The plan against the game · plan Sun Oct 18, 9:41 AM · report Tue Oct 20, 11:41 AM Cousins carves a Buffalo secondary that bends. Cook and Allen punch it in on the ground. script partly held 7 hit4 missed projection C · 67 open
Buffalo should lead and finish drives with its legs; Las Vegas answers through the air, because Buffalo's secondary gives up yards and touchdowns at the goal line.
Buffalo should control this between the 20s and convert field position into short rushing touchdowns against a Raiders defense that surrenders a score on four of five red-zone trips. Las Vegas will move it through the air underneath — Bowers, Mayer and Jeanty on checkdowns — because Buffalo's coverage has conceded four-plus explosive passes every week and the most passing yardage in the league. The Raiders' run game is non-functional, so any deficit turns into a 38-plus attempt night for Cousins. Expect a scoreboard that stays live into the fourth without the game ever feeling close on a per-play basis.
Buffalo won 33-30 but failed to cover -5.5, and the 63 points blew through a 53.5 total: the Bills moved at 0.185 EPA/play yet turned five red-zone trips into three touchdowns and two Bass field goals, while Cousins threw for 313 on 41 attempts against a Buffalo secondary that gave up four 20-plus passes and 33 DPI yards.
The plan read the shape of this game well and the lean on one side exactly: Buffalo led most of the way, Las Vegas was chased into a 41-attempt night with its run game abandoned, Cousins threw two scores, and Jeanty's carries dried up while his receiving work held. Where it went wrong was Buffalo's own identity — it called a run lean and Buffalo passed at 0.547 against a 0.502 season rate, which took the Allen rushing over, the Allen pass-TD under and the Allen pass-yard under with it in one chain. The most interesting miss is mechanical, not scripted: the plan assumed Las Vegas pressure turns into Allen scrambles, and instead it turned into two sacks and one 2-yard scramble. The Bowers miss was the bounce (9 targets, 4 catches, -9.7 CPOE team-wide), not the read.
The script: Buffalo led for 2,069 of the game's seconds and Las Vegas trailed on 60 of 66 snaps, exactly the favorite's branch; the Raiders got to 41 attempts and 20 carries, Jeanty to 10 carries for 22, as the plan's underdog mechanism described. But the favorite path specified Allen under 30 attempts with a quiet passing box score and touchdowns on the ground — he threw 33 for 255 with 2 passing TDs and 1 rushing yard, and `checks` has Buffalo's lean as pass (0.547 vs 0.502 season), so the lean call failed. Pace held: 125 combined plays against 121.5 league, called average.
Expected: Pass rate near 0.50, 28-32 Allen attempts, Cook in the 17-21 carry range if the Bills lead, Allen 7-10 carries including scrambles, and the touchdowns arriving on the ground rather than through the air.
Happened: Carries (29) and Cook's workload (22 for 95, 1 TD) landed in range, but pass rate came in at 0.554 not ~0.50, Allen threw 33 not 28-32 with two passing touchdowns, and he had 5 carries for 1 yard, not 7-10 including scrambles. Buffalo's touchdowns were 2 through the air and 1 on the ground — the inverse of the expectation.
Expected: 34-40 Cousins attempts if the Raiders trail, Bowers at nine-plus targets as the alpha, Jeanty's value in the passing game rather than on the ground, and the points arriving as touchdowns rather than kicks given Buffalo's goal-line record.
Happened: Cousins threw 41 times trailing, Bowers took 9 targets as projected, Jeanty's value was receiving (4-36) not rushing (10-22), and LV drew 33 DPI yards from Buffalo's coverage. The one part that broke: the points came as 2 TDs and 3 field goals, not as touchdowns — LV's own red-zone EPA was -0.001.
Buffalo should control this between the 20s and convert field position into short rushing touchdowns against a Raiders defense that surrenders a score on four of five red-zone trips. Las Vegas will move it through the air underneath — Bowers, Mayer and Jeanty on checkdowns — because Buffalo's coverage has conceded four-plus explosive passes every week and the most passing yardage in the league. The Raiders' run game is non-functional, so any deficit turns into a 38-plus attempt night for Cousins. Expect a scoreboard that stays live into the fourth without the game ever feeling close on a per-play basis.
Buffalo scores early, Allen's scrambles and Cook's carries eat the middle quarters, and the Bills finish two or three short-yardage drives on the ground. Allen throws under 30 times and the Bills win 30-23 with a quiet passing box score.
Las Vegas trades scores by finishing in the red zone against the league's softest goal-line defense, Cousins clears 38 attempts feeding Bowers and the backs, and a Buffalo giveaway or a drive-extending flag keeps it one score — a 31-27 type where every Raiders passing number plays up.
The read
Buffalo is the better team by a distance between the 20s (0.224 field EPA per play, 9.05 yards an attempt, 5.37 a carry) and is laid 5.5 in a dome. The way it scores is on the ground: every touchdown in Week 3 was a one- to three-yard run, Allen has a rushing score in all five games, and Cook has four rushing touchdowns in the last four weeks. Las Vegas gives up 5.21 yards a carry (31st) and 140.6 rushing yards a game (28th), and allows a touchdown on 80% of red-zone trips (30th). That is the Bills' half.
Las Vegas is not a team that gets shut out of this. Buffalo's defense has allowed 295.8 passing yards a game (31st), four or five explosive passes every single week, and a touchdown on 83.3% of red-zone trips against it — the worst finishing-defense number on the board. Cousins has thrown three touchdowns in three of five starts; Bowers has gone 116, 86 and 110 receiving yards in three games since returning and practiced in full Thursday (FantasyPros, Thu).
Why
The mechanism that decides volume is possession. Las Vegas runs 64.6 plays a game and cannot run the ball at all (2.99 a carry, 1.40 before contact against a 2.54 league mark), so once it trails, Cousins is throwing — 52 attempts the last time that happened, against Kansas City. Buffalo's run lean, by contrast, is conditional: our Week 4 review of the New England loss found the Bills threw on 60.3% of snaps because they never led. Favored here, the lean returns, which caps Allen's attempts and hands Cook and Allen the short-yardage scores.
The Raiders' counter-lever is pressure — second in the league at 22.8% — which pushes Allen out of the pocket. That raises his scramble yards more than it lowers them.
What would change it
DJ Moore and Josh Palmer are both questionable (Rotowire, Fri); Moore practiced in a non-contact jersey. If Buffalo falls behind, Allen's passing lines flip and Cook's carries evaporate — the exact break that cost us in Week 4. And if Buffalo blows it open, Cousins may not finish, as in Week 5.
The most efficient offense in this sample between the 20s: 0.224 field EPA per play, 9.05 yards per attempt, 5.37 a carry with 3.49 yards before contact against a 2.54 league mark. Five explosive passes or more in four of five weeks.
Chunk passing to set up short-yardage runs. Every Week 3 touchdown was a 1-, 1- or 3-yard run; Allen has a rushing score in all five games and Cook four in four weeks.
Pressure and their own flags. Four sacks and 32% pressure in Week 3 produced five turnovers; 13 flags and six pre-snap in Week 5; holding wiped a 43-yard touchdown in Week 3. And a script where they trail — Week 4 at 0.603 pass rate, Cook capped at 12 carries.
Las Vegas generates pressure at the second-highest rate in the league, which is the one thing that has derailed this offense. But the Raiders surrender 5.21 yards a carry (31st) and 140.6 rushing yards a game (28th) and allow a touchdown on 80% of red-zone trips (30th) — their rushing EPA rank rests on stuffs, not on yardage.
DJ Moore (shoulder) and Josh Palmer (shoulder) both questionable and both limited all week; Moore was in a non-contact jersey Friday (Rotowire, Fri). Coleman, Shakir and Bell handled nearly all receiver routes in Week 5 and would again. Ty Johnson is clear of the hamstring tag. On the Raiders' side, Hemingway is out (13% snaps) with Booker and Fatukasi absorbing interior snaps.
Pass rate near 0.50, 28-32 Allen attempts, Cook in the 17-21 carry range if the Bills lead, Allen 7-10 carries including scrambles, and the touchdowns arriving on the ground rather than through the air.
A short-area passing offense with no run game: 0.101 field EPA per play, 7.32 yards an attempt, and 2.99 yards a carry with 1.40 before contact against a 2.54 league baseline. Cousins has thrown multiple touchdowns in four of five.
Underneath volume to Bowers, Mayer and the back, plus one or two shots to Tucker. Short fields off takeaways have supplied the rest — the Week 3 win at New Orleans was built on three fourth-quarter giveaways.
Its own red zone (0.562 TD rate, 1-of-3 against Kansas City with five field goals) and any front that keeps it one-dimensional. Protection cracks: 9 QB hits at the Chargers, 3 sacks each at New Orleans and New England.
Buffalo has allowed 295.8 passing yards a game (31st), four or five explosive passes every week, and a touchdown on 83.3% of opponent red-zone trips — the worst finishing defense in the sample. Its pass rush is concentrated in Rousseau; outside him the heat is intermittent (14.3% pressure in Week 4). Buffalo does hold yards per carry down (3.92, 12th), which does nothing for a team that cannot run anyway.
Bowers full Thursday and Friday off the knee (FantasyPros, Thu); Jeanty full off the foot but visibly limited in Week 5 (10 carries, 53% snaps). Nailor cleared concussion protocol; Thornton not activated. Buffalo's interior is thinned: Ed Oliver questionable after two DNPs (Rotowire, Fri), and slot corner Dee Alford questionable, which would put Hamlin inside against Bowers and the backs.
34-40 Cousins attempts if the Raiders trail, Bowers at nine-plus targets as the alpha, Jeanty's value in the passing game rather than on the ground, and the points arriving as touchdowns rather than kicks given Buffalo's goal-line record.
Each call is one player, one market, one side of the line Kalshi quoted when the plan was written; graded off the box score, with the report's read of whether the reason given showed up. The rungs are the same call down the Kalshi ladder, Edge halfway to its last rung and Edge+ the last, at Kalshi's last pregame price.
In the game: Cook scored from 3 yards at Q4 4:49 on a goal-line carry after his own 26-yard catch set it up — precisely the mechanism (Buffalo's short-yardage scoring handed off into a soft LV goal line).
Las Vegas concedes a touchdown on four of every five red-zone trips, and Buffalo's scoring inside the 10 is handed off. Cook has found the end zone in four straight.
LV allows a 0.800 red-zone TD rate (30th); Cook 1, 1, 1, 2 rush TDs weeks 2-5 on 17-25 carries.
vs #15 EPA allowed; engine over-probability well clear of the market's. Anytime-TD calls are 8-of-12 this season.
In the game: Johnson played 17.6% of snaps, saw 2 targets and caught both for 24 yards including the 14-yard 3rd-and-8 at Q1 12:49. Passing-down role exactly as described.
He is the passing-down back and has caught everything thrown at him since returning, with Buffalo's receiver room banged up. No injury designation after a full Friday.
4-4-50, 2-2-9, 4-4-72 in weeks 3-5 on 22-32% snaps; cleared the hamstring tag Friday (Rotowire, Fri).
vs #26 pass EPA allowed; engine above the line, market barely above — the role, not the yardage, is the call. Thin sample, three games.
In the game: 6 targets, 4 catches, 36 yards on 19 air yards with 34 YAC — checkdown volume from a trailing script, the stated mechanism.
The receiving work is the live half of his usage, and the more Las Vegas trails the more checkdowns he gets. Buffalo's linebackers have been beaten underneath all season.
6, 7, 3, 8, 6 targets; 45, 15, 37, 68, 50 rec yds; 68 of those came in the 52-attempt chase game at KC.
vs #26 pass EPA allowed; engine mean sits clearly above the line with the market near even — the gap is the trailing-script attempts.
In the game: Two touchdown passes, both to Tucker (11 yards at Q4 8:13, 44 at Q4 1:39), on a night of 41 attempts. The plan's goal-line premise was partly fortunate — the 44-yarder was not a red-zone throw — but the volume and the soft Buffalo coverage were real.
Buffalo is the league's softest defense once an opponent reaches the goal line, giving up touchdowns on nearly every red-zone trip. Cousins has thrown multiple scores in four of five starts.
BUF allows an 0.833 red-zone TD rate per trip and 295.8 pass yds/gm; Cousins 3, 3, 3, 2, 1 pass TDs weeks 1-5.
vs #26 pass EPA allowed; engine 1.91, market at 0.595 over — this rests on the red-zone half of the split, not on yardage. Pass-TD calls are 8-of-13 this season.
In the game: Allen finished with 1 rush yard on 5 carries and a single 2-yard scramble. The pressure arrived (6 hits, 2 sacks) and converted to sacks, not scrambles; the short-yardage keepers went to Cook instead.
Las Vegas pressures at the second-highest rate in the league, and pressure on Allen turns into scrambles, not just sacks. Short-yardage keepers add the rest.
121 rush yds in Week 5, 69 in Week 2 with 5 scrambles for 49; LV allows 5.21 yds/carry (31st) and 140.6 rush yds/gm (28th).
vs #8 rush EPA allowed — that rank is built on stuffing designed runs, not on scramble yardage, and the same front's yards-per-carry and yards-per-game ranks are bottom-five. Engine above the line, market slightly above it too.
In the game: Role held — 9 targets, 98.5% snaps, 33.5% air-yard share — but he caught 4 and the line missed by 9.5 yards. Cousins was 23-of-41 with -9.7 CPOE; the volume thesis was right, the completions weren't there.
He is the whole target tree on a team that has to throw to score, and a trailing script adds attempts. Our model's share still lags the 11-plus targets a game he has taken since returning.
34 targets in 3 games back — 116, 86 and 110 rec yds; 70-82% snaps; full practice Thursday and Friday (FantasyPros, Thu).
vs #26 pass EPA allowed and 31st in pass yds allowed; engine mean sits well below the line, the market near a coin flip — the engine's share is the stale part.
In the game: Hit at 38 yards, but not for the stated reason: Kincaid led Buffalo in targets with 8, ahead of Coleman and Shakir. He was capped by depth (15 air yards on 8 looks), not by role in the order.
He has fallen to fourth in the target order behind Coleman, Shakir and the backs, and a run-leaning script keeps Buffalo's attempts modest. The line is priced off September, not the current role.
2, 2, 3 targets in weeks 3-5 for 38, 7 and 7 yards; snap share 59-70% without the volume.
vs #26 pass EPA allowed — a soft matchup he is not being fed in; the model's mean is far below the line while the market sits near even. The same read hit in Week 4.
In the game: Allen threw two touchdowns (Shakir 9 yards Q2 7:17, Coleman 12 yards Q3 6:30) and Buffalo's inside-the-10 scoring went through the air twice to once on the ground. The premise that Buffalo finishes with legs did not appear.
Buffalo's red-zone scoring runs through legs, not throws — Allen's and Cook's. Against a defense that will be chased into a lead script, the short-yardage carry is the call inside the 10.
0, 1, 1 pass TDs in weeks 3-5 while Allen ran for 2, 1 and 1 scores; all three Week 3 TDs were 1-3 yard runs.
vs #26 pass EPA allowed; the model's mean is under the line while the market prices the over — the market is buying the matchup, not the way these points arrive.
In the game: 255 yards on 33 attempts — Buffalo led and still threw above its season rate (0.547 vs 0.502). The conditional the call rested on (leading implies low attempts) was satisfied and the attempts came anyway.
Favored in a dome with the ground game as the scoring plan, Buffalo's attempts stay in the high 20s, and two of its top wideouts are banged up. Entirely conditional on the Bills leading.
26 and 19 attempts in the two games Buffalo ran first; 0.502 season pass rate; Moore and Palmer both questionable (Rotowire, Fri).
vs #26 pass EPA allowed — a matchup he could beat if forced to throw; model mean under the line, market also under. Same read broke in Week 4 when Buffalo never led, and that branch is still live.
In the game: Zero interceptions on 33 attempts, and the volume stayed moderate at 36 dropbacks. Call and reasoning both fine, though a 0-INT game is substantially bounce.
This is a volume call, not a safety one: favorites in control throw fewer times and bleed clock, and our model over-counts attempts for the side that leads.
Allen's picks came in the two games with the most dropbacks (33 and 36 attempts); Buffalo projects near 30 here with a 0.502 pass rate.
vs #26 pass EPA allowed; LV generates takeaways at an above-league rate, which is the risk — the model's mean is above the line while the market already leans under.
In the game: 10 carries for 22 yards with 2 stuffs and -3.79 rushing EPA, exactly the mechanism: carries shrink as Las Vegas falls behind, behind the worst before-contact blocking in the sample.
His carries shrink every week as Las Vegas falls behind, and the blocking in front of him is the worst in the sample. Buffalo gives up EPA on the ground but holds yards per carry down.
23, 21, 19, 15, 10 carries; 3.3 yds per carry on the season; LV at 1.40 yds before contact vs a 2.54 league mark; BUF 3.92 ypc allowed (12th).
vs #28 rush EPA allowed — soft by that metric, which is why this is the contrarian side; the model's mean is above the line and the market sits even. The volume trend is the mechanism.
Legs the same script produces together: they move as one, which is what a same-game parlay pays for.
One script produces all four: Buffalo leads, the Raiders' run game is abandoned, Cousins throws 38-plus, and the volume lands on Bowers and the back out of the backfield while Jeanty's carries dry up — with the scores arriving as touchdowns against the league's softest goal-line defense.
If the Bills lead and finish drives with runs against a defense that concedes a touchdown on four of five red-zone trips, the carries and the scrambles go up and the passing touchdowns go down — the same lever moves all three.
Both rest on the same branch — Buffalo in front, under 30 attempts — with the shrunken tree hitting the fourth option hardest.
grade · 67
| Proj | Actual | Diff | ||
|---|---|---|---|---|
| Points | 24.0 | 30 | +6 | |
| Pass yds | 250.5 | 313.0 | +62.5 | |
| Rush yds | 100.4 | 65.0 | -35.4 | |
| Pass att | 34.8 | 41 | +6.2 | |
| Carries | 27.2 | 20 | -7.2 |
| Proj | Actual | Diff | ||
|---|---|---|---|---|
| Points | 29.5 | 33 | +3.5 | |
| Pass yds | 237.2 | 255.0 | +17.8 | |
| Rush yds | 132.8 | 103.0 | -29.8 | |
| Pass att | 30.5 | 33 | +2.5 | |
| Carries | 27.6 | 29 | +1.4 |
The grade is a C (0.666) because the team shapes were close but the market calls went 1-for-3: Buffalo won, the spread lost and the under lost by 9.5. Buffalo's line was nearly exact — 33 vs 29.5 points, 255 vs 237 pass yards, 33 vs 30.5 attempts, 29 vs 27.6 carries — the only real miss being rush yards (103 vs 132.8), and that was Josh Allen's rushing disappearing (1 yard on 5 carries, no scrambles beyond one for 2), not Cook's (95 vs 91.9). Las Vegas is where the projection leaked: 41 attempts vs 34.8, 313 pass yards vs 250.5, 20 carries vs 27.2 (the one 'big' flag), because the Raiders trailed for 60 of 66 snaps and the engine's blended 0.577 pass rate never contemplated a 0.672 actual. The player book held better than the team book (32 of 47 lines, 0.794), with the misses clustered in one place: the Raiders' secondary receivers (Tucker 97, Mayer 95 on nine targets) soaking up the volume the model gave to Jeanty on the ground. Nothing in the grade is a modeling scandal; it's a correct game shape with the scoring rate underestimated on both sides.
Pattern: Buffalo's defense is a field-tilt loser and a goal-line loser at once: opponents move the ball through the air for chunks (83.3% opponent RZ TD rate entering, 0.24 RZ EPA allowed), and Buffalo's own offense gives points back through penalties, turnovers and stalled trips. It still wins because its field EPA (0.224 on offense) is elite.
This week: Las Vegas tried exactly this and mostly executed: 8 deep attempts, 8.6 aDOT, four explosives and 33 free DPI yards, and it still couldn't close because Cousins' CPOE was -9.7 (23-of-41) and the Raiders kicked three field goals from inside the 20.
What an opponent needs: A quarterback who can hit the deep middle at better than a -9.7 CPOE clip, two-plus pass catchers who win vertically, and red-zone finishing above 0.60 TD/trip. Teams that can also force Buffalo into a trailing script — where its EPA was 0.461 here — are the ones that actually cash it.
Pattern: Las Vegas is a one-dimensional passing offense whose score depends on red-zone luck, defended by a front that gets yards run at it and a goal-line unit that concedes. Fall behind by one score and the Raiders get into a throwing-only script that produces yards (313) without enough touchdowns.
This week: Buffalo tried the blueprint and had the efficiency: a 5.37 YPC season rate against LV's 5.21 allowed, 29 carries, three explosive runs, and an offense that ran 0.509 early-down pass rate — but even it only finished 3 of 5 inside the 20, and the pre-snap flags were Buffalo's problem too.
What an opponent needs: A rushing offense above league YPC that will stay with it, plus red-zone finishing — the Raiders give trips away, they don't give points away automatically. A defense that can get a hand on an 8.6-aDOT passing game without blitzing (BUF blitzed 41% and got zero sacks) keeps the Raiders from answering.
When a team's run game is at 2.99 YPC with 1.40 yards before contact, don't project its season carry volume into a game where it's likely to trail — the carries evaporate below even the engine's conservative blend. LV: projected 27.2 carries, actual 20.
Trigger: Underdog RB with YBC/carry under 2.0 and a team projected to trail; fade the carry/rush-yard over, buy the receiving yards.
A defense that gives up explosive passes every week but also gives up DPI is a totals over, not a side fade — the free yardage doesn't show in passing props. BUF allowed 33 DPI yards on two scoring drives here.
Trigger: Defense with DPI yards allowed above ~10/game (league 6.45) plus 4+ explosive passes allowed per week; take the over, not the opposing receiver props.
Mobile-QB rushing props are a pressure-type bet, not a pressure-rate bet. Allen faced 6 QB hits and a 33% blitz rate and finished with 5 carries for 1 yard and a single 2-yard scramble; the pressure produced two sacks instead.
Trigger: A QB rushing-yards over justified by 'opponent pressure creates scrambles' — check the opponent's sack rate vs hurry profile and whether the QB's team expects to lead (Buffalo's leading-state pass rate was 0.545, not run-heavy).
Two teams with poor red-zone finishing can still smash a total if both move the ball between the 20s. Combined field EPA was 0.205/0.161 and 7 field goals; the 63 points came from volume of trips, not conversion.
Trigger: Both offenses above 0.10 field EPA/play with sub-0.60 RZ TD rates and a number at or below 54 — the field goals add up faster than the stalls subtract.
Target-tree misses on a trailing team land on the second and third options, not the alpha. Bowers got 9 targets but only 70 yards; Tucker (97) and Mayer (95) ate the surplus volume.
Trigger: A heavy-underdog passing script with a priced-up TE1 — take the mid-tier receiving overs rather than the alpha's inflated number.
The short version
Buffalo 33, Las Vegas 30, and both market calls that mattered lost: the Bills won outright but gave back the 5.5, and 63 points cleared a 53.5 total by nearly ten. The projection had the shape of this game almost right — Buffalo's entire team line is within a few points and a few attempts — and still graded a C, because the Raiders threw 41 times instead of 34.8, for 313 yards instead of 250.5, and the scoreboard ran hot on both ends. The thing that decided it was a single possession's worth of EPA: Cousins' third-quarter interception, Buffalo's immediate touchdown off it, and the three-point cushion that survived a 44-yard Tucker touchdown with 1:39 left.
How it unfolded
The first half was pure trade. Buffalo opened with a 30-yard, eight-play field-goal drive built on Allen's 14-yard third-and-8 throw to Ty Johnson (Q1 12:49, +2.77 EPA); Las Vegas answered with a 64-yard march keyed by a 26-yard Tucker catch on third-and-7 (Q1 8:50) and Jeanty's one-yard plunge. Buffalo's 90-yard touchdown drive in the second quarter — Coleman 21, Cook 23, Shakir nine for the score at 7:17 — was the game's longest sustained stretch of offense, and Las Vegas answered it with two field goals, the second built on a 14-yard defensive holding flag on Igbinosun at :19 that tied it 13-13 at the break.
The third quarter is where the margin was made and then partly given back. Buffalo got a 47-yard Bass field goal only because its own holding penalty at 9:57 erased an 18-yard Kincaid touchdown. Then Cousins threw into traffic for Mayer at 9:00 and Dorian Williams picked it off at the LV 32 — a -5.03 EPA play — and Allen hit Coleman from 12 on third-and-8 to go up 23-13. Las Vegas answered with a 16-play, 75-yard, 7:50 possession that produced three points, which is the Raiders' season in one drive.
The fourth was a sprint. A 91-yard LV drive from its own 9, capped by Tucker's 11-yard touchdown on third-and-8 (Q4 8:13, the game's biggest WPA swing at +0.211), tied it at 23. Buffalo needed 3:24 and seven plays to retake it, the decisive snap being Allen's third-and-3 checkdown to Cook for 26 to the LV 3. Bass's 38-yarder at 2:00 made it 33-23, and Tucker's 44-yard score — set up by a 22-yard DPI on Alford at 1:44 — made the final respectable without putting the ball back in Las Vegas's hands.
Why the game diverged from the projection
The divergence is almost entirely on the Raiders' side and almost entirely script. The engine blended LV to a 0.577 pass rate and 27.2 carries; the Raiders trailed on 60 of 66 snaps and finished at 0.672 and 20. That alone moves 60-plus passing yards and -35 rushing yards onto the board. The engine's own inputs — 2.99 YPC, 1.40 yards before contact, the worst in the file — already said this run game would be the first thing jettisoned; the volume projection just didn't act on it.
The second divergence is finishing. Ten red-zone trips produced five touchdowns and five field goals, and the field-versus-red-zone split tells the story on both sides: Buffalo 0.205 EPA/play on 51 field snaps against 0.114 on 14 red-zone snaps, Las Vegas 0.161 against -0.001. That is a yards-without-touchdowns game for both teams, and it is what let a 63-point total coexist with a one-possession margin.
On Buffalo's side the only real miss was Allen's legs: projected 35.6 rush yards, actual one. The Raiders' pressure was real — six QB hits, two sacks, a 41.3% blitz rate per charting — but it produced sacks rather than scrambles, and the short-yardage keeper never came.
The plan against the game
The plan's underdog mechanism was close to perfectly described and the favorite still won by it: Las Vegas chased, abandoned the run, reached 41 attempts, and Jeanty's carries withered to 10 for 22 while his receiving line landed within a yard and a half of projection. What broke was Buffalo's half. The plan read the Bills' 0.502 season pass rate as a lead property and expected a quiet passing box score with touchdowns on the ground; Buffalo led for 2,069 seconds and threw at 0.547 anyway, with an early-down pass rate (0.509) matching the expected rate exactly. That one error chained — Allen's rushing over, the pass-yards under and the pass-TD under all come off the same assumption. The Bowers miss is a different thing entirely: the role was exactly as described at nine targets and a 33.5% air-yard share, and he caught four, because Cousins completed 23 of 41 at -9.7 CPOE. That is the bounce, not the read.
Player by player
Cook was the model's best call and the game's most useful player: 22 carries for 95 with three ten-plus runs, plus 58 receiving yards on three targets at -8 air yards — 66 YAC — and the three-yard winner at 4:49. His receiving yardage (projected 5.7) is systematically underpriced against defenses that sit back. Coleman, with Moore out and Palmer at a 25% snap share, was the downfield alpha at 88 air yards and a 47.8% share. Kincaid led Buffalo in targets with eight but on 15 total air yards — receptions, not yards.
On the other side, Bowers kept his role and lost his yardage; Tucker (7-5-97-2, 108 air yards) and Mayer (9-7-95, 68 YAC) took the surplus. Mayer's nine targets against a 3.9 projection is the kind of secondary-tree spike a trailing script manufactures every time.
The defenses
Buffalo blitzed on 41.3% of LV snaps and generated zero sacks and a 14.6% pressure rate — below league. That is now the second game in two weeks where the pass rush disappeared, and the coverage behind it gave up four explosive passes (26, 28, 29, 44) and 33 DPI yards, two of which kept scoring drives alive. Las Vegas held Buffalo to 3.55 yards a carry on 29 attempts, which is better than its season allowance, but conceded three touchdowns on five trips and never got off the field on the two drives that mattered — Buffalo was 2-for-2 on the third downs at Q4 6:48 and 4:56.
The blueprint
To beat Buffalo you throw at it: four-plus explosives a week in all five games now, plus coverage flags that donate yardage. Las Vegas had the design and the targets and still fell short because its quarterback was 23-of-41 and its red-zone EPA was zero. A future opponent needs both — the vertical threat and the finishing. To beat Las Vegas you run at it and finish: 5.21 YPC allowed entering, 80% opponent red-zone TD rate, and the 'run game' tag in each of the last two weeks. Buffalo had the rushing efficiency to cash that and did, barely, with three stalled trips of its own.
What the model should take from this
Three things. Buffalo's run lean is not a lead property — price Allen at 32-34 attempts regardless of script. A trailing team's carry projection should be dragged toward its efficiency, not its season volume, when before-contact blocking is this bad. And red-zone splits from one game are nearly worthless as signal: Buffalo kicked twice partly because a holding penalty erased a touchdown, and Cousins' lone interception is the least persistent number in the file.
Bills offenseagainst theRaiders defense
Projected BUF offense output — equivalently, what the LV defense allows. The engine blends each offense's season-to-date rates with the defense across from it, anchored on the market's implied points.
| Backs | Car | Ru yds | Ru TD | Tgt | Rec | Rec yds | Rec TD |
|---|---|---|---|---|---|---|---|
| RB | 18(22) | 92(95) | 1.2(1) | 2.2(3) | 1.7(3) | 6(58) | 0(0) |
| RBFP | 0(1) | 1(3) | 0(0) | 3.3(2) | 2.3(2) | 37(24) | 0.2(0) |
| RB | 1(1) | 2(4) | 0(0) | 0.7(1) | 0.5(1) | 11(5) | 0.1(0) |
| Receivers | Tgt | Rec | Yds | TD |
|---|---|---|---|---|
| WR | 5.6(6) | 4.1(4) | 53(34) | 0.3(1) |
| WRFP | 4.4(6) | 2.7(4) | 44(59) | 0.3(1) |
| WRQ | 3.4 | 2.1 | 24 | 0.1 |
| WRQ | 2.5(2) | 1.6(1) | 14(37) | 0.1(0) |
| Tight ends | Tgt | Rec | Yds | TD |
|---|---|---|---|---|
| TE | 4.0(8) | 2.9(7) | 32(38) | 0.2(0) |
| TE | 2.7(1) | 1.9(1) | 8(0) | 0(0) |
BUF offenserank of 32LV defenseEPA / play
Expected points added per play, garbage time excluded: the one-number read of an offense, and of what a defense allows.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Success rate
Share of plays that gain enough to stay on schedule.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Points / game
Points scored per game against points allowed per game.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Plays / game
Offensive snaps per game, a pace number: more plays means more volume to go around, not a better team.
Engine 60.4: the blended per-game value the projection used for this game; the league average is 60.8.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Pass rate
Share of snaps that are dropbacks: how the volume splits between the pass and the run.
Engine 54%: the blended per-game value the projection used for this game; the league average is 56%.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Explosive plays
Share of plays that go for a chunk: 20+ passing, 10+ rushing.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Third down
The offense's third-down conversion rate against the defense's third-down stop rate.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Red-zone TD rate
Touchdowns per trip inside the 20: finishing drives against the defense that bends or breaks.
Engine 65%: the blended per-game value the projection used for this game; the league average is 58%.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Giveaways / game
Turnovers the offense gives away per game against the ones the defense takes.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Flags / game
Accepted penalties per game committed by each unit.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
BUF offenserank of 32LV defenseRush EPA / carry
Expected points added per designed run, and per run allowed.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Rush yards / game
Rushing yards per game, gained and allowed.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Yards / carry
Rushing yards per carry: the offense's own against what the defense gives up, blended into the projection.
Engine 4.81: the blended per-game value the projection used for this game; the league average is 4.36.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Yards before contact
Yards before a defender touches the runner, per carry: the line-and-scheme share of the run game (PFR charting). Feeds each back's efficiency tilt.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
O-line grade
A composite of the line's pass protection and run blocking, offense only.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Box defenders
Average defenders the defense puts in the box on charted snaps (FTN). Ranked most to fewest with no good or bad: a heavy box crowds the run and dares the pass. Defense only; a look, not a grade.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
BUF offenserank of 32LV defensePass EPA / play
Expected points added per dropback, and per dropback allowed.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Pass yards / game
Passing yards per game, gained and allowed.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Yards / attempt
Passing yards per attempt, blended into the projection.
Engine 7.78: the blended per-game value the projection used for this game; the league average is 7.02.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Completion rate
Completions per attempt thrown, and per attempt faced.
Engine 65%: the blended per-game value the projection used for this game; the league average is 64%.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Passer rating
The quarterback rating the offense posts against the one the defense allows.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Sack rate
Sacks per dropback: taken by the offense, generated by the defense.
Engine 7.1%: the blended per-game value the projection used for this game; the league average is 6.4%.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Pressure rate
Sacks or quarterback hits per dropback, the play-by-play pressure proxy.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
INT rate
Interceptions per attempt: thrown by the offense, forced by the defense.
Engine 2.6%: the blended per-game value the projection used for this game; the league average is 2.2%.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Blitz
Offense: its EPA per dropback when the defense sends five or more rushers, higher is better. Defense: the share of its pass snaps it sends five or more, ranked most to least with no good or bad (FTN charting). Display only, not an engine input.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Yards / catch
Receiving yards per reception, offense only: where the ball goes once it is caught.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
DPI yards / game
Pass-interference yards per game: drawn by the receivers, committed by the secondary.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Raiders offenseagainst theBills defense
Projected LV offense output — equivalently, what the BUF defense allows. The engine blends each offense's season-to-date rates with the defense across from it, anchored on the market's implied points.
| Backs | Car | Ru yds | Ru TD | Tgt | Rec | Rec yds | Rec TD |
|---|---|---|---|---|---|---|---|
| RBFP | 17(10) | 67(22) | 0.4(1) | 5.3(6) | 3.9(4) | 38(36) | 0.3(0) |
| RB | 5(9) | 24(43) | 0.2(0) | 0.7(1) | 0.5(1) | 1(7) | 0(0) |
| RB | 2(1) | 7(0) | 0(0) | 0.8(1) | 0.5(1) | 4(-1) | 0(0) |
| Receivers | Tgt | Rec | Yds | TD |
|---|---|---|---|---|
| WR | 4.9(7) | 3.0(5) | 42(97) | 0.3(2) |
| WR | 2.7(2) | 1.8(0) | 19(0) | 0.1(0) |
| WRFP | 2.6(4) | 1.6(1) | 13(9) | 0.1(0) |
| WR | 1.7(0) | 1.0(0) | 2(0) | 0(0) |
| Tight ends | Tgt | Rec | Yds | TD |
|---|---|---|---|---|
| TEFP | 7.8(9) | 5.6(4) | 60(70) | 0.6(0) |
| TE | 3.9(9) | 2.7(7) | 25(95) | 0.2(0) |
LV offenserank of 32BUF defenseEPA / play
Expected points added per play, garbage time excluded: the one-number read of an offense, and of what a defense allows.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Success rate
Share of plays that gain enough to stay on schedule.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Points / game
Points scored per game against points allowed per game.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Plays / game
Offensive snaps per game, a pace number: more plays means more volume to go around, not a better team.
Engine 64.3: the blended per-game value the projection used for this game; the league average is 60.8.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Pass rate
Share of snaps that are dropbacks: how the volume splits between the pass and the run.
Engine 58%: the blended per-game value the projection used for this game; the league average is 56%.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Explosive plays
Share of plays that go for a chunk: 20+ passing, 10+ rushing.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Third down
The offense's third-down conversion rate against the defense's third-down stop rate.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Red-zone TD rate
Touchdowns per trip inside the 20: finishing drives against the defense that bends or breaks.
Engine 63%: the blended per-game value the projection used for this game; the league average is 58%.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Giveaways / game
Turnovers the offense gives away per game against the ones the defense takes.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Flags / game
Accepted penalties per game committed by each unit.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
LV offenserank of 32BUF defenseRush EPA / carry
Expected points added per designed run, and per run allowed.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Rush yards / game
Rushing yards per game, gained and allowed.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Yards / carry
Rushing yards per carry: the offense's own against what the defense gives up, blended into the projection.
Engine 3.69: the blended per-game value the projection used for this game; the league average is 4.36.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Yards before contact
Yards before a defender touches the runner, per carry: the line-and-scheme share of the run game (PFR charting). Feeds each back's efficiency tilt.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
O-line grade
A composite of the line's pass protection and run blocking, offense only.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Box defenders
Average defenders the defense puts in the box on charted snaps (FTN). Ranked most to fewest with no good or bad: a heavy box crowds the run and dares the pass. Defense only; a look, not a grade.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
LV offenserank of 32BUF defensePass EPA / play
Expected points added per dropback, and per dropback allowed.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Pass yards / game
Passing yards per game, gained and allowed.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Yards / attempt
Passing yards per attempt, blended into the projection.
Engine 7.21: the blended per-game value the projection used for this game; the league average is 7.02.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Completion rate
Completions per attempt thrown, and per attempt faced.
Engine 65%: the blended per-game value the projection used for this game; the league average is 64%.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Passer rating
The quarterback rating the offense posts against the one the defense allows.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Sack rate
Sacks per dropback: taken by the offense, generated by the defense.
Engine 6.2%: the blended per-game value the projection used for this game; the league average is 6.4%.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Pressure rate
Sacks or quarterback hits per dropback, the play-by-play pressure proxy.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
INT rate
Interceptions per attempt: thrown by the offense, forced by the defense.
Engine 2.2%: the blended per-game value the projection used for this game; the league average is 2.2%.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Blitz
Offense: its EPA per dropback when the defense sends five or more rushers, higher is better. Defense: the share of its pass snaps it sends five or more, ranked most to least with no good or bad (FTN charting). Display only, not an engine input.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
Yards / catch
Receiving yards per reception, offense only: where the ball goes once it is caught.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
DPI yards / game
Pass-interference yards per game: drawn by the receivers, committed by the secondary.
Ranks run 1 to 32 on each side, best first: green is a top-10 unit, amber a bottom-10 one.
The elevens, rated
Defenders colored by rating percentile within their position group (2026 wks 1-5 + ½·2025); the offense by production percentile among the position's starter pool. Linemen are unrated. Each defense is last game's snap leaders in a nickel base, minus this week's Out / Doubtful.
Who is listed
Everyone on the clubs' reports, the best-rated man first in his group: the designation, what is hurt, the practice line and the share of the snaps he took last game. Ratings are the same percentiles as the field above.