Fundamental analysis for prediction markets

The core idea

A Polymarket price is a probability: 35¢ means the market thinks about a 35% chance. Fundamental analysis means forming your own probability independently, then comparing. If you think 45% and the price is 35¢, buying YES has positive expected value, as long as your 45% is better than the market's 35%. Most of the work is making sure it is.

Step 1: start with a base rate

Before reading any news, ask: of all past situations like this one, how often did the event happen? How often do incumbent parties hold a seat like this? How often does a central bank move when markets price a 70% chance? How often does a film with these early reviews top the box office? This “outside view” protects you from stories that feel convincing but describe something rare.

Choose the reference class carefully. “Tech CEOs resigning this year” and “CEOs under active board investigation resigning within a month” give very different numbers. When unsure, look at several reference classes and land somewhere between them.

Step 2: adjust with evidence

• Polls: use averages, not single polls, and know each pollster's record. Polls have errors that tend to point the same way across a whole election.
• Models: published forecasting models for elections, sports and economics combine many inputs. Treat them as one strong input, not the answer.
• Market data: futures, options and bond prices often imply probabilities for rate decisions or price levels. Other prediction markets are useful cross-checks; see Polymarket vs Kalshi.
• Experts and primary sources: official schedules, court dockets, filings and statements beat commentary about them.

Move away from the base rate in proportion to how strong and specific the evidence is. Vivid anecdotes deserve small adjustments; a hard data point deserves a large one.

Step 3: price the contract, not the headline

Your forecast must be for the exact question in the rules: the named source, the deadline and time zone, and edge cases such as revisions, cancellations or ties. “Will X happen?” and “Will source Y report X by 11:59 pm ET on date Z?” can have very different probabilities. Our guide to reading market rules covers what to check, and how markets resolve explains disputes.

Step 4: update on news, carefully

When news arrives, ask two questions. How much more likely is this news if the event will happen than if it won't? And how much of it is already in the price? Big headlines often change little, while a small procedural detail can matter a lot. Update in steps rather than jumping to 0% or 100%, and avoid two common mistakes: ignoring news that cuts against your position, and overreacting to the latest story. For trading around scheduled events, see news trading.

Step 5: decide whether to trade

Compare your estimate with the price you would actually pay, the ask, not the midpoint, plus the fee. A 3-point edge in a 0.04 category near 50¢ shrinks once you account for spread and fees. Use the EV calculator and size with a fraction of Kelly using the Kelly calculator. Remember that the price reflects many people's information; if you can't explain why the market is wrong, it probably isn't.

Calibration

You are well calibrated if events you call 70% happen about 70% of the time. Most people are overconfident: their 90% calls come true far less often than 90% of the time. Calibration is measurable, and it is the clearest sign of whether your analysis is adding anything. A common score is the Brier score: the average of (forecast − outcome)², where the outcome is 1 or 0. Lower is better; always forecasting 50% scores 0.25.

Keep a forecast journal

Write down every forecast before you trade, including ones you don't trade. A simple spreadsheet works:

After 50 to 100 resolved forecasts, group them by probability band and compare with what actually happened. Also check whether you beat the market price you saw at the time. If you don't, reduce your size until you do.

Frequently asked questions