A
First-per-question scoring
The headline uses the first production forecast recorded for each resolved question. Later reruns are not substituted into the headline result.
Historical track record
A dated record, not a live performance claim.
This page preserves the public scorecard produced by Cenva’s previous forecasting system. It is frozen while a replacement forecasting core is integrated.
As of13 July 2026
Scoring viewFirst forecast per resolved question
StatusFrozen prior-production archive
Prior production archive
These results do not describe the forecasting core now being integrated. No live forecasts, active-update feed, or current-production cohort is shown on this website during the transition.
01 / Scorecard
Prior cohort at a glance.
Lower Brier scores are better. The market comparison below uses only questions for which a valid market baseline existed.
Resolved questions 544 First forecast per question
Average Brier 0.232 All scored prior-production questions
Comparable subset 147 Questions with a market baseline
Cenva / Market 0.293 / 0.285 Comparable subset; market was better
B
Comparison on matched questions only
Overall Cenva Brier and market-comparison Brier use different denominators. A market comparison is shown only for the subset with an available baseline.
C
No market-beating claim
On the comparable subset, the market score of 0.285 was lower than Cenva’s 0.293. The archive should be read as evidence of measurement discipline, not outperformance.
D
Version boundary
Results from the replacement core will begin a separate cohort. Prior and current systems will not be pooled into one headline number.
03 / Cohort notes
The denominator is part of the result.
Scorecards are only interpretable when the included questions, exclusions, and version policy are explicit.
Tweet-count markets
These markets were treated as out of scope for production forecasting. The exclusion remains disclosed with the archived scorecard.
The overall Brier score includes questions without a market baseline.
Market comparison uses only the 147 questions with a valid market baseline.
The prior system did not outperform the market on the comparable subset.
Continuous update-loop metrics were still immature and are not presented as headline data.
Why the old live feed is gone
A live-looking page tied to a retired engine would create false recency. The archive is static by design; the next current scorecard will launch only with a stable version and cohort policy.