This page is a live audit trail — each row is a timestamped prediction paired with the real outcome. Explore the record by league, filter for hits or misses, and judge calibration for yourself.
PREDICTIONS SCORED
—
Completed matches with known outcomes
PEAK ACCURACY (HIGH)
—
When the model is most confident · TT-Elite
EARLIEST RECORD
—
Start of tracked history
LAST UPDATED
—
Most recent match processed
METHODOLOGY · EVERY SPORT · STATED VS. REALIZED
How this number is measured
"HIGH confidence" means the model assigned a probability above 80% to the outcome it predicted. "Stated" is the average of those predicted probabilities; "Realized" is how often that prediction actually won. A well-calibrated model keeps the two close. Predictions are scored only after the match is final and the outcome is known — no prediction is graded against a result it could have seen in advance. Each sport reports the accuracy of a single named model revision (never a blend of retired and current models), and a draw counts as a miss against a team-win call in every three-way market. Figures below cover every prediction in the record, not a filtered subset.
SPORT
STATED
REALIZED
GAP
HIGH-TIER SAMPLE
LOADING…
GAP = STATED − REALIZED · A LARGE GAP MEANS OVERCONFIDENCE, NOT AN EDGE
FILTER LEAGUE
CONFIDENCE TIERS · TT-Elite · BEST-PERFORMING LEAGUE
Calibration check
HIGH ≥ 80% · MED 60–80% · LOW 50–60%
A well-calibrated model wins more often when it is most certain. High-confidence forecasts should consistently outperform medium and low tiers.
High = model >80% sure · Medium = 60–80% · Low = 50–60%
ACCURACY OVER TIME · 7-DAY ROLLING · BY CONFIDENCE TIER
Rolling accuracy
HIGH ≥ 80% · MED 60–80% · LOW 50–60%
7-day rolling accuracy by confidence tier. A widening gap between High and Low confirms the model is well-calibrated — it wins more when it is most certain.
NO DATA
ROLLING ACCURACY · BY MONTH · FORWARD-CHAINED
Accuracy by month
OVERALL AVG · —
NO DATA
RAW LOG · EVERY PREDICTION RECORD
Prediction audit trail
Every row is one match. "Fav %" is the probability assigned to the predicted winner. "Source" is the model version that produced the forecast.
0 ROWS · PAGE 1
WHEN
LEAGUE
MATCH
FAV %
PREDICTED
ACTUAL
HIT?
SOURCE
Unexpected token '<', "<!DOCTYPE "... is not valid JSON