Every quarterback since 1999, game by game

Most of the ways we talk about quarterbacks are season numbers. A season is a snapshot, and it hides the path: a quarterback who started hot and faded looks the same as one who started slow and figured it out. So we tracked every quarterback since 1999 one game at a time. After each game, each point shows how well he'd been playing, using only the games he'd played so far.

The QB trajectories page has all 190 quarterbacks with 16 or more games since 1999, and it updates every Tuesday. Here are four careers that show the range of shapes: Peyton Manning, good from his first game; Tom Brady and Josh Allen, who started near average and climbed; and Patrick Mahomes, who was at the top almost immediately.

The number is the Margin QB Grade. It counts the points a quarterback adds per play over every play he touches (dropbacks, sacks, scrambles, his own runs and pass penalties), with an interception a flat −4 and no play worth more than 4 points either way, on a scale where 50 is an average quarterback play from 2006 to 2021. 60 is clearly good. The best stretches in the data reach the high 60s.

How to read it

Nothing on the chart uses later games. Each dot is the estimate after that game, from the games before it. A single game's grade has a spread of about 25 points, so the estimate is a running average: games count by dropbacks, a game 32 games back counts half as much as the latest one, and every career starts with 218 plays of average built in. The line through the dots is built the same way. A smoothed line that also used later games would start bending toward a breakout before the breakout happened, which makes a career look more predictable than it was. It trails a real change in form by a few games, and that's the price of being honest about what anyone could have known. A version that does use later games is available as a toggle, labeled as hindsight.

Every line has a range. Fifty games of Brock Purdy and 333 games of Tom Brady shouldn't be drawn with the same confidence. The shaded band is where the estimate lands 80% of the time. We checked that rather than assuming it. A textbook noise model covered only 65% of what quarterbacks actually graded over their next 16 games, because players really do change, so the band is widened by a factor of 1.9 until it covers 80% across more than 10,000 checks. Fitted only on starts from 1999 to 2012, the same widening covered 77% of what came after.

The rest of the league is on the chart. A few famous careers can make late blooming look normal. Brady was below average after 32 career games. So was Drew Brees. Allen was barely above it. All three became great. So we counted. Of the 56 quarterbacks whose estimate sat at or below 50 after 32 games (careers that began in 2000 or later), 2 later reached 60: Brady and Brees. Twenty-six of the 56 played 32 or fewer games after that. Of the 37 who were already above 50 at game 32, 9 got to 60.

So when your team's second-year quarterback is average, the history says the most likely outcome is a quarterback who stays roughly average, or who is gone in two years. It isn't impossible. It happened twice among quarterbacks who started their careers since 2000.

Who is where right now

The second view is the one we wanted most. Instead of lining quarterbacks up at their first game, it lines them up at their most recent one and counts back 17, 34 or 51 games. The right edge is where everyone stands today.

Through Week 3 of 2026, Purdy leads at 65.8. Behind him, the next eight sit within four points of each other: Allen (62.0), Mahomes (60.4), Lamar Jackson (59.2), Jordan Love (58.9), Prescott (58.8), Joe Burrow (58.2), Drake Maye (58.1) and Jared Goff (58.1). Purdy has only 50 career games, so his 80% range is ±7.5 and runs from 58 to 73. That overlaps every quarterback in the top nine. The honest read is that Purdy has the best estimate in the league, the eight behind him are bunched, and the numbers can't order them.

A few other things stand out:

• Mahomes is still elite, but he's not the outlier he was. His estimate peaked at 68.9 in Week 12 of 2020, his 42nd career game, the highest point any quarterback has reached on the chart. It has drifted to 60.4. That's still third in the league.
• Drake Maye is the biggest riser. He was at 50.1 seventeen games ago and is at 58.1 now, eighth in the league. With 32 career games, his range is wide (±8.2), so this is a trend to watch. It isn't settled.
• Jared Goff is within a point of his career best. His highest estimate, 58.8, came in 2025, his 143rd game. He's at 58.1 now. Most quarterbacks peak much earlier.
• Jayden Daniels has come back to the pack. He was at 61.2 seventeen games ago and is at 54.4 now, with only 25 career games, so his range is one of the widest in the league (±8.5).

Quarterbacks peak earlier than you'd think

Goff is unusual. Among the 22 finished careers of 100 or more games that began in 2000 or later, the median quarterback's best estimate came at about game 74, a little over four seasons in. The middle half peaked between game 29 and game 125. After the peak most careers drift down slowly rather than falling off a cliff. Brady, who sat at 60 or better in 170 of the 171 games between his 100th and his 270th, is the exception.

What this can't tell you

The grade measures the whole play. The line, the receivers and the play-caller are all in it, and nothing here pulls them apart. When Goff went from the Rams to the Lions, this number can't say how much of the change was Goff.

We tested the obvious improvement, and it didn't earn its place. Adjusting every play for the pass defense it faced barely moves anything: across 983 quarterback seasons, the adjusted and unadjusted grades correlate at 0.99, and the median quarterback moves one spot in his season's ranking. Over a season, schedules mostly even out.

The chart also starts in 1999, so Manning's rookie year isn't in it, and his first game on the chart is his 1999 opener.

Pick any six quarterbacks on the QB trajectories page. The full table of every quarterback, sortable by peak, latest estimate or recent change, is there too.