Pre-registered
Everyone quotes the odds. We keep score.
The Scopes Fair Value is our independent estimate of the odds; a divergence is where it parts from the market price.
Collecting 562,688 snapshots since July 2026.
735 divergences resolved.
When market price and Scopes Fair Value disagreed:
Market right · 387Fair value right · 348
52.7%
47.3%
Still within chance (n = 735, p ≈ 0.16) What does this split mean? →See the record →
Methodology
The record holds 998 resolved flags, scored as 735 distinct events. Set aside from scoring: Sports 226, energy 20, stocks 17. How this number is computed →
conflicts · versioned · append-only · archived · verifiable · last ingest 28s ago
Every flagged divergence carries independent reads, model and human, recorded before the outcome and scored after. See the standing record →
Published before the answer was known.
Every Scopes study is registered with its tests and verdict templates before the outcomes it measures exist. The registration receipts are on each page.
Pre-registered
Pre-registered
Pre-registered
Published
Does context make a model commit?
Given only two probabilities the models abstained; given a public NWS forecast they committed. The effect is on whether a model answers, not on its accuracy.
Scopes is administered by Jon Shields, CFA. The methods are published, the record is open, and every figure on this site can be recomputed by anyone who disagrees with it. About the administrator →
What you can do with it
One record of every venue price. Four things we do with it.
One is self-serve today; the rest are available by engagement. Start with Reference Reports.
PublishPublished
Calibration IndexThe independent benchmark of how well prediction-market prices are calibrated against outcomes.Monthly · free to cite · series licensedRead the methodology →MeasureBy engagement
Market QualityHow a series behaves: volume, spread, adverse moves, stale quotes, dispersion, and what it cost. Per series.Per series · page, PDFSee the method →ProveStart hereLive now
Reference ReportsWhat the market was at any moment, with a proof anyone can check.Self-serve · page, PDF, JSONRequest a report →ScoreBy engagement
VerificationAnyone who publishes a probability, scored on the record under published rules.Per engagement · published rulesSee a sample →The Scopes Calibration Index
How much better venue prices did than a naive base-rate guess, scored against outcomes by desk and horizon. The Scopes Fair Value is not scored.
SCI-ALL, H1, trailing
0.445
backfill · July 15 to October 31, 2026
- Sports0.13990% [0.105, 0.173] · 1474 obs · 753 events
- Energy0.61490% [0.513, 0.694] · 853 obs · 28 events
- Stocks0.58290% [0.046, 0.615] · 227 obs · 15 events
- Equal-weight mean of the desks that clear the publication gate: at least 30 observations and 10 distinct events. Desks differ in how predictable their contracts are, so compare each with itself over time, not with each other.
BACKFILL: mechanically calculated using methodology v1.0, but the rules were finalized after the outcomes were known. LIVE and BACKFILL →
The Week in Review
A free weekly recap of the biggest market-vs-fair-value moves across every desk. Read the latest →
We keep score, and we publish the ones we get wrong.
Every figure on the record can be recomputed from the timestamped data behind it. That is the whole point.
Or put it to work:Cite itQuery or license the dataCommission a benchmark