Scorecard · method

How this number is computed

The scorecard asks one question: when the market and the independent Scopes Fair Value disagree enough to flag, which side proves right? Every figure below is read live from the same record the scorecard and homepage read, as of Oct 19, 6:00 PM EDT.

1.What one resolved divergence is

A flag is one market-vs-fair-value disagreement that crossed the desk's published threshold at the moment it was recorded: the gap is |market − Scopes Fair Value|, and it must clear 5 points on the curated desks (rates, polls, stocks, energy) or 6 points on sports, on a market that is not thin (the liquidity floor). Each flag is stamped with the exact time it was recorded and the benchmark version behind its fair value. A flag becomes a resolved divergence when the event settles and the record appends which side was right. There is one flag per binary event: when an event produces more than one, the widest-gap one represents it.

Worked example: on 10/16/2026, the stocks desk flagged SPX 7725–7749.9999 at a 17.6-point gap. It resolved to the market side. See it in the record →

2.What is counted and what is set aside

Set aside, broken out by the rule that removed it:

Rules apply from their effective date forward and are never applied backward to remove an inconvenient result. A set-aside flag stays in the record, unscored; nothing is deleted.

3.Ties, stale prices, unresolved events

When two outcomes of one event tie on the widest gap, the affirmative (YES) side is taken, so an event never contributes two flags. When a venue price is stale relative to the benchmark at flag time (the pregame or pre-settlement line no longer describes the same event), the flag is marked not comparison-valid and set aside, not scored. An event that never resolves, or is voided by the venue, is simply never scored: it carries no side and stays out of the counts until it settles.

4.The test

The test is a two-sided exact binomial test against p = 0.5, on market-vs-Scopes-Fair-Value wins over the scored divergences. The null is that both sides are equally good. At the current n = 558, the Scopes Fair Value side was right 248 times (a 44.4% win share, 95% interval 40.4% to 48.6%), with a p-value of 0.01.

This test compares the market against the Scopes Fair Value on flagged divergences only. It does not establish whether prediction markets are well calibrated in general; the Scopes Calibration Index measures that, under its own methodology.

5.Why the scored set is not chosen after the fact

A flag is generated by rules published before the event resolves, not picked once the answer is known. The record is append-only: resolutions and corrections are appended with timestamps, never overwritten. Every day of the record is fingerprinted and anchored to the Bitcoin blockchain (the integrity ledger). Rule changes are logged with their effective dates in the Oversight Log, and the set-aside counts are shown above, not hidden.

6.Reproduce it

Download the scored rows and the set-aside rows (event, venue, flag time, market price, Scopes Fair Value, resolution, winner, digest date). Venue price fields follow the same rights gate as the rest of the record (withheld without a verified entitlement, with the rights note); the winner column and every Scopes-owned field are always present, so the headline reproduces without them.

Download the scored + set-aside CSV → · verify a day or a row →

Recipe (recomputes the headline from the CSV):

# scored rows only; winner column: "fair" or "market"
scored=$(grep ',scored,' scorecard-scored.csv)
n=$(printf '%s\n' "$scored" | grep -c .)
fair=$(printf '%s\n' "$scored" | grep -c ',fair$')
market=$(printf '%s\n' "$scored" | grep -c ',market$')
echo "n=$n  fair_right=$fair  market_right=$market"
echo "fair win share: $(awk "BEGIN{printf \"%.1f%%\", 100*$fair/$n}")"
# compare n / fair_right / market_right to section 4 above.

7.Versions

The changelog of scorecard rule changes (in-play contamination, the binding artifact, near-settlement stocks and energy, each with its effective date) is the Oversight Log.

Research and information only, not investment advice.