When Do Prediction Markets Resolve Uncertainty? Belief Variation During Trading and Resolution Jumps on Polymarket
Author: Seiryu Ando (Graduate School of Economics, Kyoto University)
Keywords: Prediction Markets, Polymarket, Information Aggregation, Event Uncertainty, Realized Belief Variation, Resolution Jump
Research Question
Do prediction-market prices gradually incorporate information while trading is active, or does substantial uncertainty remain until the outcome is officially resolved? This study addresses that question by separating each market's price variation into two components:
- Trading realized variation (Trading RV): price variation while the market is open
- Resolution jump: the movement from the last pre-resolution price to the final outcome
Comparing the two components describes how much uncertainty the market resolved during trading and how much remained until resolution.
Data and Method
The analysis combines price histories and market attributes for closed Polymarket markets retrieved as of July 3, 2026.
| Analysis stage | Markets or events |
|---|---|
| Closed markets retrieved | 1,578,120 markets |
| Eligible population | 540,162 markets |
| Reproducible stratified sample | 30,000 markets |
| Unique events in the sample | 21,281 events |
| Markets with usable daily histories | 11,531 markets |
| Unique events in the daily analysis | 8,807 events |
The 30,000-market sample uses a fixed seed and proportional allocation across a missing-volume stratum and four observed-volume quartiles. Stacking multiple horizons for the daily analysis produces 23,621 market-horizon observations.
Main Measures
- JumpShare: the share of total variation attributable to the resolution jump
- TradeShare: the share of total variation attributable to realized belief variation during trading
The two measures sum to one, allowing a direct comparison of whether information appeared in prices during trading or at resolution.
Main Results
Among markets with usable daily price histories, meaningful variation appears both during trading and at resolution.
| Measure | Result |
|---|---|
| Mean Trading RV | 0.0686 |
| Mean resolution jump | 0.0692 |
| Mean total variation | 0.1378 |
| Mean JumpShare | 0.3903 |
| Median JumpShare | 0.1715 |
| Mean TradeShare | 0.6097 |
JumpShare depends strongly on the observation horizon. On the same rows, it is 0.1663 at the 30-day horizon, 0.3403 at 12 hours, and 0.2163 at 6 hours. Evaluations of when uncertainty is resolved therefore need to state both the time window and the price frequency being used.
Scheduled or official events have the highest JumpShare, while narrative or social events have the lowest. The analysis also does not find clear evidence that greater cumulative market volume reduces JumpShare. Total market volume is therefore interpreted cautiously as a measure that may reflect market size and attention as well as liquidity.
Practical Implications
- Evaluating a prediction market requires more than asking whether the final forecast was correct; it also requires examining how much uncertainty was reduced before resolution.
- Events settled by an official announcement can retain material resolution risk even when trading is active.
- Decision systems and event-linked products that use prediction markets should treat the terminal resolution jump as a distinct risk.
Interpretation Caution
JumpShare is a descriptive measure conditional on a specified horizon, sampling frequency, and the availability of usable price history. It is not an unconditional population average across all 540,162 eligible markets.
The decomposition describes when price variation occurs. It does not establish that event categories or market volume causally change JumpShare. Differences between markets with and without usable price histories also need to be considered.
Paper
When Do Prediction Markets Resolve Uncertainty? Belief Variation During Trading and Resolution Jumps on Polymarket
Posted on SSRN: July 24, 2026