Original data, built 6 November 2026
Markets priced at 90% or more a week out came true 96% of the time
We scored 3,000 resolved markets on Polymarket and Kalshi against what actually happened. A week before close they scored 0.164 on the Brier score, which measures how far off a price was: 0 is perfect and always saying 50% scores 0.25. Everything below is measured from public price data, and the full table is a free CSV.
Markets that said 90% or more a week out came true 96% of the time, and markets under 10% came true 4%.
- resolved markets scored
- 3,000
- Brier score a week out, where a coin flip scores 0.25
- 0.164
- of markets priced 90% or more a week out resolved yes
- 96%
- of markets priced under 10% a week out resolved yes
- 4%
Across 70 markets priced at 90% or more a week before close, 3 lost. Across 425 markets priced under 10%, 16 won.
A market that said 70 to 80% a week out came true 73% of the time.
Pick a platform, a category and a day
When the market said 70 to 80%, it happened 73% of the time. 59 markets
1,437 markets · Brier 0.164 · coin flip 0.250
| Price bucket | Happened | Markets |
|---|---|---|
| 0 to 10% | 4% | 425 |
| 10 to 20% | 27% | 135 |
| 20 to 30% | 32% | 151 |
| 30 to 40% | 39% | 148 |
| 40 to 50% | 43% | 187 |
| 50 to 60% | 52% | 122 |
| 60 to 70% | 57% | 91 |
| 70 to 80% | 73% | 59 |
| 80 to 90% | 76% | 49 |
| 90 to 100% | 96% | 70 |
Economy markets were priced closest to the outcome and sports markets furthest.
How far off, by category, a week before close
- Economy · 101 markets0.067
- Culture · 69 markets0.088
- Politics · 276 markets0.101
- Crypto · 76 markets0.104
- Sports · 893 markets0.206
Coin flip 0.25
A week out, economy scored 0.067 across 101 markets and sports scored 0.206 across 893. Lower is better.
Polymarket prices were closer to the outcome than Kalshi prices in all 4 categories where both platforms had 15 or more markets.
Polymarket and Kalshi, a week before close
| Brier score | Polymarket | Kalshi |
|---|---|---|
| Politics | 0.097247 markets | 0.13629 markets |
| Sports | 0.190581 markets | 0.236312 markets |
| Economy | 0.06185 markets | 0.09816 markets |
| Culture | 0.04041 markets | 0.15928 markets |
| All markets | 0.1441,035 markets | 0.214402 markets |
Prices got sharper as close approached, with the same 463 markets scoring 0.095 at 30 days and 0.054 at 1 day.
How far off, 463 markets scored at every distance
| Same markets | 30 days | 7 days | 1 day |
|---|---|---|---|
| Polymarket · 352 | 0.081 | 0.059 | 0.042 |
| Kalshi · 111 | 0.141 | 0.123 | 0.092 |
Only 3 of the 70 markets priced 90% or more a week out lost.
- Burnley at Arsenal: Totals, Over 1.5 goals scoredKalshi · closed 18 May 202698%
- Inflation in Nov 2025? (CPI YoY), Above 2.8%Kalshi · closed 18 December 202595%
- Paraguay vs Australia: Total Goals, Over 0.5 goals scoredKalshi · closed 26 June 202691%
Here is exactly how we counted.
- Window. Markets that closed from 6 November 2025 to 6 November 2026, resolved yes or no. Voided and ambiguous markets are left out.
- Floor. At least $50,000 traded, and open for at least 3 days. Shorter markets, such as 5-minute crypto, are left out.
- Pool. 41,336 Polymarket markets and 108,541 Kalshi markets met the floor. Kalshi does not publish dollars traded, so its pool is counted in contracts and the dollars of each drawn market are estimated from its price history.
- Sample. A seeded random draw of up to 1,500 per platform, seed 20261106. 1,500 Polymarket and 1,500 Kalshi markets were scored. 3,271 drawn Kalshi markets came in under $50,000 once estimated, and the draw moved on to the next one.
- Price. The last traded price 1, 7 and 30 days before close, left out where the market did not exist yet.
- Gaps. 1,035 Polymarket and 402 Kalshi markets have a price 7 days out, and 353 and 113 have one 30 days out, because many were open for less time.
- Mix. Sports are 68% of the scored markets, so the all-market scores lean towards sports.
- Score. Brier is the average squared gap between the price and the outcome, 0 for perfect.
To cite this study, use the line below.
BitPredict, "Prediction market accuracy", built 6 November 2026. https://www.bitpredict.io/prediction-market-accuracy
Frequently asked questions
What is a Brier score?
It is the average squared gap between a market's price and what happened, where a yes is 1 and a no is 0. 0 is perfect, and always saying 50% scores 0.25.
How many markets were scored?
3,000: 1,500 on Polymarket and 1,500 on Kalshi, drawn at random from the markets that met the floor.
Can I use this data?
Yes. The CSV is free to download at https://www.bitpredict.io/data/prediction-market-accuracy.csv, with one row per scored market. Link back to this page when you use it.
