size — and is it under my per-market cap (5%) and per-category cap (20%)?
- What is my exit plan — both win and lose, with explicit price and time triggers?
- Is there an information event coming that could re-price me?
If you can answer all seven, place the order. If you cannot, don't trade. The single most underrated trading skill is the ability to walk away from a position you cannot fully describe. Every trade you skip because you did not understand it is a trade you did not lose money on.
Common risk-management mistakes
Three patterns I see repeatedly in traders who blow up.
Doubling down on losses. A trade that goes against you is not "cheaper" — it is information that your thesis was at least partially wrong. Doubling down on a moving market without updating your thesis is the fastest path to ruin.
Sizing to bankroll, not to liquidity. A 5% position in a market with $20k of depth is fine. A 5% position in a market with $500 of depth is not — you have priced yourself out before you have even traded.
Treating each trade independently. Your bankroll is your real exposure. Three uncorrelated 4% positions are very different from three correlated 4% positions. Always look at category-level and theme-level exposure.
Ignoring the resolution layer. A great trade with a fuzzy resolution rule is not a great trade. The resolution layer is where many "winning" positions turn into disputes that lock capital for weeks. See binary vs multiple outcome markets for the design dimensions that affect resolution clarity.
Skipping the journal. Every profitable trader I know keeps a trade journal. Every losing trader I know does not. The journal is what turns trading from a string of bets into a learning process.
The boring secret
Profitable prediction-market traders are obsessive about position sizing and exit discipline. They do not have crystal balls. They lose less when they are wrong and let the rest of the math do its job over thousands of trades.
The reason this looks boring is because it is. The reason it works is because it removes the emotional, in-the-moment decisions that most traders rely on. A spreadsheet, a checklist, and a journal are not glamorous. They are also the difference between a long career and a six-month blowup.
The tortoise wins again.
Frequently Asked Questions
What is the Kelly criterion and how do I apply it to prediction markets?
The Kelly criterion calculates the optimal fraction of your bankroll to bet given a known probability edge. For a binary prediction market with price p and your estimate p_you, the Kelly fraction is (p_you - p) · p / (1 - p). Most prediction market traders use a fractional Kelly — typically ¼ to ½ of full — to manage variance. Full Kelly is mathematically optimal but psychologically unsustainable for most traders.
How big should my position be in a prediction market?
The safe answer is the smaller of (a) quarter Kelly given your edge, (b) 5% of bankroll, (c) 1/4 of available depth at your target price. If any of those three constraints rules out the trade, walk away. Position sizing is more art than science, but those three bounds keep you alive long enough to compound.
What is fractional Kelly and why use it?
Fractional Kelly uses a multiple (e.g., 0.25 or 0.5) of the Kelly criterion to size positions. It produces 75% of the expected growth of full Kelly with about half the volatility, which most traders find psychologically much easier to live with. The fractional approach also provides robustness against errors in your edge estimate — if your probability is off by 5 percentage points, full Kelly can lead to ruin, whereas quarter Kelly survives easily.
How do I track my calibration over time?
Keep a trade journal with the question, your probability, the market price, your reasoning, and the date. After resolution, group your forecasts into buckets (50–60%, 60–70%, etc.) and check whether the actual hit rate in each bucket matches the expected rate. The full methodology is in our forecast calibration playbook.
What is a per-category cap and why does it matter?
A per-category cap limits how much of your bankroll can be exposed to a single theme — politics, crypto, macro, sports. Even if no single market is more than 5% of bankroll, three highly correlated 4% positions add up to 12% of single-theme exposure. Capping the category at 20% (or whatever your risk tolerance allows) ensures that a single news event cannot wipe out your bankroll.
Should I use stop-losses on prediction markets?
Yes, in the form of pre-committed exit prices. Unlike stop-loss orders on a centralised exchange, you generally cannot set a programmatic stop on most prediction markets — you have to monitor and execute manually. Either way, the discipline is the same: decide before you enter what price would mean your thesis is broken, write it down, and act on it.
How do fees and slippage affect my Kelly sizing?
Fees and slippage reduce your effective edge. If your raw edge is 5¢ and your round-trip costs are 2¢, your effective edge is 3¢. Plug the effective edge into the Kelly formula, not the raw edge. Many traders skip this step and overweight positions in fee-heavy or thin markets.
What is the biggest risk-management mistake new prediction market traders make?
Treating prediction markets like sports betting — looking for confident calls and sizing based on conviction rather than edge. Conviction is not the same as edge. You can be 100% convinced about an outcome that is priced at exactly the right probability, and the trade has zero expected value. Risk management starts with edge math, not gut feeling.
Where to go next
You now have the risk management foundation. The natural next steps:
- For the bankroll architecture that sits above this discipline, see bankroll frameworks for information traders.
- For the calibration loop that makes your edge estimates honest, see the forecast calibration playbook.
- For the microstructure layer that determines whether your edge survives execution, see market microstructure for forecasting.
- For the broader context of how prediction markets work, our introduction to prediction markets is the foundational primer.
Risk management is not about predicting the future. It is about surviving long enough that your edge has time to compound. The traders who internalise that discipline are the ones still in the market five years from now.
