For the complete documentation index, see llms.txt. Markdown variants of the ranked boards (dynasty and fantasy rankings for NFL, NBA, MLB and CFB) are available by appending .md to those URLs or by sending Accept: text/markdown; other pages answer 404.
DYNATYZE
This page's interactive half has not started on this device. Nothing was read, and nothing was invented in its place. Reload the page.
DYNATYZE
Loading
This page's interactive half has not started on this device. Nothing was read, and nothing was invented in its place. Reload the page.
Drawn from the weekly play-by-play file, which paints on the Monday 4:00 AM ET sync after the game.
The model against the market, the momentum underneath, every drive — scrub to replay it
How the game was won
The model and the market were 2.4 win-probability points apart on the average play, and 11 apart at the widest — with the model the more convinced about Patriots.
MarketModelmodel higher on Billshigher on Patriots148 of 157 plays carry both curves
(5:48) (Shotgun) 10-M.Jones pass deep left intended for 15-N.Agholor INTERCEPTED by 23-M.Hyde at BUF -2. Touchback.
Swing +9.2 for BillsExpected points +4.46 for BillsThe model had Patriots 7 points higher than the market did.play 23
The solid line is the betting market's home win probability; the dashed line is the play-by-play model's. The shaded band between them is the disagreement — emerald where the model liked the home team more, red where it liked the away team more. Drag across the curve or use arrow keys to replay the game; gold diamonds mark the biggest swings.
Highest-leverage plays
1st · 5:48NE +9% win prob(5:48) (Shotgun) 10-M.Jones pass deep left intended for 15-N.Agholor INTERCEPTED by 23-M.Hyde at BUF -2. Touchback.
1st · 6:28BUF +7% win prob(6:28) (Shotgun) 10-M.Jones scrambles right tackle ran ob at BUF 34 for 16 yards (91-E.Oliver).
1st · 12:52BUF +6% win prob(12:52) (Shotgun) 17-J.Allen scrambles left end ran ob at NE 31 for 26 yards (32-D.McCourty).
1st · 7:24BUF +5% win prob(7:24) (Shotgun) 10-M.Jones pass deep right to 85-H.Henry pushed ob at 50 for 30 yards (30-D.Jackson).
1st · 14:53BUF +4% win prob(14:53) (Shotgun) 17-J.Allen pass short middle to 14-S.Diggs to BUF 39 for 9 yards (8-J.Bentley).
1st · 5:40BUF +3% win prob(5:40) (Shotgun) 17-J.Allen pass deep left to 88-D.Knox to BUF 44 for 24 yards (8-J.Bentley).
1st · 11:18
SettledThe modelled figures are the league play-by-play model's, not ours.
How the points landed
The margin ledger
Dawson Knox's 8-yard touchdown catch with 9:58 left in the 1st put Buffalo Bills ahead for good. The lead never changed over the 9 scores that followed.
4th1:49Kendrick Bourne·4-yd TD catch4-yard touchdown catchfrom Mac Jonesfrom Mac Jones+717–47
10 scores · bar length is the lead after each one
Down-to-down truth
Situational splits
Patriotscategory by categoryBills
Patriots0–9Bills
305Total yards484
5.1Yards per play9.5
38%Success rate69%
50%Third-down rate100%
2Red-zone touchdowns6
5Explosive plays10
18First downs27
3Sacks taken0
2Turnovers0
The read
Bills ahead 9–0 on categories, including the biggest gap of the duel — sacks taken.
Expected points added, per play
The centre line is an average play. A bar to the right of it is a side that gained more than the situation was worth; a bar to the left is a side that lost ground. All six bars share one scale, ±0.94 points.
Every snapexpected points added per play, all downs
+0.07Patriots: +0.07 expected points added per play
+0.77Bills: +0.77 expected points added per play
Dropping backexpected points added per dropback
+0.12Patriots: +0.12 expected points added per play
+0.94Bills: +0.94 expected points added per play
Running itexpected points added per rush
−0.05Patriots: −0.05 expected points added per play
+0.56Bills: +0.56 expected points added per play
Neither side “wins” volume or play-calling mix, so neither is drawn as a duel: Patriots ran 60 snaps at a 70% dropback rate against Bills’s 51 at 55%. Red-zone trips: Patriots 2, Bills 7.
NoteBills won the down-to-down efficiency fight: 0.77 expected points added per play vs 0.07.
Derived from the play-by-play, no model
The Read
7 of 7 readings cleared their snap floor and were tested against four comparisons: this week's whole slate, the team's own season, the week before, and the team's own ordinary spread. Every bar below is set in that team's own week-to-week swing, never in league points — one team's swing runs 1.6 to 3.1 times the entire league's spread.
Nothing cleared the bars. No reading led the week's slate, set a season mark, reversed the week before, or sat two of the team's own standard deviations off their number — which is what most Sundays look like, and is the finding rather than a gap.
The season-grain rooms, pointed at this one game
Was this normal for them?
7 of 8 readings cleared their snap floor. Every deviation below is measured in that team's own game-to-game swing, not in league points — because one game's swing runs from 1.6 to 3.1 times the whole league's spread.
This is not a regular-season game, so every season number below is that team's 2021 regular season — the only stretch long enough to be a habit.
their week-to-week swingall 32 teams
upper band = their week-to-week swing · lower rule = all 32 teams · scale -49 pts to 37 pts
leaned — stayed blocked · −10.3 vs their 13.9%, and they swing ±7.4 week to week (−1.40× that swing) · 28 dropbacks
Bills11.9%
inside their normal wobble · −3.6 vs their 15.5%, and they swing ±7.1 week to week (−0.50× that swing) · 42 dropbacks
Play-action rate
Charted from 2022. The columns exist for 2021 and are entirely empty, which is not a season of data.
Blitz rate sent
Charted from 2022. The columns exist for 2021 and are entirely empty, which is not a season of data.
NoteNE's fourth-and-short go rate was the night's one real departure: 80.0%, 50.0 points above their season mean of 30.0% and 1.6 times their own 31.1-point weekly swing.
The Game Day Roster
Who was out, who dressed and never took a snap, and who was on the field how much. Game-day list and snap counts from the weekly roster and snap feeds. Practice-squad and injured-reserve players are not shown — they were never on the game-day list.
7 players played under 8 snaps, so their shares read as a dash rather than a number. 4 lines have no expected-points read and sit off the plane.
Expected points are modelled in full PPR, so the “vs expected” column stays in that unit whichever format the pills are set to. Switching the pills changes the points column, not the plane.
The field map
Every snap, where it happened
The whole game as one hundred yards. The Bills attack right, the Patriots attack left, so a change of possession is a reversal you can see. Reading down is the order the snaps were played — not the clock, so a two-minute drill gets the same room as any other fourteen snaps. This game is final.
Patriots←Bills→
Q1Q2Q3Q4
M.Jones pass−10% win probabilitymoment 2 of 4
(5:48) (Shotgun) 10-M.Jones pass deep left intended for 15-N.Agholor INTERCEPTED by 23-M.Hyde at BUF -2. Touchback.
Q1 5:48 · Patriots · 1st & 10 at BUF 34 · gained nothing · -4.46 EP · −10.0% win probability
146 snaps placed · 11 without a recorded yard line, left off the map · 4 marked moments — the largest win-probability swings, one per drive.
The Film Room · post-game
Post-game film: 157 plays across 17 drives, personnel groupings on 157 snaps. Everything here paints once the weekly sync ingests the game — there is no live play-level feed in this product, so nothing on this page updates while a game is being played.
Possession by possession
The field rail
10 of 17 drives ended in points. The longest march was BUF moving 89 yards from the BUF 11 for a touchdown. Each bar is the ground a drive actually covered; BUF attacks right, NE attacks left. Hatching marks a trip inside the twenty.
Every snap
Play stream
157 of 157 plays
Team
Quarter
Down
Play type
Expected points added
Event
Charted
Plays shown
157
EPA per play
+0.25
Success rate
46%
Yards
787
Who was on the field
Personnel groupings
Offensive grouping on every pass and run. A group needs five snaps in this game before it gets a number — under that the sample is a coin flip and the cell reads a dash instead. The names of the eleven on the field are only recorded from 2023 onward, so this game gives the grouping without the roster.
Patriots1 RB, 1 TE, 3 WR36+0.2283%
Bills1 RB, 1 TE, 3 WR34+0.9159%
Patriots2 RB, 1 TE, 1 WR9-0.1111%
Patriots1 RB, 2 TE, 2 WR6+0.6833%
Bills2 RB, 1 TE, 2 WR6+0.4850%
Patriots2 RB, 1 TE, 2 WR5-1.2380%
Bills2 RB, 1 TE, 1 WR5+0.9120%
Bills1 RB, 0 TE, 4 WR3——
Patriots2 RB, 2 TE, 1 WR2——
Patriots1 RB, 1 TE, 2 WR2——
Bills2 RB, 1 TE, 0 WR2——
Bills1 RB, 0 TE, 3 WR1——
snapsepapass rate
157 plays · ranked by the win probability each one moved
Every play
Ball
When
Play
Only
Who
157 plays
Bar length is win probability the play moved
1st quarter37 plays
Patriots0Bills6
Patriots0Bills7
Touchdown11 plays, from the Bills 30 to the Patriots 8
Intercepted11 plays, from the Patriots 13 to the Bills 34
Patriots0Bills13
Patriots0Bills14
Touchdown11 plays, from the Bills 20 to the Patriots 11
The period ended3 plays
2nd quarter47 plays
Punt3 plays, from the Patriots 30 to the Patriots 34
Patriots0Bills20
Touchdown13 plays, from the Bills 19 to the Patriots 3
Punt9 plays, from the Patriots 21 to midfield
Patriots0Bills26
Patriots0Bills27
Touchdown6 plays, from the Bills 11 to the Patriots 16
Patriots3Bills27
Field goal14 plays, from the Patriots 25 to the Bills 26
The period ended2 plays
3rd quarter35 plays
Intercepted5 plays, from the Patriots 24 to the Bills 49
Patriots3Bills33
Touchdown8 plays, from the Bills 42 to the Patriots 24
Patriots9Bills33
Patriots10Bills33
Touchdown14 plays, from the Patriots 25 to the Bills 3
The period ended8 plays, from the Bills 23 to the Patriots 36
4th quarter38 plays
Patriots10Bills39
Patriots10Bills40
Touchdown4 plays, from the Patriots 29 to the Patriots 19
Punt7 plays, from the Patriots 24 to the Patriots 42
Patriots10Bills46
Patriots10Bills47
Touchdown4 plays, from the Patriots 39 to the Patriots 1
Patriots16Bills47
Patriots17Bills47
Touchdown18 plays, from the Patriots 27 to the Bills 4
The period ended5 plays
Win probability is the league play-by-play model, 98.5% of plays · a dash means none was published