What it trained on
Loading provenance…
The Lab
The numbers, with their work showing.
This is the clubhouse's model room — and it opens with proof, not promises. The first working tool is an NFL in-game win-probability model trained on real play-by-play and backtested on a held-out season, with its calibration table published below. Move the levers and the number moves with you. A model, not a certainty — the label is part of the tool.
NFL What-If model
Loading…
Reading the model card.
Held-out proof
Loading…
Brier score and calibration load from the backtest output.
NBA foundation
Loading…
Only what the free, license-clean probe actually produced is shown.
Foundations only in this phase: no live in-game feed is wired yet. The toolbar runs the trained model in your browser on states you set; the NBA section shows the pregame foundation and states the in-game gate plainly.
NFL What-If Toolbar
See the calibration ↓Set the game state — quarter, clock, margin, possession, down, distance, field position, timeouts — and the trained logistic model recomputes the home team's win probability instantly. Every control below maps to a feature the model was actually trained on; nothing decorative, nothing faked.
Model output — home team
—
Loading the trained model…
The state you set will be restated here in plain words.
Model version, training window, and test window load from the model card.
A model, not a certainty. A 70% state still loses three times in ten over the long run — the calibration table below is where this model proves how often its numbers meant what they said on games it never trained on.
Calibration — the held-out proof
Rendered from the backtest output — no hand-typed numbersThe model was trained on earlier seasons and graded on a later season it never saw in fitting. A calibrated model's 70% calls should land near 70% in real life. The bins below are that test, row for row, next to the dataset's own win-probability column as the benchmark.
Loading the backtest metrics…
Reliability table — our model vs the nflfastR benchmark
| Predicted bin | Plays | Model avg | Actual home wins | Model | Benchmark avg | Benchmark |
|---|---|---|---|---|---|---|
| Loading bins… | ||||||
In each row's bar, the amber fill is the model's actual home-win rate and the white tick is what the model predicted on average — close together is the whole point. The benchmark column is nflfastR's own wp on the same held-out plays, shown so our number never grades itself alone.
Model card — NFL win probability
How it works, and what it is not
Loading method…
Build Your Own Line — the starter strip
Pregame · your numbers, your assumptionsPick two teams and the strip starts from the Lab's own power ratings — computed from game margins in the training window by the documented method on the ratings card. Set the home-field value and a rest edge, and your line builds in the open. This is modeling education: the output is your number from your settings, never a betting recommendation.
Your line
—
Loading the ratings table…
The arithmetic will show here, lever by lever.
No market column in this starter — on purpose. A market reference belongs beside a specific game whose line is present in sourced data. This strip works from ratings, not a named game, so no market number is shown rather than an invented one being placed next to yours.
Power ratings — the full table
Loading the ratings method…
| # | Team | Rating (pts) |
|---|---|---|
| Loading ratings… | ||
NBA — the foundation, honestly stated
Loading the NBA foundation…
Luka works in here
Set the state, he'll walk it
Luka can restate any toolbar state in plain words, explain which lever moves the number most, and quote the model's calibration — and when chat can't run the model itself, he says so and points at the toolbar readout instead of inventing a percentage.
Co-build in words
Give Luka two teams and your levers and he restates them, shows the arithmetic line by line, and narrates which lever is doing the work — his closing line never moves: your numbers, your assumptions, not betting advice.
What the Lab changes — and what it doesn't
The Lab's first NFL model prices in-game states; it is not a pregame pick engine, and Luka's pick'em picks still run on the labeled knowledge rule. Where a Lab model genuinely covers a question, its number carries the model card with it. Where none does, the take runs on tiers — never an invented percentage.