NFL Trap Games and Survivor Pools: What the Data Actually Says

September 11, 2026

Somewhere around the middle of every season, a broadcast crew reaches for the phrase. Good team, weak opponent, marquee game waiting next Sunday: “this one has trap game written all over it.” It is one of the most durable ideas in football commentary, and it has leaked into survivor pool advice, where people now pass on good favorites because a matchup feels like a trap.

The idea deserves a fair hearing, because the story is plausible. Football is played by people. A roster that spent all week hearing about next week’s opponent might practice a little sloppily. A team coming off an emotional rivalry win might come out flat against a bad one. Short week, long flight, division game on the horizon: you can see the mechanism.

The problem is that when people went looking for it in actual results, it wasn’t there.

Somebody actually tested this

In November 2012, Kevin Meers took up a challenge from Aaron Schatz and posted the first serious attempt at an answer on the Harvard Sports Analysis Collective, a student-run undergraduate analytics blog. It is one person’s blog post, not a peer-reviewed study, and the Harvard in the name is doing less work than it looks like. But it is carefully done, and it was the first place anyone bothered to define the term before arguing about it.

The definition came first, since the phrase is normally used without one. He settled on something close to how announcers use it: a team that finished the season above .500 playing a team that finished below .500, one game before facing another above-.500 opponent.

Then every game between an above-.500 and a below-.500 team from 2002 through 2011. Across 1,359 games the better team won 79.5 percent of the time, which is the baseline any trap-game effect has to beat.

In the 515 games that fit the trap definition, the better team won 80.5 percent of the time. Higher than baseline, not lower, and not statistically significant. His t-value came out at 0.60. Trap spots were indistinguishable from ordinary spots.

Worth being precise about what a test that size can and can’t see. At 515 games and roughly an 80 percent win rate, the standard error is about 1.75 percentage points, so the 95 percent interval runs about 3.4 points either side of the result. An effect big enough to change a survivor pick (five or ten points of win probability) would very likely have turned up, and didn’t. A one-point effect could hide in there, and in a sample that size it always will.

He ran the letdown version too, defined as a good team facing a weak opponent the week after beating another good team. In those 242 games the favorite won 82.2 percent of the time, again slightly above baseline and again not significant.

We ran it again on the modern game

A 2012 result is a reasonable thing to be suspicious of in 2026. Fourteen more seasons have been played since, under a collective bargaining agreement that changed how teams practice and a 17-game schedule that didn’t exist when he wrote. So we repeated the test.

Everything below is our own calculation from public game data, not a published result. Regular season only, 2012 through 2025, matching Meers’s setup as closely as we can: a team that finished the season with a winning record, facing a team that didn’t, one game before playing another winning-record team. Ties count as non-wins. As a check on the method, running the same code over his window returns 1,358 qualifying games at a 79.5 percent baseline, against his published 1,359 and 79.5 percent.

Over 2012 through 2025 there were 1,917 games between a winning-record team and a team without one. The better team won 1,535 of them, 80.1 percent. In the 842 of those that also fit the trap definition, the better team won 675. That’s 80.2 percent.

One tenth of one percentage point, on a sample 63 percent larger than the original, drawn from a completely different era of football. Fourteen more seasons still cannot find the trap game.

The letdown half is a different story

Here is where we have to be straight with you, because the same run turned up something the 2012 post didn’t.

Letdown games (a good team playing a weak opponent the week after beating a quality one) came in at 76.7 percent over 2012 through 2025, across 420 games, against that same 80.1 percent baseline. That is 3.4 points below, and it does not clear the usual bar for significance: a two-tailed z-test of the subgroup against the full-sample rate lands around p = 0.08. Narrow the window to 2019 through 2025 and the gap widens rather than shrinking: 73.4 percent over 214 games against an 80.3 percent baseline, which does clear it.

Now the reasons not to believe us, all of which we think are serious. That second window is a slice we cut after seeing the full-period result, which is the cheapest way there is to manufacture a finding. None of this is adjusted for team strength, opponent quality, or home field, and “a good team coming off a big win” is partly just a team whose schedule has been hard and isn’t about to get easy. And we ran several cuts of this data; one of them landing under p = 0.05 is roughly what you would expect even if nothing at all were going on.

So we are not claiming letdown games are real. We are saying that our own numbers don’t let us rule them out, and that an article which spends this many words on other people’s unfalsifiable claims doesn’t get to quietly drop its own inconvenient one. The trap game we can rule out at any size worth caring about. The letdown game is unsettled, and anyone who tells you otherwise in either direction is ahead of the evidence.

The adjacent folklore

Trap games travel with a family of scheduling superstitions, and most of those have gotten the same treatment lately.

Rest is the big one. The bye-week advantage used to be real: two NFL data scientists, writing in Frontiers in Behavioral Economics in 2024, used Bayesian state-space models to estimate rest effects and found that before 2011, teams coming off a bye were worth about 2.2 extra points per game. Then the collective bargaining agreement changed the rules on when teams could practice during a bye. Since that change the estimated advantage is about 0.31 points, with a credible interval wide enough to include zero in both directions, and neither the bye nor the so-called mini-bye currently shows a significant edge. What used to help was never the rest. It was the practice time, and the new rules took that away.

The West Coast body-clock curse is thinner than it sounds too, though the public work on it is thinner still. The most thorough version we have found is a 2025 post by the analyst Max Moacanin, who pulled 36 seasons of Pacific-time teams traveling east, 1989 through 2024. Their raw record is losing, which is how the myth survives. Measured against what his own Elo model expected of those same teams, they came in slightly ahead rather than behind: 46.6 percent in early kickoffs against a 43.8 percent expectation. Take that for what it is. The benchmark is one analyst’s unvalidated model, and he is careful to say the gap isn’t significant. It is not proof the curse is fake. It is the best available look, and it does not point the way the curse needs it to.

Short weeks: the market got there first

Short-week road trips are the piece of this folklore with a real-looking gap behind it, so it’s worth walking through slowly.

Again, our own calculation from public game data, not a published result: every regular season game from 2006 through 2025, Week 2 onward. Week 1 is excluded because no team has a previous game to rest from, so its rest values are a placeholder rather than a measurement.

Road teams on five or fewer days of rest won 44.2 percent of the time (n=283). Road teams arriving on more than seven days won 46.9 percent (n=994). That looks like a rest effect, and it’s the split people quote.

Notice first that the ordinary Sunday-to-Sunday road team, on exactly seven days, won 43.2 percent (n=3,261), below the short-week group. Whatever is happening at the long-rest end, nothing is happening at the short end.

Then price it. Strip the vig out of the closing moneylines on both sides of each game and you can ask what the market expected of these same road teams. Short rest: expected 44.4 percent, won 44.3 percent (n=282). Seven days: expected 42.9 percent, won 43.2 percent (n=3,186). More than seven days: expected 45.0 percent, won 47.1 percent (n=983).

The market prices short-rest and long-rest road teams within two-thirds of a point of each other, and both above the ordinary road team. That is most likely because both groups are disproportionately the teams networks want in standalone windows, not because of the calendar. Short-week road teams then finish one tenth of a point off what was expected of them. The only residual left is the long-rest group, two points ahead of its price, which is about one and a half standard errors on that sample. That is well inside noise, and about what the Frontiers work above would lead you to expect from a modern bye.

Both groups are also confounded in obvious ways. Of the 283 short-rest road teams, 272 are Thursday-night visitors. That isn’t a rest study, it’s a Thursday-night study. And the long-rest group is three different things in a trench coat: roughly a third off a full bye, a third off the mini-bye that follows a Thursday game, and a third just playing a Monday night after the previous Sunday.

So the schedule does less than the folklore needs it to. What little shows up points at rest surplus rather than deficit, none of it survives contact with the closing line, and the people setting that line have already charged you for whatever is left.

Why none of this helps a survivor player anyway

Set the evidence aside and grant, for argument’s sake, that trap games exist. It still wouldn’t change your pick.

Trap-game talk is cover talk. When a commentator warns you about a trap, the implicit claim is that a favorite will play down to its opponent: win 24-21 instead of 34-13. That’s an against-the-spread idea, and survivor pools don’t pay for margin. They pay for outright wins.

Which is the part nobody can find. Every version of the test above says the same thing a different way: whatever trap-game talk is describing, it does not show up in who wins. There is no measured shave to apply. Discounting a favorite for it means discounting it for an effect that two decades of results can’t locate.

Then there’s the pricing problem. The short week, the cross-country flight, the division rival waiting next Sunday: all of it was on the schedule in May, and the people setting the number know about it. By the time a storyline is obvious enough for a broadcast to mention, it has been in the price for months. Same reason expected value in survivor pools starts with the market line rather than trying to out-argue it.

The part worth keeping: fade the narrative, don’t obey it

Here’s where the trap game becomes useful, just not in the direction people expect.

A trap-game reputation moves picks. That half is reasoning rather than measurement (nobody publishes pool-wide ownership data clean enough to prove it), but it follows from watching how people talk themselves out of a pick. It doesn’t move outcomes, and in survivor those are two different currencies. If your pool spends the week hearing that a strong favorite is walking into a trap, some share of your pool takes somebody else. That favorite’s ownership drops while its win probability sits exactly where it was.

Low ownership on a high win probability is the whole game. If the week goes badly for the chalk you’re one of a small group left standing, and if it goes fine you gave up nothing to be there. The pick-equity argument makes that arithmetic explicit; the short version is that a team the crowd is avoiding for unserious reasons is the cheapest edge available in an average week.

Which flips the advice. Don’t dodge the trap game. Notice who else is dodging it.

The traps that are actually real

There are traps in survivor. They’re structural, not narrative, and it’s worth ranking them by how many seasons they end.

The consensus pick. This is the big one, and 2024 is the exhibit. ESPN’s free Eliminator Challenge drew 942,806 entries that year. Through three weeks, 94.5 percent of them were gone, leaving 52,049 alive. The most-selected team in the contest lost every single week, and in Week 3 four of the top five most-selected teams lost, taking 77.2 percent of the surviving field down in that single week. The high-stakes version went the same way: Circa Survivor, with 14,266 entries, was down to 642, a 95.5 percent elimination rate. Circa Sports CEO Derek Stevens told ESPN he had been playing these pools for roughly 30 years and had never seen anything like it. To be fair to a normal bad year: the same ESPN piece notes that since 2020 the previous low through three weeks was 60,690 entries out of 503,963, about 12 percent, in 2022. So 2024 is the worst case, not the typical one, and Stevens put it down to parity and to the fact that not one double-digit point-spread favorite appeared through those three weeks. But notice what that trap was built out of. Not a lookahead spot or a letdown spot. Chalk, held by almost everybody at once.

Week 17 and 18 rest situations. The one genuinely schedule-driven trap, and it’s real because the incentive is real. A team with a playoff seed already locked has little reason to expose its starters, and that decision often surfaces late in the week, after many pools have locked. You aren’t reading motivation, you’re reading a roster that may not be the roster you picked.

A quarterback ruled out after your pool locks. Pools lock at the first kickoff of the week. News does not. Any pick you submit early is a pick made without information that will exist before kickoff, and the earlier the deadline, the more of that information you’re giving up.

Divisional familiarity. A note of caution rather than a number. Teams that play twice a year know each other’s tendencies and the staffs have years of tape, and the widespread belief is that this compresses margins. The rigorous public work is thin and the numbers in circulation are mostly cherry-picked, so treat it as a small haircut on your confidence in a divisional favorite, not a rule. It matters least when the talent gap is enormous, which is how the games against the league’s bottom tier stay usable.

The team you already spent. The trap that ends the most survivor seasons isn’t a game at all. It’s arriving in Week 13 with a slate full of coin flips and discovering the team that would have carried you was burned in September on a game a mid-tier team could have handled. That’s the entire subject of the complete survivor pool strategy guide, and it costs more entries than every narrative on this page combined.

What to do instead

When somebody tells you a game is a trap, run three checks and move on.

Start with the line, because it already contains the short week, the travel, and next week’s opponent. Then check ownership, because that’s the number the narrative actually moved: if the crowd is off a strong favorite for a soft reason, you may have found a cheap week. Then check your own inventory and ask the only question that has ever mattered here: can a team I care less about get me through instead?

New to the format? Start with how survivor pools work, because everything above assumes you already know that a pick is a one-time expense.

Trap games are a way of talking about football, not a way of deciding anything. Survivor Caddy is built around the things that actually move a pick: our own Elo-based power ratings, the future cost of every team, and the deadline that’s sneaking up on you. Join the waitlist and we’ll email you once at launch.