GUIDES · BETSLIP DOCTOR
Which play context and model version does an NFL expected-stat explanation use?
An expected statistic should be read with its model vintage and its event-time references. NFL's historical explanations place some Completion Probability factors at the throw and others at the catch, define Time to Throw across the snap-to-throw interval, describe Expected Rushing Yards from the handoff, and describe Expected Yards After Catch at reception. Those distinct clocks matter: the historical Completion Probability explanation does not describe every input as known at release. None of these older articles establishes a current feature set, weight set, or feed.
For NFL fans reading model-based passing, rushing, and post-catch analysis. · Updated
The historical Completion Probability article spans the play
The 2018-season NFL explainer describes Completion Probability as a model estimate based on more than ten in-play factors. It ties quarterback speed and the distance to the nearest pass rusher to the time of the throw. It defines Time to Throw as the duration from snap to throw.
Other references in that same explainer reach later in the play. Its air-distance definition connects the quarterback's location at the throw with the receiver's location at the catch. Its worked examples also refer to receiver separation and sideline position at the catch or arrival. The article therefore does not support a shorthand claim that every listed feature was available when the quarterback released the ball.
The source does not document the complete feature timestamp, capture-latency, or availability rules for each model input. Preserve the event times it states, and mark the rest unknown rather than reconstructing them.
- At throw: the article names quarterback speed and quarterback-to-nearest-rusher separation at the throw.
- Across the play: Time to Throw spans snap to throw.
- At arrival or catch: air distance uses the receiver's catch location, and the examples describe receiver and sideline context at the catch.
Rushing and post-catch explanations use different anchors
The NFL Expected Rushing Yards explainer builds on the 2020 Big Data Bowl and describes player location, speed, and acceleration at the moment of handoff. It describes a distribution of possible rushing outcomes and names related measures such as RYOE. That source is a historical account of the model developed from the competition, not proof of today's feature set or calculation.
The NFL xYAC article says the expected post-catch measure is framed at the moment of reception. It describes context including receiver movement, nearby defenders, and blockers, and explains that expected yards before a defender closes in is an input to its described model. A later actual run after the catch is a different event from the expectation anchored at reception.
A historical composite is not a current specification
A separate NFL Passing Score article describes a seven-model composite and a predicted value of a pass before the throw. Its model list includes earlier Completion Probability and xYAC models alongside other expected-value and outcome models. Read that as the lineage described in that article, not as evidence that the same components, model versions, or weights are used now.
The Completion Probability and xYAC articles identify their descriptions with the 2018 season. The rushing article connects its method to the 2020 Big Data Bowl. The Passing Score page does not show a publication date in its article header. Current feature timing, weights, calibration, outputs, and availability are not established by those historical descriptions.
Three fictional plays, three different clocks
Fictional handoff: a running back receives the ball while the defense is spread across the line. The historical Expected Rushing Yards description anchors player location and movement features at the handoff; the back's later path and final gain happen afterward, and the example assigns no expected value. Fictional forward pass: the quarterback releases while moving and a rusher is nearby. Those are throw-time references in the 2018 Completion Probability article, but the article's air-distance measure also uses the receiver's location at the catch, and its examples include catch-time separation context. The estimated probability is unknown here. Fictional completed pass: the receiver catches the ball with defenders and blockers in different paths, then runs several steps. The xYAC article anchors its expectation at reception; the later run is not information that can automatically be moved back to that moment. A historical Passing Score article describes a separate pre-throw pass-value component and a seven-model composite, but this example does not assign any score or current model lineage.
Related questions
Does the 2018 Completion Probability source say every input is known at release?
No. It names some throw-time factors, defines Time to Throw from snap to throw, and describes air distance using the receiver's location at the catch. The source does not state that all inputs are available at release.
Do the older articles establish the current NFL model versions?
No. They document historical descriptions and selected model lineage. They do not establish a current feature set, weights, calibration, outputs, or BSD data availability.
Sources and further reading
- NFL, Next Gen Stats: Introduction to Completion Probability, 2018-season methodology article; reviewed October 15, 2026
- NFL, Next Gen Stats: Intro to Expected Rushing Yards, article built on the 2020 Big Data Bowl description; publication date not shown; reviewed October 15, 2026
- NFL, Next Gen Stats: Intro to Expected Yards After Catch, 2018-season methodology article; reviewed October 15, 2026
- NFL, Next Gen Stats: Intro to Passing Score metric, historical methodology article; publication date not shown; reviewed October 15, 2026