Proof Center
Every pick, graded in public.
This page is computed live from the same database our picks are written to — nothing on it is typed in by hand. Accuracy alone is not enough, so we also show calibration (were the probabilities honest?) and value (would betting the picks have made money?). We show the misses too.
Loading latest data…
Picks are locked before events and never edited after. Every model pick is written to the database ahead of fight night with its probability and the opening line at the time. Results are graded in afterward — the pick itself is insert-only and can't be revised.
Crunching the full pick history…
Accuracy by year
Straight-up picks (the side the model gave over 50%), grouped by the year the fight happened. The gray tick marks 50% — a coin flip.
Calibration — did the probabilities tell the truth?
When the model says 65%, those picks should win about 65% of the time. Gray bar = what the model predicted on average, green bar = what actually happened.
PredictedActual
Favorites vs underdogs
Split by the opening betting line. Beating the market on underdogs is where the money is — favorites winning often is expected.
Return on investment — flat $100 per pick
What a $100 bet on every pick above each edge threshold would have returned at the opening line. Edge = model probability minus the market's implied probability.
Where the model has been wrong
The most recent graded misses, with the price we recorded at pick time. A model that only shows winners is selling something.
All models, same test
Every model measured on the identical out-of-sample window — fights from Jan 1, 2025 onward that none of them trained on. Apples to apples.
Closing line value
The sharpest test of a betting model: do our picks beat the price the market closes at?
We started capturing closing lines from a 7+ book consensus in June 2026. That sample is still too small to mean anything, so we're not going to dress it up as proof — CLV numbers will be published here once 100+ locked picks have both an opening and a closing line on record. Until then, judge us on the graded record above.
How to read this honestly
- Accuracy alone is not enough. Picking favorites gets you to ~65% for free. Calibration and value against the line are the real tests — that's why they're on this page.
- Loading pick provenance…
- Model output, not guarantees. A 70% pick loses 3 times out of 10. Nothing here is betting advice.
- Samples matter. Anything under ~100 picks in a bucket is noise. We show the counts so you can discount accordingly.
The record is the pitch.Put the model to work on the next card — every pick, edge, and price in one place.
Model probabilities are statistical estimates, not guarantees. Historical performance does not predict future results. Cannon Fight Lab is an analytics publication, not a sportsbook, and nothing on this page is betting advice. 21+ only.