Fading the backup QB: 178 first starts against the spread
A starting quarterback goes down, and the next week’s advice writes itself: fade the backup. We tested it on every NFL game since 2002 in which the QB1 was ruled out and his backup made his first start of the season, 178 games against the closing spread and 168 with a moneyline. Against the spread the theory is a coin flip, 86-85-7. The backup’s team wins outright a little less often than its moneyline says, but the gap is small, and smaller still in a second build of the data. The books have been pricing backups for a long time.
The theory
The case for fading sounds solid. The backup has had few first-team snaps, the game plan shrinks, the defense knows it, and the line only moves a few points because oddsmakers hedge. If the market underrates how much a team loses with its second quarterback, the other side is a bet. The case against is just as simple: the sportsbooks, the bettors and the models all see the same injury report, and a line that has absorbed all of that by kickoff should be right on average.
Which is it? Only the games can say. So we built the set carefully: not every quarterback change, only the ones the theory is about.
Which games count
- The QB1 went down. The outgoing starter had started the team’s previous three games, and was either the week 1 starter or had started at least half the team’s games that season. That rules out a fill-in handing the job back to the returning starter, which is the opposite situation.
- He was ruled out. Listed Out or Doubtful on that week’s injury report, placed on injured reserve, or on the report and inactive on game day. Benchings are a different story and are kept apart.
- The backup’s first start. The new starter had not started for that team that season. We also followed him into his second and third starts.
- The final regular-season week is left out, since healthy starters rest and the line knows it. Playoff games stay in.
That leaves 178 games from 2002 to week 5 of 2026, scored against nflverse’s closing consensus spread. In 131 of them the backup’s team was the underdog, by a median of 4 points.
Against the spread: 86-85-7
The backup’s team covered 50.3% of the time and lost by an average of 1.9 points more than the spread. That average has a standard error of about a point, so it is consistent with the line being right. No split of the sample is more than noise either: favorites covered less, big underdogs covered more, home and away were level, and the last ten seasons look like the rest.
| Games | Record | Cover rate | Average margin against the spread |
|---|---|---|---|
| All 178 games | 86-85-7 | 50.3% | −1.9 |
| Team an underdog | 66-58-7 | 53.2% | −1.3 |
| Team the favorite | 20-27-0 | 42.6% | −3.5 |
| Underdog of 7 or more | 31-25-2 | 55.4% | −1.2 |
| At home | 37-37-3 | 50.0% | −2.9 |
| Away | 45-46-4 | 49.5% | −1.3 |
| Line moved 4+ points against the team | 38-28-3 | 57.6% | −1.7 |
| Line moved under 1.5 points | 39-46-2 | 45.9% | −2.5 |
| 2016 to 2026 only | 54-52-4 | 50.9% | −1.8 |
One thing the table does show: the market already moves. Against the team’s line in its previous game, with home field removed, the number shifted a median of 1.5 points and an average of 2.3 points toward the opponent, and 5 points or more when a top quarterback went down. The size of that move told you nothing about the result. Games where the line barely moved covered 46%; games where it moved four points or more covered 58%; across all 178 the correlation between the move and the margin against the spread is about zero. The move is a price, and on this evidence it is about the right one.
The backup does not get found out in week two either. Where he kept the job, his next starts went:
| Start | Record | Cover rate |
|---|---|---|
| First start | 86-85-7 | 50.3% |
| Second start, if he kept the job | 52-61-0 | 46.0% |
| Third start | 39-40-0 | 49.4% |
On the moneyline: a small shortfall
The spread can be right while the moneyline is not, because the two are priced separately. Of the 178 games, 168 have a moneyline in the data (2006 on). With the bookmaker’s margin removed, those prices gave the backup’s team a 38.3% chance of winning. It won 31.5%: 53 wins where the prices expected about 64, a shortfall of just under two standard errors.
| Games (won–lost) | Won | Implied | Back | Fade |
|---|---|---|---|---|
| All 168 games (53-115) | 31.5% | 38.3% | −21.9% | +11.7% |
| Team the favorite (23-19) | 54.8% | 64.2% | −10.7% | +31.0% |
| Team an underdog (30-96) | 23.8% | 29.6% | −25.6% | +5.2% |
| Underdog of 7 or more (9-48) | 15.8% | 20.6% | −30.4% | +2.9% |
In dollars: $100 on the backup’s team in every one of the 168 games would have lost about $3,700, or $22 a game. $100 on the opponent every time would have made about $1,960, or $12 a game, two-thirds of it from the 42 games in which the backup’s team was favored and the opponent paid plus money. The underdog rows pay little because the opponent there is a favorite at about −200, and 76% of those favorites won against 70% implied.
Ten first starts you might remember
| Game | Line | Result | Covered |
|---|---|---|---|
| 2015 week 11: Brock Osweiler for Peyton Manning, Broncos v Bears | −2.5 | Won by 2 | No |
| 2017 week 7: Brett Hundley for Aaron Rodgers, Packers v Saints | +3.5 | Lost by 9 | No |
| 2017 week 15: Nick Foles for Carson Wentz, Eagles v Giants | −7.5 | Won by 5 | No |
| 2019 week 8: Matt Moore for Patrick Mahomes, Chiefs v Packers | +5 | Lost by 7 | No |
| 2021 week 8: Cooper Rush for Dak Prescott, Cowboys at Vikings | +4.5 | Won by 4 | Yes |
| 2022 week 14: Brock Purdy for Jimmy Garoppolo, 49ers v Buccaneers | −3.5 | Won by 28 | Yes |
| 2024 week 8: Jameis Winston for Deshaun Watson, Browns v Ravens | +7.5 | Won by 5 | Yes |
| 2024 week 10: Cooper Rush for Dak Prescott, Cowboys v Eagles | +7.5 | Lost by 28 | No |
| 2025 week 5: Cooper Rush for Lamar Jackson, Ravens v Texans | +2.5 | Lost by 34 | No |
| 2025 week 16: Gardner Minshew for Patrick Mahomes, Chiefs v Titans | −3.5 | Lost by 17 | No |
Three covers in ten reads like a case for fading, which is why the memory of these games is a bad guide. The whole set is 86-85-7; the famous games are the blowouts, and the blowouts are what people remember.
What to do with it
- Spread: the backup is in the number. Bet the game on its merits, and if you like the backup’s team, shop the line the usual way, because the number moves most at the books that reprice fastest. See why odds move.
- Moneyline: if you were going to back the backup’s team outright, the history says take the points instead. If you were going to fade, the moneyline has done a little better than the spread, mostly when the backup’s team is favored, and with error bars wider than the effect.
- The real question is the price, not the pattern. Compare the number you can get with the no-vig close; beating it is the only evidence that holds up over a season. See closing line value.
Data and method
- nflverse game data (Lee Sharpe, CC BY 4.0): scores, closing consensus spreads from 1999 and moneylines from 2006, and the starting quarterback of every game. nflverse weekly injury reports from 2009 and weekly rosters from 2002, for Out, Doubtful, reserve and inactive status. From 2002 to 2008 only reserve status is available, which contributes 18 injured-reserve cases.
- The team’s line is the points it was favored by, the margin against the spread is its winning margin minus that line, and the moneyline’s implied chance is the price with the vig removed across both sides. Records are win–loss–push; p-values are two-sided binomial tests against 50% with pushes dropped. The line move compares this game’s line with the team’s previous game’s, with 2.5 points removed for home field, and ignores the opponent.
- Limits: 178 games is a small sample, and every split has a p-value above 0.25. A quarterback hurt in the first three weeks, a team that alternated starters, and a game where the backup was hurt too and a third-stringer started are not counted. The scripts and the full tables are in the vigged repository under tools/backup-qb-study.
Frequently asked questions
Should you fade a team starting its backup quarterback?
Not against the spread. In 178 NFL games from 2002 to 2026 in which the starter was ruled out and his backup made his first start of the season, the team went 86-85-7 against the closing line, 50.3%, and lost by an average of 1.9 points more than the spread, which is inside normal noise. The backup’s second and third starts covered 46% and 49%.
How much does the spread move when a starting quarterback is out?
In these games the closing line moved a median of 1.5 points and an average of 2.3 points against the team compared with its previous game, with home field removed. For a top quarterback the move is 5 points or more. The size of the move had no relation to whether the team covered.
Do backup quarterbacks win less often than the moneyline says?
A little, in this sample. Teams starting a backup after their QB1 was ruled out won 31.5% of the games straight up while their moneylines implied 38.3% with the vig removed. Betting the opponent on every one of the 168 games with a price returned about 12% on the stake, but the shortfall is under two standard errors and a second build of the data found a smaller gap.
What counts as a backup quarterback’s first start in this study?
The outgoing starter had started the team’s previous three games, was the team’s week 1 starter or had started at least half its games, and was listed Out or Doubtful, on injured reserve, or on the injury report and inactive. The new starter had not started for that team that season. The final regular-season week, where healthy starters rest, is left out.
What data is the backup quarterback study based on?
nflverse’s game file (closing consensus spreads and moneylines, scores and starting quarterbacks), its weekly injury reports from 2009 and its weekly rosters from 2002, all published under CC BY 4.0. Spreads run from 1999, moneylines from 2006. The scripts and definitions are in the vigged repository.
Live odds boards: NFL odds · NCAAF odds · NBA odds · NCAAB odds · NHL odds · MLB odds · WNBA odds · MMA odds · Cricket odds · EPL odds · MLS odds · UCL odds · Liga MX odds · La Liga odds · Serie A odds · Bundesliga odds · Ligue 1 odds · ATP odds · WTA odds
21+ | Gambling problem? Call 1-800-GAMBLER. Nothing here is betting advice, and no bet or strategy guarantees a profit. Sports betting is legal only in some US states; check the rules where you are. Responsible gambling.