"He Always Kills This Team": We Tested 5,133 Games. It's a Myth.

"He always kills the Titans." "Division games are different: they know each other too well." "Revenge game, take the over."

You hear these every week. They're some of the best stories in football, and they sound true. So we tested them, the same way we test everything that goes into our boards.

The short version: team-vs-team familiarity doesn't show up in the data. Not in yards, not in touchdowns, not at the team level.

How we tested it#

For every NFL starter-game from 2019 to 2026, we compared what the player actually did with his pregame projection: our expectation for him that week, which already accounts for the opposing defense, the Vegas total and the game script.

That's the key. We're not asking "does Derrick Henry run well against Houston?" (he runs well against everyone). We're asking: does playing a familiar opponent make a player beat or miss his projection by more than chance? Each test also had a control group, so a pattern had to beat what random noise produces.

What we found#

| The claim | What the data says | Verdict | |---|---|---| | Division games (teams that know each other) | Receivers, running backs and tight ends land within 1 yard of their projection. Quarterbacks run 4.6 yards short, inside the noise. | Nothing | | The rematch: does meeting 1 predict meeting 2? | Same opponent: r = +0.02. A different opponent the same number of weeks apart: r = +0.00. Both within ±0.04. | Nothing | | "He always kills this team" (5,133 games with 2+ past meetings) | His past games against this team: weight +0.03 ± 0.05, not significant. His form against everyone: 9× more predictive. | Myth | | Players who "owned" a team before | Today: +0.04 standard deviations against projection (±0.07). Noise. | Myth | | Revenge games (vs a former team, 369 games) | −4 yards and −0.07 touchdowns on average. If anything, slightly worse. | Myth | | Team level: 20 years of division rematches (1,920) | Meeting 1's surprise predicts meeting 2's at r = −0.01. Offenses that exploded in meeting 1 were right on their implied total in meeting 2 (+0.6 ± 1.2 points). | Nothing |

(r measures how well one game's surprise predicts another's: 1 is perfectly, 0 is not at all.)

Two small flickers turned up: quarterbacks 4.6 yards short in division games, and quarterbacks in revenge games about 15 yards short. Both sit inside their error bars. Across roughly 15 tests, a couple of near-misses is exactly what chance produces.

Why the story feels true anyway#

• The projections already cover the real part. How good the defense is, the game script and the Vegas total explain most of what "familiarity" is supposed to explain. Once you account for those, nothing is left.
• The samples are tiny. A player sees a given team once or twice a year, so "he always kills them" usually rests on two to four games. That's a coin landing heads three times.
• What carries over is the player. His form against the whole league predicts next week far better than his history against one team.

What it means for your bets#

• Don't pay for a narrative. If a prop is priced up because of "he always kills them", that's the book (or the crowd) charging you for a story.
• Bet the player's form and the matchup, not the history. How good the defense is this season matters. Who he torched two years ago doesn't.
• We're keeping it out of our model. Adding a familiarity factor would add noise and make the boards slightly worse. It's now on the list of things we tested and left out, on how Boom Yards works.