Kansas City vs Buffalo in the playoffs has become NFL theater at its finest. This year’s installment built to a crescendo with the Bills facing 4th and 5, two minutes remaining, and their season on the line at Arrowhead.
The ensuing play sparked intense debate across football media as many staked out firm positions on which hypothetical alternative outcomes were best. We here at SumerSports saw an opportunity to leverage for a frame-by-frame examination of the counterfactuals. By analyzing each moment as it unfolds, without the benefit of knowing the outcome, our AI offers a unique perspective free from the hindsight bias that often colors human analysis.
Let’s walk through the play, starting with the defensive pre-snap picture, examining how the coverage evolved post-snap, and finishing with an analysis of the target options Josh Allen had given the hand he was dealt.
A Spagnuolo Special: Pre-snap Chaos
Steve Spagnuolo (“Spags”), enroute to potentially helping orchestrate the NFL’s first three-peat, remains one of the Chiefs’ most underrated assets. On this critical 4th and 5, Spags deploys dime personnel but shows aggressive intent with six defenders on the line of scrimmage – both Reid and Bolton mugged up. This pre-snap picture creates significant uncertainty for the protection plan: our AI calculates a 54% probability of a stunt and 41% chance of a team blitz. The Bills respond by deciding to shift their protection left. While it’s hard to fault sliding protection toward Chris Jones’ side, we’ve identified a potential vulnerability – Reid showed a 29% pre-snap rush probability, yet the protection scheme appeared to have no answer if he came.
At the snap, Spagnuolo executes a masterful pressure design. Both Danna and Bolton pop drop into coverage while Reid rushes – but with another surprising wrinkle as McDuffie corner blitzes from his condensed press alignment. This corner blitz was particularly deceptive, as our AI calculates that from this pre-snap alignment, a corner blitz occurs in less than 1% of similar situations in NFL play. The combination creates a two-man overload to the offense’s right, immediately compromising the protection. Within 1.5 seconds of the snap, our AI shows the two unblocked DBs with around 80% pressure probability each. The first down probability and expected yards for the Bills plummets from 49% and 7.3 pre-snap to just 23% and 1.8 yards by the 1.5 second mark, quantifying the devastating effect of this pressure design in conjunction with the protection shift left. The play had basically collapsed before Allen could even finish his dropback.
Smoke and Mirrors: The Coverage Shell Game
Unfortunately for Josh Allen, Spagnuolo didn’t just attack with pressure design – he had the coverage shell rotating at the snap: Danna and Bolton pop dropped into shallow zones, Chamarri Connor sank from his nickel alignment to cover the deep middle, while Hicks and Cook shifted to the backside to cover for the blitzing McDuffie. Our AI’s team coverage recognition mirrors the complexity, as its predictions start ambiguous pre-snap, morphs to Quarters as the defenders move but ultimately finishes in a Cover-3 prediction.
In a span of mere seconds, Spagnuolo had sprung a dangerous defensive trap – combining an overload pressure with shape-shifting coverage that would stress even the most seasoned quarterback’s processing ability.
Making Tough Decisions: Analyzing Allen’s Options
To break down Josh Allen’s options objectively, we introduce one of our AI’s latest views: Targeted Success Probability – a comprehensive measure that combines both completion probability and expected gain to estimate the likelihood of achieving a positive play by EPA for each target at each moment. We will focus on the critical window where Josh Allen gets the ball (frame 12) and when was his last opportunity to make a play in structure (frame 20).*
* Footnote: We exclude Kincaid and Hollins from this view because it is clear from the film that they are running deeper routes and are untargetable during this window.
Was Shakir the Obvious Choice?
Much of the debate around this play has focused on whether Allen should have hit Shakir’s orbit return early before the pressure got home. Taking a look with our AI, we see that even though a swing pass to Shakir around frame 15 is very likely to be complete, the model only expects a 25% chance of conversion if he is targeted at this point. This may seem counterintuitive initially as it looks like he is in wide open space, but with 15 yards to gain and the defense in eyes-up zone coverage able to rally to the ball, the model is skeptical on Shakir’s ability to convert.
But why should we believe the model? As ESPN Analyst Ben Solak points out (40 second mark) this orbit motion has been targeted previously by the Bills in critical situations and to great success before. To build some trust in our AI, let’s examine the 4th and 3 play Solak cites from last year’s playoffs.
Unlike for the controversial play this year, our AI agrees strongly with Solak on this one by showing that Shakir is clearly the best option with an 88% Targeted Success Probability, and Josh Allen capitalized on it near perfectly. The stark difference in success probability could stem from several key factors: the defense was in man coverage, the yards needed from the catch point were more manageable, and the defensive spacing at that moment lacked horizontal width to effectively constrain Shakir. Predicting yards gained and conversion probability many frames into the future is a notable strength of tracking-data-based models such as our own at SumerSports, and can probably do a better job at evaluating that than us humans.
The Underrated Choice: Shallow Crossers
So, if not Shakir, our AI believes that throwing to either of the shallow crossers of Ty Johnson or Amari Cooper were quality options between frames 12 and 16, each with Targeted Success Probability of around 48%. Ty Johnson in particular stands out as a quality hot option, as McDuffie’s blitz left his crossing route largely uncovered. If completed, Bolton would be the only defender with a chance to prevent the first down, but with his hips oriented to the strong side, he would have struggled to redirect and make a tackle in time.
The film confirms Allen should have had vision of Ty Johnson but given that the Bills were not expecting an overload from the right, it is possible that Ty Johnson being a hot option was not in the concept and would not have seen a pass thrown early at him.
The Hero Moment
Once these shallow crossers are out of the equation, Josh Allen tries to escape to his right by Frame 17 as the unblocked DBs descend upon him. A heartbeat later, he sees George Karlaftis taking away that escape path as well. He has about one second to get his eyes up backpedal and uncork a ball. At Frame 32, when he lets go of it, our AI sees the exact same thing that Allen does: Dalton Kincaid breaking free as the safety in that space crashes towards the line of scrimmage. At that moment, Kincaid was the best option with around a 40% Targeted Success Probability. This value is only marginally less than the earlier options on the crossers, so it would be hard to fault Allen’s process here; he still made a heroic throw in tough circumstances.
Through The AI’s Eyes: What We Learned
The frame-level analysis of this pivotal play reveals the layers of complexity in modern NFL defense. Spagnuolo’s design created a perfect storm: capitalizing on a faulty protection shift, an unexpected corner blitz creating an overload, and coverage rotations to exacerbate the dire offensive situation. Even through all this, our AI’s analysis suggests that while there may have been a theoretically better option in Johnson early, the option Josh Allen finally chose was only slightly worse.
