Challenger model

GriffBet

The challenger. Same transparent, market-blended approach as BiffBet, with a starter-neutralized base option, slate-level discipline, and a richer CLV split graded against the sharp close. Measured head-to-head against the champion below.

GriffBet recordWins–losses (–pushes when present) over GriffBet's own graded picks only. BiffBet's record never mixes in.15-20
Graded35
CLV vs sharp closeAverage closing-line value of the blended pick stream vs the sharp close — the primary edge measure at small samples.+2.22%
Beat the close15/37

Head-to-head referee

Same games · records labeled per model · judged on process

BiffBet is the frozen champion and control; GriffBet is the challenger. We compare them on calibration (are the probabilities honest?), CLV vs the sharp close (did we beat the truest price?), and EV monotonicity (do bigger predicted edges actually realize bigger returns?). Each column shows that model's own record for transparency — but at small samples the record is noise, not the verdict.

Small sample — fewer than 100 graded GriffBet bets. These numbers are noisy and not yet a verdict. Treat as a process check, not a ranking.

BiffBet

champion · control

33-30 · BiffBet record · 63 graded

Calibration — blended prob

Brier 0.259Log loss 0.711

Calibration — raw model prob

Brier 0.271Log loss 0.740

Closing-line value

Obtainable (best close)+0.99% · 29/57 beat
Blended vs sharp close-0.22% · 8/21 beat

44 earlier pick(s) had no sharp close (gap, not backfilled).

EV monotonicity

Pred. EVBetsFlat ROIKelly ROI
3%..5%28-28-4.5%-11.0%
5%..8%5-2+30.3%+7.3%

GriffBet

challenger

15-20 · GriffBet record · 35 graded

Calibration — blended prob

Brier 0.219Log loss 0.630

Calibration — raw model prob

Brier 0.240Log loss 0.672

Closing-line value

Obtainable (best close)+4.59% · 23/37 beat
Blended vs sharp close+2.22% · 15/37 beat
Raw model vs sharp close+2.22% · 15/37 beat

EV monotonicity

Pred. EVBetsFlat ROIKelly ROI
3%..5%14-18-13.6%-13.6%
5%..8%1-2-39.3%-23.2%

GriffBet slate

Tue, Jul 21, 2026

St. Louis Cardinals @ Los Angeles Angels · 9:38 PM ET

Los Angeles Angels (home)

Matthew Liberatore vs Walbert Ureña

Projected 4.0 – 5.0 · total 9.0

Fragile edgeLeanLean: a small bet (Adjusted EV 2–5%). Lower conviction; stake capped at 0.5% of bankroll.Projected lineupSharps fade 3.6ppYou're betting against the sharp books' consensus on this side by this many percentage points. Higher = more contrarian to the smartest money (already penalized in Adjusted EV).-108
Raw EVRaw expected value: the model's edge from its win probability vs the price, before any context adjustments. +5% ≈ you'd net 5% of stake per bet long-run if the model is right.+3.7%
Adj EVAdjusted EV: Raw EV after small context haircuts/boosts (sharp support/fade, projected lineup, fragile edge). This is the number used to size the bet.+1.7%
ModelThe model's win probability for this side, after blending toward the market. Used to compute EV.53.8%
MarketThe market's fair win probability for this side — the book's moneyline with the vig (juice) removed.49.5%
KellyRecommended stake as a fraction of bankroll, from the Kelly criterion (quarter-Kelly, then capped by the bet tier).0.5%
ConfModel confidence (0–100): data completeness, pitcher sample size, edge magnitude, and how much the model's components agree.74
CLVClosing Line Value: how much better your bet price was than the closing line. Positive = you beat the close — the key long-run skill signal.+5.0%
SizingsmallKelly 0.5% 1.0%

adj EV 1.7% < 2% but raw EV cleared the pick threshold -> floored to small (selection is by raw EV; haircuts only size)

Show model breakdown ▾

Dynamic adjustments

Base win prob (home): 45.1%

  • Starting pitcherrate H=3.25(statcast_rate) A=4.87(statcast_rate)+10.40%
  • Bullpenrelief FIP H=4.35 A=4.18-0.38%
  • Bullpen availabilityH 2/3 leverage arms available (-0.5%); A 2/3 leverage arms available (-0.5%)+0.00%
  • LineupLINEUP NOT POSTED YETH projected lineup; A projected lineup+0.00%
  • Park / offensePF=101 wRC+ H=95 A=96-0.00%
  • Home field+0.025 home+2.50%
  • Recent formxwOBAcon H season=0.284 | A season=0.409 (normal cap +/-1.5%)+1.50%

EV adjustments

Raw +3.7% → +1.7% (used for sizing)

  • · -1.0pp sharp fade (3.6pp)
  • · -1.0pp fragile edge

Market intelligence

Blend (high · data conf 88): model 59.1% × 0.45 + market 49.5% → 53.8% (home, used for EV)

Open -108 → Close -120 · CLV 5.0%

Sharps (n=3): 50.2% home · Squares (n=3): 49.5% home · S−Sq +0.7pp · dispersion 0.5pp

Detroit Tigers @ Chicago Cubs · 8:05 PM ET

Chicago Cubs (home)

Framber Valdez vs David Peterson

Projected 5.0 – 4.9 · total 9.9

Fragile edgeLeanLean: a small bet (Adjusted EV 2–5%). Lower conviction; stake capped at 0.5% of bankroll.Projected lineupSharps fade 3.8ppYou're betting against the sharp books' consensus on this side by this many percentage points. Higher = more contrarian to the smartest money (already penalized in Adjusted EV).-120
Raw EVRaw expected value: the model's edge from its win probability vs the price, before any context adjustments. +5% ≈ you'd net 5% of stake per bet long-run if the model is right.+3.1%
Adj EVAdjusted EV: Raw EV after small context haircuts/boosts (sharp support/fade, projected lineup, fragile edge). This is the number used to size the bet.+1.1%
ModelThe model's win probability for this side, after blending toward the market. Used to compute EV.56.2%
MarketThe market's fair win probability for this side — the book's moneyline with the vig (juice) removed.52.2%
KellyRecommended stake as a fraction of bankroll, from the Kelly criterion (quarter-Kelly, then capped by the bet tier).0.5%
ConfModel confidence (0–100): data completeness, pitcher sample size, edge magnitude, and how much the model's components agree.73
CLVClosing Line Value: how much better your bet price was than the closing line. Positive = you beat the close — the key long-run skill signal.+0.8%
SizingsmallKelly 0.5% 0.9%

adj EV 1.1% < 2% but raw EV cleared the pick threshold -> floored to small (selection is by raw EV; haircuts only size)

Show model breakdown ▾

Dynamic adjustments

Base win prob (home): 66.1%

  • Starting pitcherrate H=4.90(statcast_rate) A=4.04(statcast_rate)-5.61%
  • Bullpenrelief FIP H=4.02 A=3.88-0.30%
  • Bullpen availabilityH 3/3 leverage arms available (+0.0%); A 3/3 leverage arms available (+0.0%)+0.00%
  • LineupLINEUP NOT POSTED YETH projected lineup; A projected lineup+0.00%
  • Park / offensePF=100 wRC+ H=108 A=97+0.00%
  • Home field+0.025 home+2.50%
  • Recent formxwOBAcon H 30d=0.433/season=0.392 | A 30d=0.315/season=0.345 (normal cap +/-1.5%)-1.50%

EV adjustments

Raw +3.1% → +1.1% (used for sizing)

  • · -1.0pp sharp fade (3.8pp)
  • · -1.0pp fragile edge

Market intelligence

Blend (high · data conf 89): model 61.2% × 0.45 + market 52.2% → 56.2% (home, used for EV)

Open -120 → Close -122 · CLV 0.8%

Sharps (n=3): 52.4% home · Squares (n=3): 52.1% home · S−Sq +0.3pp · dispersion 0.5pp

+ 13 game(s) analyzed and passed (below the EV threshold).