Prediction markets › Accuracy
How accurate are crypto prediction markets?
If markets priced at 70% happen about 70% of the time, their odds are well calibrated. Here is that check on every resolved crypto market we track.
One day before close
| Market said | Markets | Avg. predicted | Happened |
|---|---|---|---|
| 0–10% | 3,281 | 1.3% | 1.8% |
| 10–20% | 228 | 14.3% | 20.6% |
| 20–30% | 157 | 25.4% | 34.4% |
| 30–40% | 147 | 35.4% | 32.0% |
| 40–50% | 269 | 45.6% | 49.4% |
| 50–60% | 413 | 52.8% | 61.7% |
| 60–70% | 130 | 64.3% | 70.0% |
| 70–80% | 150 | 74.6% | 78.7% |
| 80–90% | 139 | 85.2% | 89.2% |
| 90–100% | 1,198 | 98.6% | 99.7% |
By market type (Brier, lower is better): Price above 0.035 (n=2544) · Price range 0.072 (n=1647) · Price reaches 0.111 (n=516) · Price dips to 0.025 (n=410) · Up or down 0.259 (n=405) · Other 0.051 (n=344) · Price below 0.000 (n=127) · Token launch 0.000 (n=88) · Airdrop 0.005 (n=16) · FDV after launch 0.053 (n=7) · Listing 0.000 (n=5) · ETF 0.250 (n=3)
One week before close
| Market said | Markets | Avg. predicted | Happened |
|---|---|---|---|
| 0–10% | 486 | 1.7% | 3.7% |
| 10–20% | 45 | 14.8% | 26.7% |
| 20–30% | 35 | 24.4% | 25.7% |
| 30–40% | 37 | 35.6% | 43.2% |
| 40–50% | 18 | 43.6% | 50.0% |
| 50–60% | 36 | 54.1% | 77.8% |
| 60–70% | 24 | 63.4% | 87.5% |
| 70–80% | 26 | 74.8% | 92.3% |
| 80–90% | 27 | 84.8% | 100.0% |
| 90–100% | 18 | 94.7% | 94.4% |
By market type (Brier, lower is better): Other 0.084 (n=244) · Price reaches 0.166 (n=221) · Price dips to 0.033 (n=161) · Token launch 0.000 (n=98) · Airdrop 0.008 (n=16) · FDV after launch 0.047 (n=7) · Listing 0.000 (n=5)
Method
For each crypto market that has resolved, we take the last recorded “Yes” price at least 24 hours (and 7 days) before the market closed — the earlier of its end date and the moment it settled — and compare it with the outcome. Prices come from Polymarket's public price history (hourly near the close, daily before that) and from our own 10-minute snapshots since 28 Sep 2026. Resolved markets were seeded from the 90 days before that date; very short up/down markets (5 minutes to 4 hours) are excluded. Markets with no price in the window are left out rather than guessed, and so are markets with under $1,000 traded or a price from their first 12 hours — a new market's order book shows about 50% before anyone trades, which is not a forecast.
Past calibration says nothing certain about any single market. Nothing here is financial advice.