edge lab · factor writeup · 6 min read

Defensive Line Strength: Does the Better Front Beat Its Price?

Season-to-date sack differential looks predictive until you notice it correlates 0.51 with the moneyline itself. Strip the price out and the signal disappears.

Every survivor column eventually reaches for the trenches. Take the team with the better line, the argument goes, because pass rush travels and pressure decides games. It is a good story, and the raw numbers appear to back it. They back it for a reason that turns out to be the whole finding: the betting market already knows.

What we tested

The hypothesis: a team whose defensive line and pass protection are better than its opponent's wins more often than its de-vigged market price implies. Not "wins more often" — that part is obvious — but wins more often than the price already says.

How we tested it

The metric is a two-sided net sack differential, computed season-to-date through the prior week only, so nothing from the game being predicted leaks into its own prediction: (own sacks made per game minus own sacks allowed per game) minus the same figure for the opponent. Regular season only, and both teams needed at least four completed games, which starts the usable sample in week 5. That left 2,052 team-weeks across 2021 through 2025, joined to real de-vigged bookmaker moneylines and real final scores with no unmapped teams.

The first thing that fell out was not the answer to the hypothesis — it was a warning. The differential correlates with the market's de-vigged win probability at r = 0.509. Half the variance in "who has the better line" is already inside the price. Run a raw correlation against outcomes and it comes back positive, and it means nothing, in exactly the way our pick-popularity proxy looked predictive until it was separated from win probability.

So the real test regresses de-vigged win probability out of the differential first and keeps only the residual — the part of line strength the market has not already priced. That residual is then tested against each team's outcome relative to its implied win probability.

What the numbers said

Sample
2,052 team-weeks
Window
2021–2025, weeks 5–18
Collinearity
r = 0.509
Residual signal
r = -0.006, z = -0.27

The residual is flat. Correlation with outcome-versus-price is -0.006, a z of -0.27, which is indistinguishable from zero in a sample this size. Sorting the residual into five equal buckets does not rescue it: every bucket finishes within 1.2 percentage points of its implied win rate, the largest bucket-level z is 0.52, and the pattern is not even monotonic — the bucket with the strongest line edge finished slightly below its price.

  • Weakest residual bucket (411 team-weeks): 51.3% actual vs 50.2% implied, z = 0.52.
  • Middle bucket (410 team-weeks): 50.0% actual vs 50.1% implied, z = -0.04.
  • Strongest residual bucket (410 team-weeks): 48.8% actual vs 49.9% implied, z = -0.51.

We ran the same test on Pro Football Reference's charted pressure data instead of sacks, because pressure is the cleaner measure of a pass rush. It behaves identically: collinearity of 0.537 with the price, residual r = 0.013, z = 0.59. Worth a coverage caveat — that source averages 27.5 charted player rows per team-week (range 11 to 39), so it is thinner and noisier than the box-score sacks, and its flat result is a weaker flat than the sack version's.

Three weeks where the line said one thing and the price said another

These are the real cases where the residual pointed hardest against the market — the underdog with the much better trench profile.

Across the eight largest such divergences, three of the eight underdogs won against an average implied rate of roughly 33%. That is what "no edge" looks like up close: the extreme cases land almost exactly where the price said they would.

Already in the number

Line strength is not irrelevant to football. It is irrelevant to you, because by the time you see the moneyline, someone has already charged you for it.

What it means for your picks

Do not override a survivor pick because one team's front looks better on paper. Half of that judgement is already inside the price you are reading, and the half that is not has no measurable effect on who wins. This is the tenth factor tested here and the seventh to come back flat, which is the ordinary result: markets are efficient about the things everybody can see, and a defensive line is about as visible as football gets.

Nothing in the live equity engine changed on the back of this test. Every factor we have run, including the ones that worked, is in the Edge Lab, and the pricing method itself is in the methodology.

Track your real pool free — 10-day trial, no card →