Polymarket vs Metaculus
Polymarket vs Metaculus
Metaculus is a forecasting tournament where reputation comes from calibrated Brier scores; Polymarket is a traded venue where the same belief is expressed as a bid that must be filled. The quantitative products are different in kind: Metaculus gives you distributions, Polymarket gives you a book.
Figures measured as of 2026-09-24 on the published PolyOrderbooks archive.
Short answer
Scores versus trades
Metaculus collects probabilistic forecasts, aggregates them, and scores each forecaster with strictly proper scoring rules; no money changes hands. Polymarket collects bids and asks on a real order book and settles them to USDC at a resolution boundary. Both are forecasting instruments; only one produces fills.
The researcher's angle: Metaculus is the best free distributional dataset in the category, while Polymarket is the best data on what committing capital does to a price. The two are complements, which is why quantitative forecasts often quote both.
Resolution
How questions close
Metaculus resolutions are adjudicated by its own resolution procedures and published with explanations. Polymarket crypto Up/Down contracts resolve off a Chainlink reference price at a predetermined timestamp, so resolution is part of the data row — every archived book carries the winning outcome.
The advantage of reference-based resolution is reproducibility: a study can replay the final minute against the stored books, as the settlement corpus of 298 resolved BTC contracts does here, and know exactly when and why each market closed.
Depth
Distribution versus ladder
Metaculus expresses belief as a distribution over outcomes, which is ideal for risk analysis. Polymarket expresses belief as two orders — and on this venue those orders are frequently one-sided: 16.9% of 5-minute snapshots carry a single side, so the price is often a floor with nobody bidding the other way.
That structure is a feature for a trader and a warning for a forecaster: the aggregation is thinner and faster than a Brier-weighted ensemble, and any model importing Polymarket prices must read the executable side, not the mid.
Data
Data access compared
Metaculus offers an open API with forecasts, comments, and resolutions ideal for research. Polymarket's official API is free for live state and 1-minute price history; recorded 250ms depth with prices and metrics on the same grid is available from the archive here, downloadable as CSV, JSON, or Parquet.
For a paper that needs both, the practical merge is on the question identifier and the UTC second: Metaculus holds the distribution, the archive holds the book, and the two tell the same story from different instruments.
Honest pick
Which to choose
Choose Metaculus to study calibration and aggregate expertise. Choose Polymarket to study commitment, depth, and speed — especially when resolution timestamps and fills matter.
The rigorous answer is to use both: Metaculus for the prior, Polymarket for the traded reality check, and to keep the recorded books as the evidence layer.
FAQ
Does Metaculus use real money?
No, Metaculus is a tournament based on forecasting accuracy scores (Brier and related), while Polymarket trades real USDC on an order book.
Which is better for research data?
Both. Metaculus offers free distributional forecast data; Polymarket offers recorded 250ms order books with prices and metrics, which support fill-level backtesting.
How does resolution differ?
Metaculus adjudicates resolutions with published explanations; Polymarket crypto Up/Down contracts resolve off a Chainlink reference price at a fixed UTC timestamp, which is stored on every archived row.