Nasty Board

slate: 504 players, season 2026 week 1 ============================================================================== A. DO NOT BET — high-usage overs [-9.8% 2024, -17.0% 2025] ============================================================================== no high-usage overs on this slate — nothing to decline. ============================================================================== B. CEILING PLAYS — spike candidates [PRICEABLE, PROFIT UNTESTED] ============================================================================== Shortlist for the longest-reception / longest-rush markets, which are not in odds_client.MARKETS and have never been bought. Player Pos Tm Opp Stat p50 p90 p90/p50 ------------------------------------------------------------------------------ Xavier Smith WR LA SF REC_YDS 9.0 47.0 5.22 Jimmy Horn Jr. WR CAR CHI REC_YDS 8.0 38.0 4.75 Terrance Ferguson TE LA SF REC_YDS 9.0 41.0 4.56 Jacob Cowing WR SF LA REC_YDS 9.0 41.0 4.56 Bo Melton WR GB MIN REC_YDS 9.0 40.0 4.44 Ty Johnson RB BUF HOU REC_YDS 9.0 39.0 4.33 Joshua Dobbs QB DET NO RUSH_YDS 6.0 26.0 4.33 Dont'e Thornton Jr. WR LV MIA REC_YDS 10.0 43.0 4.30 Tutu Atwell WR LA SF REC_YDS 10.0 43.0 4.30 Josh Oliver TE MIN GB REC_YDS 8.0 34.0 4.25 Isaac TeSlaa WR DET NO REC_YDS 9.0 38.0 4.22 Jordan Whittington WR LA SF REC_YDS 9.0 38.0 4.22 Tim Patrick WR NYJ TEN REC_YDS 10.0 42.0 4.20 Luke McCaffrey WR WAS PHI REC_YDS 11.0 46.0 4.18 Jared Wayne WR HOU BUF REC_YDS 10.0 41.0 4.10 Jaylin Noel WR HOU BUF REC_YDS 12.0 48.0 4.00 Carson Wentz QB MIN GB RUSH_YDS 6.0 24.0 4.00 Emari Demercado RB DAL NYG RUSH_YDS 8.0 32.0 4.00 Justice Hill RB BAL IND RUSH_YDS 6.0 24.0 4.00 Sam Darnold QB SEA NE RUSH_YDS 6.0 24.0 4.00 Quinshon Judkins RB CLE JAX REC_YDS 8.0 32.0 4.00 Woody Marks RB HOU BUF REC_YDS 9.0 35.0 3.89 Caleb Douglas WR MIA LV RUSH_YDS 8.0 31.0 3.88 Cyrus Allen WR KC DEN RUSH_YDS 8.0 31.0 3.88 Malik Benson WR LV MIA RUSH_YDS 8.0 31.0 3.88 ============================================================================== C. FAVORITE UNDERS — 7+ point favourites [NOT SUPPORTED ON 2025] ============================================================================== Watch list only. Unders by own-team spread returned +2.0% at 10+, -3.5% at 4-9.5, +1.7% as a dog — no gradient. Do not bet this section off this section. See the module docstring. Player Pos Tm Stat Line p_under Spread ------------------------------------------------------------------------------ KC Concepcion WR CLE REC 3.5 68.8% -7.5 Denzel Boston WR CLE REC 2.5 66.5% -7.5 KC Concepcion WR CLE REC_YDS 33.5 64.6% -7.5 Tyler Shough QB NO RUSH_YDS 16.5 61.5% -7.0 Quinshon Judkins RB CLE RUSH_YDS 59.5 60.8% -7.5 Juwan Johnson TE NO REC 4.5 60.5% -7.0 Deshaun Watson QB CLE RUSH_YDS 19.5 60.3% -7.5 Harold Fannin Jr. TE CLE REC 4.5 58.6% -7.5 Denzel Boston WR CLE REC_YDS 23.5 56.4% -7.5 Marvin Harrison Jr. WR ARI REC_YDS 49.5 54.4% -10.5 ============================================================================== D. LONGSHOT TOUCHDOWNS — +800 or better [REPLICATES 2024+25, WK 2+] ============================================================================== Betting these blindly loses. Across 12,276 legs of real closing prices, anytime-TD overs returned -12.77% in 2024 and -9.45% in 2025, and the deep longshots were the worst of it (-30.9% and -27.0% in 2024's +800-1600 and +1600+ buckets). The favourite- longshot bias is real and it is against you. What survives is SELECTION inside the bucket. Graded on 2024 AND 2025 closing prices, volume models refit per season and touchdown history held to strictly prior seasons, taking every +800-or-better leg where the model's probability beats the price (weeks 2-18): n=2,283 pooled +45.1% 2024 +33.9% 2025 +55.8% z +3.87 Positive in both seasons, and it sharpens as the price lengthens: +400 floor +28.3%, +800 +45.1%, +1200 +64.2%. The control is what makes it credible: the same edge>0 rule on legs SHORTER than +800 returns +0.3% (z +0.52). This is not the model being good in general, it is longshot touchdown prices specifically failing to hold the information the model has. The cut is edge>0, NOT the biggest edges. The top decile of edge went -7.5% in 2024 and +116.1% in 2025 — the largest model-minus- market gaps are where the model is most overconfident. The FADE side is the more consistent half: +800-or-better legs with edge<0 hit 2.01% against a 6.12% breakeven, -77.0% (-75.7% 2024, -78.4% 2025). Those are hard avoids. SUPPRESSED — 100% of this slate is cold start. The rule above was graded on weeks 2-18. Applied to week 1 it loses in BOTH seasons: -35.3% pooled (2024 -50.5%, 2025 -21.8%, n=204, z -2.15) against +45.1% in weeks 2-18. That is what the calibration work predicts — with no in-season usage the model's ranking inside a longshot bucket is the part with least in it. This section turns itself back on once players have games. (12 legs would have been listed. The FADE half is the more durable signal even here, but it was not graded separately on week 1 either, so it is withheld too.)