How Boom Matchups works

The honest answer first

We wanted a board that beats the spread. So we built several and tested every one of them the same way we test Boom Yards and Boom TD!: on seasons the model never saw. None of them beat the closing line. To make money at −110 you have to win 52.4% of the time. Here's how ours did:

The closing spread already holds everything we measure, plus injuries, weather and the sharpest money in the world. So we do show our own number on every game, with every ingredient and its record, to explain the game, but never as a pick to bet against the line. Our player boards are different: player props are priced less carefully, and that's where our model has tested well on seasons it never saw.

Betting trends: mostly noise

We tested the popular angles over 20 seasons, split into two eras (2006–15 / 2016–25) to see which ones hold up:

Test fifteen trends and a couple will land near 55% by chance alone. None is strong enough to build on, so we don't sell them to you.

What does work: every line as a real chance

The line is the best guess at the middle of what happens. Bettors need the whole range: how often does a 2.5-point favorite win by 7? How often does the underdog win outright? We answer that from 5,246 NFL games (2006–2026): for each game, the past games with a similar line, and how they actually finished. Every game page has a “Show the games” chart, so you can see them yourself.

That gives every game a ladder: each line, its real chance, and the fair odds. Compare the fair odds to your book's price, and you know whether a bet is worth it.

Our accuracy check, re-run on every update

We hide the most recent 5 complete seasons (2021–2025, 1,359 games), predict them using only 2006–2020, then compare with what actually happened. “The line + 3” asks: did the favorite win by more than the spread plus 3? (For totals: did the final score beat the posted total plus 3?)

Where we're off, we say so: the rows in red missed by 2 points or more. Games have landed on exactly 7 less often lately than history says (4.9% vs 6.3%), so treat the 7 numbers with a little care. Last checked Sep 29, 2026, 1:47 PM ET.

Key numbers: why half a point can matter

NFL margins bunch up on 3 and 7 (a field goal, a touchdown). When a team is favored by 3, it wins by exactly 3 in 9% of games. Favored by 7, it wins by exactly 7 in 6%. So when the line is 2.5, the underdog at +2.5 covers 50%, but at +3.5 it covers 59%.

For your bet: On lines of 2.5, 3.5, 6.5 and 7.5, the half-point is worth real money. Each game's page shows exactly what it's worth, so you know how much extra juice to pay to buy it. Totals don't have key numbers like this, so don't pay to move a total.

Touchdowns move games

From our TD projections, made before every game (2019–2026):

For your bet: A player's TD and his team covering (or the over) go together: the pair hits about 19% more often than the two chances multiplied. That's the math behind a good same-game parlay. Books price some of this in, so compare the parlay price to the real chance.

Power rankings

Every play this season, adjusted for the opponent, turned into points: “+4.8” means about 5 points better than an average team on a neutral field. We rank each unit too (pass and run, offense and defense), and show the market's rating next to ours.

• The stat: EPA per play: expected points added, how much each snap changed the expected score. It rewards a 3rd-and-8 conversion more than a 3-yard run on 3rd-and-8.
• Opponents: every team's offense and defense are solved together (a ridge regression, strength 2), with home field, so beating up a bad defense counts for less.
• Weights: every game this season counts the same; last season's games count 25%. We tested this on 2019–25: counting the latest games more (hot streaks) predicted the rest of the season worse, not better, and last season at 10–25% beat ignoring it or counting it fully.
• Points: EPA per play × 33.4, the scale that best turned ratings into final margins on 2017–20 games.
• The market's rating: the same model solved from this season's spreads instead of plays (each line = one team's rating − the other's + home field).
• Updated: Sep 29, 2026, 2:37 PM ET, through Week 4.

When the two disagree, we give the reason only when the data shows one, like a new starting QB. They update every morning with the pipeline. They explain the line; they don't beat it.

How much do three games tell you?

Every season is different, so we checked how well each view after three games matched where teams actually finished (1.0 would be perfect):

Last season matters less every year: rosters turn over faster. Three games in, the plays already say more than last season does, about as much as the market. But three games still leave a lot unknown, which is one reason our number and the line can disagree early. Research September 2026.

Edges and pace

Edges are this week's biggest unit mismatches (say, the best passing offense against a weak pass defense) with the players they move on Boom Yards. Pace is how many plays both teams run, because more plays means more yards to go around.

Our projected spread, and its record

Every game also gets our own number, built from everything we have, with each piece shown on the game's page:

our spread = 1.00 × power-rating gap + 1.95 × expected-TD gap + home field

• Power-rating gap: the two teams' ratings in points, from the rankings above.
• Expected-TD gap: each team's starters' expected touchdowns on Boom TD! (the QB's TD passes count half, since his receivers already hold them).
• Home field: the average home margin over the previous five seasons (2.1 points); zero at neutral sites.
• The weights are whatever best matched final margins in 1,599 games (2019–2024).
• Left out: projected yards. Next to the ratings and TDs, yards got a weight of −3.1 points per 100 yards: negative, a sign they only echo the other two. So they're shown on each game for context, not counted.

The honest record. On 319 games it never saw (2025–2026), our number was off from the final margin by 13.14 points on a typical game; the line by 12.35. Our number was closer 40% of the time. Betting it against the line would have won 43% of 315 games (38% when we disagreed by 3+). That's why it explains games on this board, and why we never post it as a pick.

Each game's projection is locked at kickoff and never rewritten. Games scored after they were played are marked as re-runs and kept apart.

Every formula

• Chance = past games that finished that way ÷ all the past games with a similar line.
• Fair odds from a chance p: if p is 50% or more, −100 × p ÷ (1 − p); if less, +100 × (1 − p) ÷ p. Pushes are set aside first.
• Value per $1 at your price = chance of winning × what $1 wins − chance of losing. Within 2¢ either way, we call it fair.
• Implied score = (total ± spread) ÷ 2: the sportsbook's line split into each team's points. It's theirs, not ours.
• “If he scores” = in past games, how often the team covered (or the game went over) when a starting RB, WR or QB scored, vs when he didn't. Averages across every starter, not one player.

Where the data comes from

• Games and closing lines: nflverse, the open NFL data project (5,246 regular-season games, 2006–2026).
• Play-by-play: every run and pass since 2016, also from nflverse, stored game by game.
• This week's lines and players: the same data our Boom Yards and Boom TD! boards use, refreshed every morning.
• Updates: the numbers on this board rebuild every morning; last run Sep 29, 2026, 1:47 PM ET.

Research September 2026. Data: nflverse play-by-play and closing lines. For entertainment, not betting advice. 21+. Gambling problem? Call 1-800-GAMBLER. Back to this week's games