What happens when the backup QB takes over

Through week 2 in the 2026 NFL season, we've seen Caleb Williams, Jayden Daniels, Jaxson Dart, and Sam Darnold all go down with injuries. Michael Penix was already out for the first two weeks recovering from his ACL surgery. And so we've seen a lot of backup QBs come in and fill in for these starting QBs. So naturally, I wanted to know what usually happens to the rest of the offense when a backup QB comes in.

The question I actually care about is a Giants one. I'm in multiple dynasty leagues and have multiple shares of Jaxson Dart, Malik Nabers, and Cam Skattebo through my various leagues. The 2025 Giants offense under Jaxson Dart was low volume (51% pass rate, 28.5 attempts a game). Jameis Winston came in for Dart mid-game and had a rough outing, but with a tough game script - who can blame him? So what is going to happen to the Giants' offense the rest of the season? What happens if the Giants acquire the perennial backup QB Joe Flacco?

Setting up the data

I pulled every regular season case from 2012 to 2025 where a team's week 1 starter played at least 3 games as the primary passer, then a backup took over for at least 2 clean games in a row (clean = one QB threw 85%+ of the team's passes). The game the starter got hurt in is thrown out. That gives 106 cases. The median window is 6 games before the switch and 3 after.

To keep ourselves honest, we need a baseline. A raw before/after picks up drift and regression to the mean which isn't a QB factor. For example, take the WR1 who led the team in targets in the before window - is it talent or a hot streak? After the switch his share drops 1.7pp. The baseline will tell us how likely this would've been to happen regardless of the QB change.

So to create the baseline, we run 486 "fake" QB changes. The QB was actually static all year, but we pretended there was a switch and see what the "drop" is in the WR1's target share. On the fake QB changes, WR1 loses about 1.1pp.

Net Δ = Raw Δ − Same-QB Δ
      = (−1.7pp) − (−1.1pp)
      = −0.6pp

So our takeaway here is that the WR1 loses about 0.6pp of target share on average when a backup QB comes in.

What actually changes

Not much, in the average case.

Teams keep the same pass/run mix. Volume holds. Efficiency slips a little across the board (fewer yards per attempt, worse completion percentage, fewer TDs), but none of those individually clears the noise. About 11 fewer passing yards a game is the honest headline for "average backup".

Target shares are even stickier than the volume:

The TE bump is right on the edge of significance and it comes out of the WRs. That's the safety blanket effect, and it's worth about a point of share.

By role, once you subtract regression, no WR1, WR2 or RB1 moves by more than a point of target share. The WR1 loses about 0.6 yards per target and 0.7 PPR points a game. The TE1 and TE2 gain a few tenths of a point. That's it. The fantasy hit from a backup comes from lower yards per target and fewer TDs, not from targets moving around. Your WR1 is still the WR1, he's just a slightly worse one.

So in the average case, the backup QB doesn't change the offense much. And if that's where you want to leave it, go ahead. However... there's a little more to the situation than meets the eye.

The backup's profile changes the answer

The average blends two opposite effects. Split by the backup's years of experience in the league and it separates out:

Veteran backup, 6+ years (n=35). Pass attempts +0.3/g. Passing yards basically unchanged (−4/g). Completion % holds. No receiver role shows a clear fantasy change at all.

Young backup, 0–2 years (n=42). Pass attempts −2.4/g. Passing yards −24/g, and this is the clearest effect in the whole study (CI −42 to −5). Completion −2.3pp, CPOE −1.9. TEs gain +1.5pp of targets, the TE2 specifically +2.0pp. The WR1 loses a full yard per target.

So when a rookie or second year guy comes in, the offense contracts. It runs more, throws shorter with less accuracy, and leans on the tight ends. When a veteran comes in, the offense mostly stays the offense.

Do backups come in types?

After Jameis Winston entered the game for Jaxson Dart, I remember at the Browns that Winston would sling the ball. He had a few big games, but also led the league in interceptions at one point. So I thought maybe backups like Winston and guys like Joe Flacco are probably statistically different from the standard backup if they get yanked around to different teams to cover for a missing starter.

To get started - I described each of the 106 stints by the backup's own line in the games he took over, standardized, and ran k-means clustering with k from 2 to 6. This basically asks a model to group the stints into k clusters based on SOME similarity if it can find one. Where a linear regression would try and fit a line to the data, k-means is trying to draw circles around groups of points if it can find them.

To check whether any grouping is real, I ran the same clustering on the data with every column shuffled independently. That keeps each stat's spread but destroys any real groups, so if the shuffled data clusters about as well as the real data, the "groups" are noise. Here the k=2 split barely beat the shuffled data (silhouette 0.15 vs 0.14), and higher k didn't beat it at all. So backups don't come in distinct types. What you get is one blurry axis, and k=2 cuts it in half:

Experience, pass attempts, fewer sacks and fewer QB runs all move together. Flacco (2020 NYJ, 2024 IND), Winston (2018 TB, 2024 CLE), Rivers, McCown, Dalton, Cooper Rush and Minshew land on the volume side. Keenum, Huntley, Browning and Lock land on the other side.

Does the type change where the ball goes? A little. The volume type gives the WR1 and WR2 about 1.2 more targets a game each (p = 0.011 and 0.026). But their individual shares barely differ, so the extra targets are just extra passes. WRs as a group gain about 3 points of share relative to the low-volume type, mostly from the TEs. That's 18 comparisons though, and nothing survives a multiple-comparison correction. Pass attempts were also a clustering input, so part of this is baked in.

The honest read: Flacco isn't a better or worse backup for anyone's share of the offense. He gives everyone more opportunities at lower efficiency. If you want a live example, it's Cooper Rush through his two 2026 starts in Atlanta. Here's Rush and Jack Strand going into Thursday night's Week 3 game against the Packers:

Plenty of throws, all of them short, and well below expected completion rate. For comparison, Penix came back that night and went 18/25 for 256 yards: 72% completion and 10.2 yards per attempt (with a 7.1 aDOT in Thursday night's game).

Flacco, stint by stint

Flacco is the purest case of the volume type and more extreme than its average. In every backup stint, team pass volume rose sharply and the top target's weekly usage went up with it. The cost is interceptions and a CPOE that is usually below average. These are 4–7 game samples and game script is not controlled for, so take the individual numbers as illustration, not proof.

The 2024 Colts one is the biggest style swing in the set. Richardson was throwing deep (12.2 aDOT) and completing 47% (he was also walking to the sideline to catch his breath). Flacco came in, aDOT dropped to 9.0, completion went to 66%, and every WR gained fantasy points. The 2023 Browns stint is the one that built the reputation: 41 attempts, 325 yards a game, and both Amari Cooper and David Njoku became league winners. Notice the TE bump isn't consistent though. Njoku got 9 targets a game in Cleveland. In Indy and Cincy, the TEs stayed pretty flat. There are a lot of inferences you could make from this, but we're not going there this time.

Winston is the other volume-first veteran. He pushes the ball deeper than Flacco and throws more picks. In 2024 Cleveland when he took over, the offense went from 33 to 43 attempts and 173 to 315 yards a game. Jerry Jeudy went from 8.8 to 21.0 PPR/g. The RB target share got cut in half (19% → 9%), a much more dramatic version of the average RB dip, and in line with his aDOT. In his two 2025 Giants starts (tiny sample), Wan'Dale Robinson hit 37% of targets and 19 PPR/g.

What this means for the Giants

The 2025 baseline under Dart was low volume. An average backup would leave that roughly unchanged. Flacco (hey it could happen) and Winston are not average backups. In every stint in this data, both raised pass attempts by 2–10 a game.

• Target shares won't reshuffle much. Across 106 changes, no WR1, WR2 or RB1 share moved more than about a point net. The Giants' top target stays the top target. With Flacco, expect that player's weekly target count to go up. Chase, Cooper and Downs all gained 2–4 targets a game.
• The TE is the likeliest share gainer, but mostly with the young/low-volume backup type, not the Flacco/Winston type. With Flacco it depended on the team. Isaiah Likely was already trending up - so expect that to continue. He'll probably be a league winner this year.
• The RB1's receiving role usually shrinks a bit. Volume QBs throw to backs less. Rushing work isn't in this report.
• Flacco vs Winston. Flacco completes more at a shorter aDOT. Winston throws deeper for more yards per attempt and more interceptions. For the pass catchers, both are a volume upgrade over a Dart-style run-heavy offense, and efficiency will likely land below a healthy starter's. Could we see a few big games from Malik Nabers?

What this means for your fantasy team

Pulling the above into the questions you actually have on waiver day:

• Don't drop the WR1. His share doesn't move. He'll lose a little efficiency (about 0.7 PPR/g on average), which is a downgrade, not a bench. If the backup is a veteran pocket passer, he might actually get more targets a week.
• The TE streamer is real but small. About a point of share league-wide, and it mostly happens when the backup is a young, low-volume guy. That's a "he's a fine bye-week fill-in" bump, not a "go spend FAAB" bump. Check which type of backup it is before you bid.
• Look at the backup's age before anything else. A rookie or second year backup is the one case where the whole passing game gets a haircut (−24 yards a game). Everyone on that offense goes down a tier. A 6+ year vet leaves the offense mostly intact.
• Volume vets are a buy for pass catchers. Flacco, Winston, Dalton, Minshew. When one of these comes in, the WR1 and WR2 each pick up about a target a game. Chase, Cooper, Jeudy and Downs all went up, not down. The catch is it comes with worse efficiency and more picks, so it's a floor raise more than a ceiling raise. Don't buy the team, but buy the game script and the fantasy opportunity.
• Sell the RB's receiving work under a volume vet. Volume QBs throw to backs less. If the RB1's value was leaning on targets, a Flacco or Winston takeover is a quiet downgrade for him even as the WRs go up.
• Don't chase the memorable stints. The 2023 Browns and 2024 Colts Flacco runs are the reason he has a reputation, and they're 5–7 game samples with no game script control. The median backup stint is 3 games. Plan for the average, and treat a Flacco-style blowup as upside, not the base case.

Caveats

Not controlled for: game script (backups tend to be on teams that are losing), opponent strength, and receivers' own injuries beyond only counting games they played. Pass rate is estimated from box scores (attempts + sacks), not play-by-play. Injuries and benchings are pooled because the injury report field in our data is too sparse to separate them. Uncertainty is from 2,000 bootstrap draws with case and baseline resampled independently. The veteran/young splits are n≈40 each, so they're noisier than the headline numbers.

Quick Note

I used AI to assemble the data and build the report. The writing was done by me with some AI review and editing.