TiltDataLabsUnlisted preview

Weekly NFL Highlight

on EdgeOfSunday

Record6-3

One pick a week, published before kickoff and settled in the open — win or lose.

Week 10 · Sunday 1:00 PM ET

KCKansas City at AtlantaATL

The market says Kansas City by 3 on the road. The model projects Kansas City by 7.1 — a difference of 4.1 points. The side that follows is KC -3.

The disagreement

Our pick is Kansas City -3. Kansas City is favored by 3 at Atlanta, the number at every book we track but FanDuel, which holds 3.5 at +100. The model has Kansas City by 7.1, a 4.1-point disagreement. Where the number sits matters. Three is the most common final margin in football, 14.6% of the 2,097 settled games in the model's training data, and four is another 5.0%. At -3 a Kansas City win by three pushes and a win by four covers; at -3.5 the three-point win loses. The 7.1-point projection clears both and sits just past seven, the next key number at 8.4%. At 3.5 the model puts Kansas City's chance of covering at 51.7%, against the 50.0% that an even-money price needs.

Kansas City is 9th in recency-weighted net EPA at +0.132, and Atlanta is 21st at -0.116. Kansas City's offense ranks 10th at +0.073 against Atlanta's 24th at -0.085, and its defense ranks 11th at -0.059 against Atlanta's 21st at +0.031, where lower is better. Kansas City's turnover margin of +0.299 ranks 8th. Both special teams units are near the bottom: Kansas City's ranks 28th at +0.008 and Atlanta's 31st at -0.040.

Kansas City is 6-2 and has outscored its opponents 202-143. Two of its six wins came by exactly three, 33-30 over Indianapolis and 30-27 at Las Vegas. It lost 20-3 at Seattle and 28-24 at Denver before beating the NY Jets 35-12 last week. Atlanta is 3-6 and has been outscored 223-188. All six of its losses came by seven points or more, the last two 16-0 at Tampa Bay and 34-17 to Cincinnati. Its three wins were 35-14 at Green Bay, 45-24 at New Orleans and 29-20 over San Francisco, the only team to beat the 49ers this season.

The ratings

2025 ratingRankNet EPAOffDefSTSuccessTO margin
KCKansas City9+0.132+0.073−0.059+0.00847.1%+0.299
ATLAtlanta21−0.116−0.085+0.031−0.04042.2%−0.198

Net EPA per play, averaged over the rating window stated at the foot of this page. For defense, negative is good — it is points of expected value allowed. Rank is out of 32.

The quarterback input

This is the only path from an injury to the projection, so it is worth showing rather than describing. The model reads the expected starter’s rating:

Expected starterRatingPlays
Kansas City+0.15585%
Atlanta+0.035100%

When a starter is ruled out, the depth chart promotes his backup, so the play probability above reads 100% and the post-model quarterback adjustment does not fire. The absence shows up in the rating instead, which falls to the backup’s — and that rating is one of the twenty features the model selects. A play probability below 100% with no designation filed is the default weighting applied to any unfiled starter, not a report that he is hurt.

How stable is it

Every input is perturbed by the standard error of a full 17-game season average and the game re-projected, four hundred times. The question each trial answers is whether the model still favors KC -3 after the perturbation.

It does in 95.8% of trials at that realistic noise level, and 84.0% when the noise is doubled.

The other side of this game is priced too — ATL +3 — and the model puts it at 48.1% against 50.5% implied by its own price. It is the same projection seen from the other end, not independent evidence — preferring this side and declining that one are one decision, not two.

Patrick Mahomes practiced in full with no designation and reads a play probability of 0.85; Atlanta's quarterback reads 1.00. The net downgrade is 0.040, under the 0.05 trigger, and the quarterback adjustment does not fire.

The market

KCOpenNowTicketsMoney
KC -3-4.5-3.587%84%

KC opened at -4.5 and sits at -3.5: the market has moved 1 point toward ATL. 87% of tickets and 84% of the money are on this side. That is reverse line movement: the line moved toward ATL while the majority of tickets were on the other side, which usually means the move came from bettors the public splits do not count.

Kenneth Walker III is the injury that matters, and it is not in the number. He is questionable with a groin injury after not practicing all week, at 71.7% of the running-back snaps. He has carried 159 times for 5.5 yards a carry and a 46.5% success rate, against a league average of 3.9 yards and 39.2% on designed runs. If he cannot go, the carries go to Emmett Johnson, a rookie taken 161st overall out of Nebraska, who has 26 NFL carries: 4.3 yards a carry and a 46.2% success rate. Those 26 carries are his whole track record. The model has no injury input beyond the quarterback adjustment, so the 7.1 is the same with or without Walker.

Public betting splits from Action Network, captured 2026-11-13T17:36-05:00, across 15,787 tracked bets. They describe the market; they are not an input to the model.

Injuries

The side held in 95.8% of trials and 84.0% under doubled noise. Put plainly: every input to the model was nudged at random by the standard error of a season's average, 400 times over, and Kansas City stayed ahead of 3.5 in 383 of them. At twice that noise, 64 of the 400 flipped.

Final report: Wednesday for a Thursday game, Friday for the weekend. A player ruled out or listed as doubtful counts as not playing; a questionable non-quarterback plays about 62% of the time, a questionable quarterback about 35%. Reserve lists are folded in separately because players on IR never appear on a weekly report.

KCKansas Citydepletion 1.386

Peter WoodsDTdid not practiceNeck — Out29.7% of snaps
Kenneth Walker IIIRBdid not practiceGroin — Questionable71.7% of snaps
Jaylon MooreTdid not practiceAnkle — Out17.2% of snaps
Tyquan ThorntonWRdid not practicereserve/injured — Out15.5% of snaps
Patrick MahomesQBfullno designation filed98.9% of snaps
L'Jarius SneedDBdid not practicereserve/injured — Out13.1% of snaps
Mansoor DelaneCBfullShoulder — Questionable16.2% of snaps
Xavier WorthyWRfullno designation filed21.2% of snaps
Travis KelceTEfullno designation filed19.8% of snaps
Rashee RiceWRfullno designation filed15.7% of snaps

ATLAtlantadepletion 1.306

Divine DeabloLBdid not practiceForearm — Doubtful33.0% of snaps
Samson EbukamDEdid not practiceHamstring — Out30.9% of snaps
Brian Robinson Jr.RBdid not practiceHamstring — Out30.3% of snaps
Cameron ThomasDEfullno designation filed42.8% of snaps
Zachariah BranchWRdid not practiceHamstring — Out5.9% of snaps
Kendal DanielsLBfullno designation filed29.6% of snaps
Da'Shawn HandDLdid not practicereserve/injured — Out4.4% of snaps
Michael JerrellOLdid not practicereserve/injured — Out4.3% of snaps
Jawaan TaylorTlimitedno designation filed19.0% of snaps
Jake MatthewsTlimitedno designation filed18.0% of snaps

Depletion is the share of a normal week’s snap workload expected to be missing: each player’s share of his position group’s snaps in prior games, multiplied by his probability of not playing. 1.00 is roughly one full starter’s workload. It is context, not a model input.

The case against

The case against starts with Kansas City's form: it has lost two of its last three, and scored 3 points at Seattle. Two of its six wins came by exactly three, the margin that pushes at -3 and loses at -3.5. Its special teams rank 28th, its lead running back is questionable without a practice all week, and his replacement is a rookie with 26 carries; none of that is in the 7.1. Atlanta is the only team to beat San Francisco this season, and it won by three touchdowns at Green Bay.

How the model works

Play-by-play from every game since 2019 — roughly 345,000 plays — plus snap counts, depth charts and multi-book closing lines. An ensemble of gradient-boosted trees, a regularized linear model and a random forest predicts the point margin directly; that continuous prediction is converted into a probability against any line using the model’s own out-of-sample error distribution.

Twenty features, ranked fresh each week inside the training window and never using the games being predicted. Validation is strictly walk-forward: every season is predicted using only the seasons before it. Betting odds are never permitted as model inputs — the model is not allowed to see the market it is being asked to beat.

See every Week 10 game on the edge board →

2026 results

Kansas City6–2
WkOpponentScoreMargin
1vs DENW 31–10+21
2vs INDW 33–30+3
3at MIAW 24–10+14
4at LVW 30–27+3
6vs LACW 22–6+16
7at SEAL 3–20-17
8at DENL 24–28-4
9vs NYJW 35–12+23
Season202–143+59
Atlanta3–6
WkOpponentScoreMargin
1at PITL 13–20-7
2vs CARL 3–34-31
3at GBW 35–14+21
4at NOW 45–24+21
5vs BALL 23–30-7
6vs CHIL 23–31-8
7vs SFW 29–20+9
8at TBL 0–16-16
9vs CINL 17–34-17
Season188–223-35

Every game played this season, with the margin. These are results, not ratings: the ranking above is recency-weighted expected points added per play, which is why a team’s record and its rank can disagree.

Numbers frozen from the 14 November 2026 run (9:36 AM ET), on Friday’s final injury reports; lines checked 14 November at 11:44 AM ET. Ratings weight every game of 2025 and 2026 by recency, halving every five games, so 2026 carries about two-thirds of the weight; they are not opponent-adjusted. Lines move and designations land on Friday; a pick can change between now and kickoff.

This is a sample of the weekly NFL write-up. The live board — every game on the slate, both sides, model probability against the best available price — sits behind the subscriber paywall at tiltdatalabs.com/dashboard/nfl.