Skip to content
Friday, April 17, 2026Subscribe →
political forecasts
◆  The broadsheet of prediction markets  ◆

How Polymarket Prices Map to Probability

A clear, worked explanation of how Polymarket and Kalshi prices convert to implied probability — and how well-calibrated those probabilities have been historically.

The mechanic

Every binary market on Polymarket has two contracts: YES and NO. Prices sum to exactly $1. If a contract for "Democrat wins Ohio Senate 2026" trades at $0.38, the NO contract trades at $0.62. Traders buying YES at $0.38 are paid $1 if the event occurs and $0 if it doesn't, giving them a profit of $0.62 per share in the win case.

Solving for the fair price that makes expected value zero gives price = P(YES). So the price literally equals the market's collective probability estimate. There's no extra translation — the number you see is the number.

Categorical markets

For races with more than two outcomes — say a primary with eight candidates — each candidate gets their own YES/NO contract, all independent. Prices across all candidates sum to roughly $1 when the market is efficient. If they don't sum to $1, there's an arbitrage opportunity in principle (in practice, fees and thin liquidity usually eat the gap).

Fees and spread

The "fair" implied probability is the midpoint between the bid and ask. For thinly-traded markets, that spread can be wide — 5 to 10 cents isn't unusual — so take widely-spread prices with a grain of salt. A $0.42 ask on a market with a $0.38 bid is consistent with anything from 38% to 42% true probability.

On Kalshi, fees are explicit: a small fee (typically 1 to 10 cents per contract depending on size) eats into expected value. Adjust your interpretation accordingly.

Calibration — is the 42% actually 42%?

A good forecast is well-calibrated: of all events you price at 42%, 42% should actually happen. Polymarket and Kalshi both publish calibration curves, and independent researchers have studied this.

The short answer: prediction markets are well-calibrated at the aggregate level, especially for liquid markets. The 2024 US presidential election is the most famous example — Polymarket priced Trump above 60% in the final week while traditional polling models had Harris around 50/50. Post-result, that implied probability was closer to correct than any major polling aggregator.

But markets are not infallible:

  • Thin-liquidity markets drift based on a few traders, not wisdom-of-crowds.
  • Markets can be manipulated — a single deep-pocketed trader can move price, especially in illiquid markets.
  • Resolution ambiguity can mis-price edge cases (what exactly counts as "the government shut down"?).
  • Tail events (99% favorites that still happen 1% of the time) are where markets can look "wrong" even when perfectly calibrated.

Practical reading guide

  1. Big, liquid markets (US presidential, major senate races with $10M+ volume): treat implied probability as a credible consensus estimate.
  2. Medium liquidity ($500k to $5M volume): directionally correct but noisy. Cross-check against polling.
  3. Thin markets (under $200k volume): interesting signal, not authoritative.
  4. Always note the spread. A 5+ cent bid-ask spread means the "probability" is really a range.
  5. Time horizon matters. Markets 12 months out are priced with enormous uncertainty. Prices 1 week out should be dramatically tighter.

How we use these numbers

On every race and market page, we cite the Polymarket or Kalshi implied probability as a reading, not the reading. We cross-reference 538, RealClearPolitics, Cook Political Report, and expert forecasters. When markets and polls diverge, that divergence is itself newsworthy — see our Polymarket vs polls methodology page.

We do not give betting advice. We do not say "Polymarket has Trump at 42%, so buy YES." We say "Polymarket has Trump at 42%, up from 38% last week, driven by [event]." The reader decides what to do with it.

Want to explore?