I have gotten some version of this question about a dozen times since July: is Drake Maye actually this good, or was 2025 a fluke I should be fading in drafts? So I pulled every play from 1999 through 2025 out of nflverse, built a QB EPA table of 983 qualified quarterback seasons, and went looking for the guys who have already run this experiment for us.

There are really two experiments worth running, and I'm going to do both, because they give different answers and both are useful.

1. What happens to quarterbacks who are great as rookies? This is the one everyone reaches for, and it's the scarier of the two.
2. What happens to quarterbacks who are great in year two? This is the bucket Maye is actually in. His 2024 rookie year was 27th of 40 qualified QBs, and the monster season came in year two.

The short version: Maye's 2025 was one of the 30 best QB seasons of the last 25 years, and almost nobody who does that does it again at the same level. But regression and collapse are two very different things, and which bucket a QB is in changes the odds a lot.

Quick disclaimer

This is historical base-rate work, not a projection of Drake Maye specifically. I'm asking "what happened to the other guys who did this," which is useful, but sample sizes get small fast; the tightest comp group below has 12 players in it. I also have zero visibility here into 2026 roster changes, coaching, or health. Take the percentages as a starting prior, not a forecast.

Definitions, so we are all on the same page

• EPA - Expected Points Added. Every play starts with an expected point value based on down, distance, field position, and time, and ends with a new one. The difference is the EPA. Positive is good.
• EPA/play - a QB's average EPA across his dropbacks (pass attempts, sacks, and scrambles) plus his designed runs. This is the number I care about most for QBs. Kneels and spikes are thrown out.
• CPOE - Completion Percentage Over Expected. nflverse models the completion probability of each throw from depth, location, and situation. CPOE is how much better than that expectation the QB actually completed. It's the closest thing we have to a clean accuracy stat.
• Qualified - 200+ dropbacks in a regular season. That gives me 983 qualified QB seasons since 1999.
• z-score - how many standard deviations above the qualified QBs in that same season a guy was. League EPA has drifted up over 25 years, so raw EPA/play across eras is apples to oranges. z fixes that.

What Maye actually did in 2025

Commentary

• 0.305 EPA/play is the 27th-best qualified season out of 983 since 1999. That's the 97th percentile. The company on that list is Brady 2007, Manning 2004, Rodgers 2011, Mahomes 2018. It isn't a "good year," it's a historic one.
• The CPOE number is the one that made me sit up. +10.8 is the highest CPOE of any qualified season in the entire 2006-2025 sample. The next best is Aaron Rodgers in 2011 at +8.8. Maye didn't just lead the league in accuracy over expectation, he beat the 20-year record by two full points.
• He got there while throwing deeper, not shorter. Air yards per attempt went 7.3 to 9.1 and his deep rate jumped to 14.2%. The cheap way to inflate completion percentage is to check down, and he did the opposite.
• He wasn't being carried by his skill guys either. His YAC per completion actually fell to 4.93, which is 37th percentile. The yards were coming from the throw.
• The one genuinely ugly number: an 8.7% sack rate, higher than 74% of qualified QBs. He held the ball. That's the part of his game that should scare you.
• Team context: New England went from 4-13 with a -128 point differential in 2024 to 14-3 at +170, won the AFC title game 10-7 in Denver, and lost the Super Bowl to Seattle 29-13.
• Fantasy-wise he finished at 20.6 points per game, QB2 on the season, dead even with Matthew Stafford and behind Josh Allen's 22.8, off 4,394 yards, 31 passing TDs, 8 picks, and 450 rushing yards.

Part 1: What happens to great rookie quarterbacks

Maye isn't in this bucket, but it's where every regression conversation starts, so let's actually look at it. I took all 81 qualified rookie seasons since 1999 and split them into tiers by how good the rookie year was.

Commentary

• Look at that last column top to bottom. Bad rookies improve 90% of the time. Good rookies improve 30% of the time. That is regression to the mean in its purest form. The rookie year tells you far more about the noise than about the player.
• Rookie QBs as a class get better. All 81 averaged a z of -0.72 as rookies and -0.11 the next year. The average rookie season is bad, and the average second season is roughly league average.
• The correlation between a rookie z and his year-two z is just 0.28. For year three it's 0.25. For reference, the year-over-year correlation for all qualified QBs regardless of experience is 0.42. So a rookie season is a worse-than-normal predictor of what comes next. It's the least informative season a quarterback will ever play.
• Only 69% of the worst-tier rookies even got 200 dropbacks again. That's the survivorship catch: the group looks like it improved partly because the guys who didn't improve got benched and dropped out of the sample.

Here are the twelve best rookie seasons individually, and what each guy did in year two:

Commentary

• Eight rookies since 1999 posted a z of 0.75 or better. As a group they went from +1.13 to +0.21 in year two, and exactly one of the eight improved: Roethlisberger, and he did it on 291 dropbacks.
• Four of those eight fell below league average outright. Dak 3rd to 22nd. RG3 3rd to 21st. Stroud 9th to 31st. Daniels 5th to 28th on only 243 dropbacks.
• The two survivors of the top group, Russell Wilson and Matt Ryan, both still declined, just gently. Nobody in this bucket held serve.
• Patriots fans already lived through the cleanest version of this. Mac Jones had the 15th-best rookie EPA in the league in 2021, and fell to 25th the very next year.
• If you drafted a hyped second-year QB off a huge rookie season in either of the last two years, you got burned twice in a row. That experience is exactly why people are nervous about Maye, and it's the wrong reference class for him.

The other side of the rookie coin

The flip is more fun and gets talked about way less. Here is every QB since 1999 whose rookie year was below average and who later posted a genuinely elite season (z >= 1.0) in year two or year three:

Commentary

• Maye's rookie season slots him second on this list, right between Lamar Jackson and Nick Foles. That's the range: a franchise-altering leap on one end, a one-year mirage on the other.
• Jared Goff's rookie year was a -3.41 z, which is the worst qualified rookie season in the sample, and he was a top-5 QB two years later. Rookie EPA is not destiny.
• Josh Allen and Tua both needed until year three. That's a reminder that the timing of the leap varies, and that a quiet year two doesn't close the book either.

Part 2: What happens to great year-two quarterbacks

This is Maye's actual bucket, and the picture changes.

First the class-level numbers. Of the 68 QBs who qualified in both their rookie and second seasons, 68% improved, going from an average z of -0.57 to -0.11. Then here's the number that reframes this whole post:

• Correlation between rookie z and year-two z: 0.28
• Correlation between year-two z and year-three z: 0.57

A quarterback's second season predicts his third about twice as well as his rookie season predicts his second, and better than the all-QB baseline of 0.42. By year two you're looking at a real player instead of a rookie surviving a script. That doesn't make Maye's 2025 repeatable. It makes it meaningful, which is a different claim.

Every elite year-two season since 1999

Here's every QB who posted a z of 1.20 or better in his second NFL season, and what he did in year three:

Commentary

• Twelve comps. As a group they went from a +1.81 z to +0.76 in year three. That's a big haircut and it's the honest expectation.
• But look where they landed. Only one of twelve fell below league average: Nick Foles, whose 2013 was the most obviously unsustainable season on the board. Compare that to four of eight in the elite-rookie group.
• 58% stayed at +0.5 or better, which is comfortably a top-10 quarterback. 33% held +1.2 or better: Warner, Mahomes, Manning, Garcia.
• Maye's 1.92 slots him fifth on this list, between Mahomes and Foles. I'm not going to pretend the historical record picks one of those for you.
• Side note worth flagging: only 65% of qualified year-two QBs qualify again in year three, versus 84% of rookies. Second-year guys have shorter leashes because the team has already made up its mind.

The biggest year-one to year-two leaps ever

Maye's jump was -0.011 to 0.305, a +0.316 improvement in EPA/play, the 6th-biggest rookie-to-year-two leap among the 68 qualified pairs since 1999. Here's the top of that list:

Commentary

• This is the bearish table, and I'm including it because it cuts against the story I'd rather tell. The 15 QBs with a leap of +0.150 or more averaged a year-two z of 0.55 and then a year-three z of -0.27. Only 13% held a z of 0.5 or better in year three.
• But read that carefully before you panic. That group's average landing spot in year two was only +0.55, and it's full of guys like Trubisky, David Carr, and Bryce Young who leapt from terrible to mediocre. A big leap from a low base is mostly the low base regressing.
• The size of the leap isn't what matters. Where you land is. Maye landed at 1.92 and led the league. That puts him in the twelve-QB table above, not in the leap-artist table, and those two groups have very different futures.
• The genuinely uncomfortable names here are Foles and Freeman, who both leapt into a top-5 season and then fell off a cliff. Two out of the top six. That's roughly the 1-in-3 downside I keep landing on with every cut of this data.

Which parts of a season actually stick

Regression isn't uniform. Some stats are skill, some are noise wearing a skill costume. I ran year-over-year correlations on every qualified QB pair from 2006 to 2025:

Commentary

• The stickiest things Maye did are accuracy, depth, and completion rate. Those are the ones I'd bet on showing up again.
• The two least sticky things in football are TD rate (r = 0.30) and INT rate (r = 0.21), and Maye was 87th percentile in one and threw the ball away from trouble in the other. That's the part I'd expect to give back. A 5.9% TD rate paired with a 1.48% INT rate is the exact combination that decays.
• EPA/play itself only carries an r of 0.44 year over year, an R² of 0.19. Nineteen percent. Four-fifths of a QB's EPA next season is not explained by this season. That single number is why I don't get too excited or too scared about any one QB year.
• On the CPOE record: I looked at the ten highest CPOE seasons since 2006 and all nine that had a follow-up season declined, from an average of +7.8 down to +5.0. Every one of them was still above average in EPA the next year. Falling off a record isn't the same as being bad. If Maye "collapses" to +6 CPOE he's still one of the most accurate passers in the league.

So what are the odds?

Two ways to answer it.

The base rates. Since 1999 there have been 64 seasons with a z of 1.5 or better. Of the 61 that had a qualified follow-up:

• 87% declined. Essentially everyone regresses. That's the ballgame.
• 30% repeated at z >= 1.5.
• 48% stayed in the top five.
• 15% fell below league average.
• The group went from +1.84 to +0.93.

Narrowing to the guys who actually led the league in EPA/play, like Maye did, 26 of them since 1999. 24% repeated as the outright leader. 48% stayed top-5. 64% stayed top-10.

The regression line. Fit next-year z on this-year z across all 667 qualified pairs and you get:

next_z = 0.100 + 0.417 × this_z     (residual SD = 0.83)

Plug in Maye's 1.92 and you get a projected z of +0.90, which at 2025's league mean and spread is about 0.177 EPA/play. That would have been 6th in the league last year. Around that projection, the residual spread gives:

• 24% chance he repeats at z >= 1.5 (another elite, top-2-or-3 season)
• 45% chance of z >= 1.0 (a comfortable top-5 season)
• 14% chance he falls below league average

The two methods agree almost exactly, which is the only reason I trust either one. Roughly a 1-in-4 shot at doing it again, a coin flip at staying elite-adjacent, and a 1-in-7 chance the whole thing was a mirage.

One more thing for the betting crowd: the Patriots, not the QB

Maye's individual regression is only half of it. New England improved by ten wins. I looked at every team since 1999 that improved its win rate by 8+ wins on a 17-game pace:

Commentary

• Ten such leaps in the sample. The average team went 4.1 wins, then 13.3, then 8.2. That's a -5.2 win average giveback.
• 90% declined. Only 20% stayed at 11+ wins. Half fell below .500.
• Washington in 2024-2025 is the freshest and most painful example, and it's the same profile: young QB breakout, huge win leap, hard fall.
• I want to be careful here. This is a team-level base rate picking up schedule regression, health, and one-score-game luck, not a judgment on this roster. But when I see a team win total priced off a 14-3 season, this table is what I'm thinking about.

Bonus: don't bury the 2025 rookie class yet

Since we just established that bad rookie years mean almost nothing, here's last year's class:

Per the tier table way up top, rookies in Ward's tier (z < -1.0) improve 90% of the time, up to an average z of -0.53 in year two. That's not "Cam Ward is going to be good." The average outcome there is still below average. It's "a bad rookie year is the single least predictive season in a quarterback's career, in both directions." Dart posting the best rookie EPA of the class at 18th overall is barely more informative.

Take Aways

1. Rookie year and year two are different animals. A rookie z predicts year two at r = 0.28. A year-two z predicts year three at r = 0.57. Maye's big season came in the one that actually carries signal, and that's the strongest argument in his favor.
2. The two comp groups land in very different places. Elite rookies fell to a +0.21 average with half of them dropping below league average. Elite year-two guys fell to +0.76 with one of twelve dropping below. Both regress; only one group craters.
3. Expect regression, don't expect a collapse. My honest projection is around 0.177 EPA/play, still a top-5 or top-6 quarterback. The most likely 2026 is "very good, not historic."
4. The accuracy is real, the touchdowns are not. CPOE and completion percentage are the stickiest things in the profile and he set the record in one of them. The 5.9% TD rate and 1.48% INT rate are the least sticky and both were career-best. That's where the points leak out.
5. In fantasy, I'm not paying his 2025 finish. He was QB2 at 20.6 points per game. If he's going as a top-3 QB I'd rather take the QB5-QB8 tier and the discount, because 24% is the historical hit rate on a repeat. If he slides to QB6 or later I'll take him everywhere. The floor in this comp group is genuinely high.
6. Watch the sack rate. 8.7% is higher than 74% of qualified QBs and it's one of the stickier stats on the board. If that comes down, my projection here is too harsh. If it goes up, the downside cases start looking more like Wentz and Herbert than Mahomes.
7. On the team side, I lean under a repeat. Not because the roster got worse, but because 9 of 10 comparable win leaps gave back an average of five wins.

I'll rerun this after a few weeks of 2026 games. CPOE stabilizes faster than EPA does, so his early accuracy numbers will tell us which branch of this we're on well before the win-loss record does.

Thanks for reading, and if you want to see our top player projections each week, create a free account by clicking this link.