The Record
We show our work, losses included.
Every pick is logged before it plays and graded in public. This is the whole receipt drawer: how often the model is right, whether it beats the closing line, how close its DFS projections land, and how its draft board has held up over a replayed decade. No cherry-picking, and the rules we grade ourselves by are written down on how we grade.
62.2%
Documented accuracy
255/410 graded picks · last 30d
+1.16%
Avg closing-line value
across 890 graded picks
7-3
Draft board vs the market
10 replayed drafts, losses published
A pick we ruled invalid before it played is taken out of the record with the reason written here. It is never rewritten.
- 2026-09-13 nfl MIA at LV: picked MIA, LV won. Model ran on the 2025 Dolphins, blind to the quarterback change (Malik Willis starting for Tua Tagovailoa); ruled invalid and flagged before kickoff on Sep 13, 2026 (11:20 AM ET, kickoff 4:05 PM). The owner's own pick was LV.
Game predictions
Model win-probability accuracy, last 30 days.
| League | Accuracy | Graded | Avg conf. | Brier |
|---|---|---|---|---|
 MLB | 57.1% | 21 | 55.4% | 0.2506 |
 NBA | 69.1% | 123 | 65.1% | 0.2011 |
 NFL | 59.6% | 57 | 63.5% | 0.2273 |
 NHL | 59.3% | 209 | 57.8% | 0.2301 |
Brier score measures calibration (lower is better; 0.25 = a coin flip). How the model works → · Browse day-by-day results →
Closing-line value
Did the market move toward our pick after we logged it? Positive = we beat the close.
MLB
+0.63%
59.3% positive · n=383
NBA
+0.83%
49.2% positive · n=124
NFL
+4.29%
64.8% positive · n=122
NHL
+0.64%
50.6% positive · n=261
CLV accrues statistical signal faster than win/loss, it's the honest early read on whether the model moves markets' direction.
DFS projections
Projected vs actual DraftKings points.
| League | MAE | Corr. | Floor hit | Ceiling hold | Graded |
|---|---|---|---|---|---|
| MLB | 4.58 | 0.4 | 56.1% | 96.8% | 495 |
| NBA | 9.69 | 0.713 | 72.8% | 93.9% | 2519 |
| NFL | 4.59 | 0.663 | 66.3% | 90.3% | 1197 |
| NHL | 4.68 | 0.368 | 64.3% | 91.4% | 7661 |
MAE = average points off per player. Lower is better; a mid-single-digit MAE with near-zero bias is a genuinely sharp projection.
Draft board
A decade of replayed drafts, graded against actual fantasy production.
7-3
vs ADP over 10 drafts (PPR)
+0.779%
avg edge vs the market
+2.49%
2024 (latest graded)
The 10-year extension showed the model is market-neutral over the full decade with its wins concentrated recently, we publish that honestly, not a cherry-picked window. See the live board →
This is the whole point.
Anyone can post winners after the fact. Verdexed logs every call before it plays and grades it in public, so you never have to take our word for it.