Skip to content

Essential storage, first-party measurement, cookieless Google Tag Manager. Sentry, visit log off unless you Accept all.

Settings
$1 first week
nfl v1.0.0

divisional-dog-spread

NFL · custom

Unverified
v×1
Backtest 51.3% ATS · n=483, training backtest Backfill · training backtest · n=483

Training backtest published Jul 5, 2026

0 forward picks — record starts at kickoff

No settled CLV yet.

A model built on the thesis that divisional underdogs are sharper than the market — familiarity compresses margins toward the dog; trained with bayesian ridge and fixed 1u@1.0.0 staking, sliced to divisional games@1.0.0, will beat the closing spread on a held-out eval window. Kill criterion: ROI below 0 vs the close after 200 graded bets.

0 tailing
Sign in to tail this model

Test games

The graded games behind this model's W-L-P.

No graded test games yet for this model.

Season heat

Weekly wins and losses by game date. Gray weeks had no decisions.

live pregame n=0 model scheduled game dates
No dated graded decisions in this window.

Decisive live-pregame outcomes grouped by scheduled game date; pushes and unavailable dates are excluded.

Receipts

0 loaded · 0 qualifying settled rows

No settled predictions are available yet.

⑂ Remix this model

Turn knobs to put your own spin on divisional-dog-spread. Your settings are saved with the remix and used when it trains.

Try it free

No graded record yet

This model has not been graded on any games, so there is nothing to simulate your knobs against. Save your spin, then train it to get a record.

Lens

Balanced

Every graded game.

Lens
Market slice
Train window
Engine

Proof before save

  1. 1. Parent release: v1.0.0 · version 60ab3d12-4341-4365-a9f5-81975f72356f · released 2026-07-05T21:15:30.33+00:00
  2. 2. Requested changes: none yet
  3. 3. Simulation: No real simulated result is available for this request.
  4. 4. Child result: unavailable until the saved draft is trained and evaluated.
Sign in to save your spin

Draft tracking

No future draft picks are declared in this model's latest version.

Tune it

Pick a dial to see how tuning this model changes its win rate on the same graded games.

No settled calibration samples are available for this model yet. Return after picks settle.
No distribution yet — no graded samples in this window.

Inputs

No feature slugs declared on this card.

Assumptions

  • A model built on the thesis that divisional underdogs are sharper than the market — familiarity compresses margins toward the dog; trained with bayesian ridge and fixed 1u@1.0.0 staking, sliced to divisional games@1.0.0, will beat the closing spread on a held-out eval window. Kill criterion: ROI below 0 vs the close after 200 graded bets.

Known limitations

  • Auto-generated by nflbets Lab from grind ledger evidence; inspect benchmark artifacts before making supported thesis claims.

Version history

  1. v1.0.0 · published Released Jul 5, 2026

Lineage

No ancestors or forks. This model stands alone — yet.