Prediction side track
NFL Edge Model Profile
Public record · not the neutral matchup product
<- Back to matchup board Model setup
Ratings inputAdj EPA/play through W0
Margin scale35.0 pts / adj EPA
Home field1.5 points
Training window2022-2024 regular season
Season to date
2026 Running Record
Through Week 0
- Scope
- The slice of model decisions included in the row.Why it matters: It distinguishes the full slate from narrower confidence subsets.
- Picks
- The number of completed model decisions in the sample.Why it matters: It reveals how much evidence sits behind the displayed rates.
- Against the spread / ATS
- Wins, losses, and pushes measured against the closing point spread.Why it matters: It tracks performance relative to market expectations rather than straight-up results.
- ATS %
- Against-the-spread wins divided by wins plus losses; pushes are excluded.Why it matters: It measures the model's historical spread record while keeping sample size visible.
- SU %
- Share of model-selected teams that won the game straight up.Why it matters: It gives outcome context separate from covering the spread.
- Model MAE
- Mean absolute error between the model's predicted margin and the final margin.Why it matters: Lower values mean the model's score margins were closer to actual results.
- Market MAE
- Mean absolute error between the market-implied margin and the final margin.Why it matters: It is the comparison baseline for judging the model's margin accuracy.
Pushes are tracked separately and excluded from ATS win rate.
| Scope | Picks | ATS | ATS % | SU % | Model MAE | Market MAE |
|---|---|---|---|---|---|---|
| - | 0 | - | - | - | - | - |
Season confidence tiers
- Abs edge / Tier
- The absolute point difference between the model margin and market margin, grouped into bands when shown as a tier.Why it matters: It measures disagreement magnitude without implying that either side is correct.
- Picks
- The number of completed model decisions in the sample.Why it matters: It reveals how much evidence sits behind the displayed rates.
- Against the spread / ATS
- Wins, losses, and pushes measured against the closing point spread.Why it matters: It tracks performance relative to market expectations rather than straight-up results.
- ATS %
- Against-the-spread wins divided by wins plus losses; pushes are excluded.Why it matters: It measures the model's historical spread record while keeping sample size visible.
- Avg edge
- Average absolute model-versus-market margin difference within the row's tier.Why it matters: It describes the typical disagreement represented by that calibration bucket.
Results grouped by absolute model edge. These describe calibration; they are not pick guidance.
| Abs edge | Picks | ATS | ATS % | Avg edge |
|---|---|---|---|---|
| - | 0 | - | - | - |
Current slate
Week 1 Model Slate
Ranked by absolute edge
- Rank
- The row's order when the table is sorted by its primary measure.Why it matters: It makes the strongest or largest values quick to find.
- Game
- The away and home teams in the matchup.Why it matters: It identifies the market and model comparison being scored.
- Edge
- Model away margin minus market away margin, expressed in points.Why it matters: Its magnitude shows how far the model differs from the market, not whether a wager is advisable.
- Market
- The away-team margin implied by the point spread.Why it matters: It provides the public baseline against which the model is compared.
- Model
- The model's predicted away-team scoring margin.Why it matters: Comparing it with the market margin produces the displayed edge.
- Abs edge / Tier
- The absolute point difference between the model margin and market margin, grouped into bands when shown as a tier.Why it matters: It measures disagreement magnitude without implying that either side is correct.
- Model side
- The team and spread corresponding to the model's margin disagreement.Why it matters: It translates the margin difference into the side the model leans toward.
- Result
- Whether the displayed model side won, lost, or pushed against the spread.Why it matters: It closes the loop on completed weekly model decisions.
| Rank | Game | Edge | Market | Model | Tier | Model side |
|---|---|---|---|---|---|---|
| 1 | BAL at IND | -7.5 | BAL by 3.5 | 4.0 vs BAL | 6+ | IND +3.5 |
| 2 | ARI at LAC | +5.1 | 10.5 vs ARI | 5.4 vs ARI | 4-6 | ARI +10.5 |
| 3 | TB at CIN | +5.1 | 3.5 vs TB | TB by 1.6 | 4-6 | TB +3.5 |
| 4 | BUF at HOU | -4.6 | BUF by 1.5 | 3.1 vs BUF | 4-6 | HOU +1.5 |
| 5 | MIA at LV | +4.0 | 3.5 vs MIA | MIA by 0.5 | 2-4 | MIA +3.5 |
| 6 | NE at SEA | -3.9 | 3.5 vs NE | 7.4 vs NE | 2-4 | SEA -3.5 |
| 7 | DAL at NYG | -3.4 | DAL by 2.5 | 0.9 vs DAL | 2-4 | NYG +2.5 |
| 8 | NYJ at TEN | -3.1 | 1.5 vs NYJ | 4.6 vs NYJ | 2-4 | TEN -1.5 |
| 9 | SF at LA | -2.6 | 3.5 vs SF | 6.1 vs SF | 2-4 | LA -3.5 |
| 10 | DEN at KC | +2.6 | 3.0 vs DEN | 0.4 vs DEN | 2-4 | DEN +3.0 |
| 11 | GB at MIN | +2.6 | 1.5 vs GB | GB by 1.1 | 2-4 | GB +1.5 |
| 12 | WAS at PHI | -2.5 | 5.5 vs WAS | 8.0 vs WAS | 2-4 | PHI -5.5 |
| 13 | CHI at CAR | +2.0 | CHI by 2.5 | CHI by 4.5 | 0-2 | CHI -2.5 |
| 14 | ATL at PIT | +1.4 | 3.5 vs ATL | 2.1 vs ATL | 0-2 | ATL +3.5 |
| 15 | CLE at JAX | -1.2 | 7.5 vs CLE | 8.7 vs CLE | 0-2 | JAX -7.5 |
| 16 | NO at DET | -0.3 | 7.0 vs NO | 7.3 vs NO | 0-2 | DET -7.0 |
Frozen training window
Backtest Profile (2022-2024)
- Scope
- The slice of model decisions included in the row.Why it matters: It distinguishes the full slate from narrower confidence subsets.
- Picks
- The number of completed model decisions in the sample.Why it matters: It reveals how much evidence sits behind the displayed rates.
- Against the spread / ATS
- Wins, losses, and pushes measured against the closing point spread.Why it matters: It tracks performance relative to market expectations rather than straight-up results.
- ATS %
- Against-the-spread wins divided by wins plus losses; pushes are excluded.Why it matters: It measures the model's historical spread record while keeping sample size visible.
- SU %
- Share of model-selected teams that won the game straight up.Why it matters: It gives outcome context separate from covering the spread.
- Model MAE
- Mean absolute error between the model's predicted margin and the final margin.Why it matters: Lower values mean the model's score margins were closer to actual results.
- Market MAE
- Mean absolute error between the market-implied margin and the final margin.Why it matters: It is the comparison baseline for judging the model's margin accuracy.
These numbers are the current proof gate for the prediction side track. They do not gate the matchup board.
| Scope | Picks | ATS | ATS % | SU % | Model MAE | Market MAE |
|---|---|---|---|---|---|---|
| Full slate | 815 | 403-384-28 | 51.2% | 63.6% | 10.36 | 9.42 |
| Top pick/week | 54 | 25-27-2 | 48.1% | 53.7% | 12.81 | 9.69 |
| Top 10/week | 540 | 263-261-16 | 50.2% | 59.9% | 10.83 | 9.41 |
Backtest confidence tiers
- Abs edge / Tier
- The absolute point difference between the model margin and market margin, grouped into bands when shown as a tier.Why it matters: It measures disagreement magnitude without implying that either side is correct.
- Picks
- The number of completed model decisions in the sample.Why it matters: It reveals how much evidence sits behind the displayed rates.
- Against the spread / ATS
- Wins, losses, and pushes measured against the closing point spread.Why it matters: It tracks performance relative to market expectations rather than straight-up results.
- ATS %
- Against-the-spread wins divided by wins plus losses; pushes are excluded.Why it matters: It measures the model's historical spread record while keeping sample size visible.
- Avg edge
- Average absolute model-versus-market margin difference within the row's tier.Why it matters: It describes the typical disagreement represented by that calibration bucket.
| Abs edge | Picks | ATS | ATS % | Avg edge |
|---|---|---|---|---|
| 0-2 | 270 | 142-116-12 | 55.0% | 0.9 |
| 2-4 | 232 | 113-113-6 | 50.0% | 2.9 |
| 4-6 | 135 | 62-69-4 | 47.3% | 4.9 |
| 6+ | 178 | 86-86-6 | 50.0% | 8.4 |