We now have 18 games worth of evidence on Drew Lock, NFL quarterback, so I thought it would be a worthwhile exercise to look at the first 18 games of every NFL QB since 2002 (which allows to bring in Drew Brees and covers everyone currently playing in the NFL except Tom Brady). That's 70 quarterbacks with at least 18 NFL starts.

The key question we'll examine is what did these QBs do after their first 18 starts? What sort of improvement can be expected and what sort of improvement would be extraordinary? For those QBs with extraordinary improvements, is there anything we can say about them which might provide an explanation? Finally, which of these QBs is Lock most similar to, and what does that mean for our expectations for Lock.

Since this will be an extraordinarily long post, I'm going to summarize the conclusions up here at the top and then give you all of the data and a bit more detail. Be warned: anyone who reads the whole thing will likely be able to tell you didn't if you only go off this summary.

• Drew Lock is a below average QB over his first 18 starts compared to the other 70. He's in the 38th percentile for completion percentage, 31% for yards per attempt, 49% for TDs, 42% for INTs, 46% for TD-INT ratio, 42% for fumbles, 38% for INTs+fumbles (not turnovers as I am just measuring fumbles, not fumbles lost, which depends more on the randomness of recovery), 45% for QB rating, 54% for rushing yards, and 62% for rushing TDs
• The average QB does not show what I would call "noticeable" improvement in any category after their first 18 starts. Most improve, but it is fairly marginal. The biggest improvement generally comes in TD/INT ratio. Specifically the average QB adds roughly 2% to their completion percentage, barely changes their yards/attempt at all, goes from 1.6 to 1.9 for TD/INT ratio, and from an 82.6 rating to a 86.3 rating.
• About 25% of QBs show real improvement and 10% show drastic improvement. That does not always mean Josh Allen stories (more on him in a bit), it can also mean Kyle Orton, Ryan Fitzpatrick, David Carr (David not Derek), and Joey Harrington
• Josh Allen and Drew Brees are the only two examples of someone who started out below average in their first 18 and turned into mega-stars. Allen improved the most of any of the 70 QBs in every single category. If you did not know this already, he is an extreme outlier
• If Drew Lock merely shows average improvement he will become a roughly 61% passer (would have been 32nd among QBs this year) with 6.7 yards per attempt (25th), a 1.7 TD/INT ratio (don't know for 2020 but it's the same as Dalton, Trubisky, and Mariota for their careers), and a rating of 83.3 (27th)
• If he shows 25th percentile improvement, Lock will be a 63% (30th) passer with 7 yards/attempt (23rd), a 2.0 TD/INT ratio (Tannehill, Rivers, Roethlisberger career mark), and a 89.6 rating (25th)
• If he shows 10th percentile improvement (Brees level), he'll be a 67% passer (12th) with 7.5 yards/attempt (13th), a 2.8 TD/INT ratio (roughly Watson/Prescott level), and a 96.4 rating (13th). If those ranks surprise you, the NFL is super passer friendly right now and what used to be great is now above average.
• Lock is most similar to Brock Osweiler and Jason Campbell, and a bit more broadly: Derek Carr, Andrew Luck, Christian Ponder, Matt Schaub, and Jameis Winston. This group showed above average levels of change, in particular Carr and Luck. If you widen a bit further, you bring in Josh Allen, Blake Bortles, Rex Grossman, Brian Leftwich, JP Losman, and Mark Sanchez. Still on the whole a group that improves, in fact one of the better clusters for improvement

OK, onto the meat.

First, the full table, QBs after 18 starts, along with averages at the bottom, sorted by QB rating. These begin with the first start the QB made, then continue in each game until they got 18. So some of them have more than 18 games in here (it's just the easiest way to pull the data and does not make a big difference). Note that we are not at all looking and when these starts occurred. You've got year 1 game 1 starters like Matt Ryan in with year 3/4 starters like Tony Romo.

A few comments before moving on: There are in fact several very good NFL QBs below Drew Lock in this list. There are a few QBs above him who regressed severely, generally running QBs who the league figured out. We'll discuss improvement/regression separately. But your main takeaway from this big list should be that for the most part, great QBs are great at the jump, good QBs are good at the jump, average QBs are average, and bad QBs are bad.

That leads nicely into the next table, which is guys who simply did not materially improve or regress after their first 18 starts (note that materially is not the same as none). In these next tables you will see what each QB (except for Lock of course who will not appear) has done over the rest of the career and how much they improved or didn't.

This is a full 35 QBs, exactly half the sample. There is a wide range in here of 1st rounders, mid-rounders, late-rounds, and undrafted guys. We have guys who were great and stayed that way, guys who were terrible and stayed that way, and guys that are stuck on mediocre stayed stuck. We remember Goff's terrible start, but by his 18th he had largely turned into the QB he is now. Some of these guys have pretty small "after 18" samples, so it's a bit early to say, but Darnold, Mayfield, and Jackson are basically who they were after their first 18. Not to say we don't see one area where somebody improves by quite a bit. Goff added 5 percentage points to his completion percentage. Rivers upped his Y/A by close to a yard. And several QBs improved their TD/INT ratio. But no one in this group became a totally different guy (at least not yet).

What about the group that showed significant, but not dramatic change?

Again I think an important thing to note here is that we still have a mix of improvement. Vince Young and Geno Smith are near the top of the improvement chart here and they still were bad. Really only Aaron Rodgers is at a star level (Minshew is a sample size thing to some degree, only a few games over 18 but this makes me think he is definitely worth a kick of the tires). Wentz as we know got up there and then regressed horribly, Luck was never as good as his reputation (though he was getting close in his last two years in the league), Stafford is better than his rep but still not a star, and the jury is still out on Murray (though early returns are good). Take a good look at Stafford, Luck, Keenum, Wentz, and Bradford. Those are examples of a very good, and statistically speaking unlikely, outcome for Drew Lock. In fact, given what we know outside of stats I would bet this is where Lock ends up. Somewhere between a Stafford and a Keenum. You'll also see by looking at this table that improvement is not an across the board thing in most cases. Eli Manning added a full 10 points to his completion percentage and also improved his yards/attempt (which is somewhat dependent on completion percentage) but his TD/INT ratio barely budged. Stafford, Luck, and Keenum also showed impressive jumps in completion percentage. Rodgers took an already impressive TD/INT ratio and made it ridiculous, but Wentz also greatly improved here. So when we think about Lock, he may very well improve "over" what we would expect in some areas and "under" in others. He may, for example, greatly improve his TD/INT ratio but not greatly improve his completion percentage (like Wentz did) or he may greatly improve his completion percentage but keep throwing those picks like Eli Manning.

Just a side-note before proceeding, note Roethlisberger on this list. That 93.5 ain't so impressive (it is barely better than Stafford).

Now, the big movers. Let's talk about Kyle Orton.

Yes, the great improvement group is not as simple as Josh Allen and Drew Brees. In fact, of the QBs on this list, they are the only ones considered superstars, though Kirk Cousins has a statistical case that is better than you probably thought. In fact, most the of the QBs who improved massively went from absolutely awful to just middling or good. Orton, Fitzpatrick, and Alex Smith are not anyone's idea of a franchise QB. But all of them improved more than Drew Brees did. David Carr and Joey Harrington are even on this list!

And I think we can cut *some* slack for people who though Siemian could be an NFL starter. DeMarcus Russell had a pretty small sample size post 18, which leaves Siemian as the holder of the true worst regression. RGIII we all know the league figured out.

Given all of this information I think it is fair to zero in on Josh Allen, Drew Brees, Kirk Cousins, Derek Carr, and Ryan Tannehill. Allen I discussed at length in this Fanpost, but just to bring up a few key points: he was a late-bloomer and not part of the QB factory through camps and such things, he went to a bad football school and had bad teammates, and he is perhaps the perfect physical specimen to play the position. And, he's A FREAKING OUTLIER. Like historically so. Is it technically possible for someone else to replicate what he's done? Sure, but it is very, very, unlikely.

Tannehill is a similar story. He was converted to a wide receiver at Texas A&M and then didn't play QB again in college until the end of his junior year. He's an athletic guy (more so than Lock) with a good arm and good size. He also quite frankly would not have made this list at all until last year. He had a hard time staying healthy in Miami. He found the right situation in Tennessee with an offense that fits his strengths and hides his weaknesses. If Lock turns into Tannehill, he's probably going to do it on a different team.

Cousins is dare I say fairly controversial. Did you know that this year he only had one game with a sub 90 passer rating? (now it was 15.9, but still) This is his stat line from 2020: 67.6% completion, 35 TDs, 13 INTs, 4265 yards, 8.3 yards/attempt, and rating of 105. QBR doesn't like him as much, 63.3, but his stats are very good. I think a lot of folks around here would not be happy if Lock turns into Cousins but I'm here to tell you that we'd probably be pretty lucky if he did. Cousins of course sat early, then wasn't very good when played. Here's the thing about Cousins: the 18 game threshold is very much why he is here. Cousins' 19th game was a perfect passer rating. His 20th was 89.2. His next six games were all over 100, including two more over 150. A wider threshold and he doesn't make this list. If we are still talking about massive improvement needed from Drew Lock at game 21, throw out the Kirk Cousins comparison.

Derek Carr is another one I don't think too many folks would be happy about but he's probably a near-best case outcome for Lock. Their 18 game passing stats are pretty similar, but the problem here is that they achieved them in very different ways. Carr was immobile and still is pretty much. He doesn't leave the pocket much. Lock does some of his best work on the move. But the real deal with Carr is that he needed the right coach. The last two years, under Gruden (who I don't like but I can't deny the results), Carr is a 69% passer with 7.9 yards/attempt, a 2.8 TD/INT ratio and a rating of 101.1. Basically the case for Lock here is that if we get him stability and the right offensive coordinator, he can blossom. Perhaps not into a superstar, but into a Derek Carr type.

That leaves us with Drew Brees who didn't really turn it around until year 3, then regressed slightly in year 4, his last in San Diego. Quite honestly, Brees didn't become the guy we think of today until year 4 in New Orleans, year 8 of his career. So another case of needing the right coach and the right system and needing to learn how he could translate his talents at the NFL level.

But, and this is a big but, there is something we haven't touched on yet at all. All of these guys played in college. Brees of course played in a different era (in fact I think we can to some degree chalk up Brees' massive improvement to the simple fact that the NFL became much more passer friendly). The leading passer in 2000 in college had a 64.6% completion percentage. Brees checked it at 60.4% in 17th place. His best season was an 8th place finish at 63.4%. Of course Brees played at a sub-par football school (similar to Lock).

Cousins had better stats, if not the good reputation which led him to be a 4th round pick. His best season included completing 66.9% of his passes, but with just 20 TDs to 10 INTs. By 2010, that completion percentage was only good enough for 10th nationwide.

Carr had a great college career, completing 66.7% of his passes of the course of it and reaching the 6th spot nationally in his last year. Tannehill struggled more, a career 62.5% passer who never got higher than 20th nationally.

All of this is a prelude to say that Drew Lock's college stats when it comes to completion percentage were not good. A career 56.9% passer, his best season was his senior one, when he topped 60% for the first time at 62.9% which got him to 36th nationally. This brings us back to Josh Allen, who famously had terrible stats for a big-time prospect, his high water mark was 56%, 77th nationally. So again, it appears that except for Allen and to a degree Tannehill, all of the QBs who showed massive improvement and became clear starters in the NFL had college resumes which indicated it was possible. And Allen and Tannehill were late-bloomers to the position, unlike Lock.

I'll wrap this up by looking at QBs whose first 18 games most closely resembled Lock's. This was accomplished using cluster analysis and I'll note that like with much of statistics, a lot of the final output here rested on my own determinations. I decided which variables to include, which meant excluding the rushing stats when the clusters kept insisting on putting Russell Wilson with Daniel Jones and Mitchell Trubisky (I did not normalize the numbers so the rushing stats had too much influence in the final result). I decided where to make the cut-off to determine cluster size. Just something to keep in mind whenever you read an article based on statistical analysis: what are the choices the author made and how did they affect the result?

Drew Lock is part of what I have labeled as Cluster 9. His cluster includes Brock Osweiler and Jason Campell (the two most similar to Lock), as well as Derek Carr, Andrew Luck, Christian Ponder, Matt Schaub, and Jameis Winston.

It is not an inspiring group to be sure, but that's what the first 18 games for Lock have been like. You can see from the data that this group showed above average improvement, though, with three QBs showing real improvement and one (Carr) drastic improvement. This trend holds if we bring in the cluster that is closest to this one, cluster 9, which includes Josh Allen, Blake Bortles, Rex Grossman, Byron Leftwich, JP Losman, and Mark Sanchez.

Allen is of course the headliner, but Bortles and Leftwich also showed real improvement. The thing is, again, improvement can be there but still leave us short of what we want. Cluster 9 was worse than Cluster 8, so Bortles and Leftwich ended up at a lower level than Schaub or Luck, but even if we get good improvement Lock may just be league average. I'd also like to point out the rushing stats, something I've largely not touched on. Josh Allen and to a lesser degree Andrew Luck provided a ton of value through their ability to run with the ball (as did Tannehill who is not in these clusters). Lock does not.

Before I conclude with the other clusters I suppose I ought to make a prediction. Quite honestly I found things I expected here (most QBs don't improve by much) and things I didn't (making the Carr/Stafford leap happens more often than I thought). From a pure eye test perspective Lock reminds me most of Blake Bortles, Ryan Tannehill, and Daniel Jones. I think he's likely to cut down a bit on the interceptions without raising that completion percentage too much so I think Drew Lock's most likely future is Ryan Fitzpatrick/Byron Leftwich/Cam Newton (without the running). It's possible he gets up to Stafford/Carr territory with the right coaching around him. It's also possible he ends up like Blake Bortles/Brock Osweiler/Jason Campbell and just is the guy we see right now.

Just for fun, here are the other clusters:

• Mahomes is by himself in cluster 1. No one has been close to as good as he was in his first 18 games.
• Cluster 2 is DeShaun Watson and Tony Romo. Really high yards/attempt and lots of TDs.
• Cluster 3 is Ben Roethlisberger and Chad Pennington. Really high completion percentage and yards per attempt. Interestingly both regressed about the same amount (cluster analysis did not use the regression data or the post-18 game data at all)
• Cluster 4 is Lamar Jackson, Dak Prescott, and Robert Griffin. This is without the rushing stats at all, they still ended up together. Low INTs, high but not crazy high yards/attempt.
• Cluster 5 is Russell Wilson, Jimmy Garappolo, Tyrod Taylor, Aaron Rodgers, and Colin Kaepernick. All had pretty good TD/INT ratios with high, but not crazy high TDs. Interestingly enough all had similar rushing stats except of course Garappolo and to a degree Rodgers.
• Cluster 6 is Joey Harrington all by himself. The mirror of Mahomes, no one else was quite as bad.
• Cluster 7 is Geno Smith, Vince Young, Kyle Boller, Alex Smith, David Carr, Ryan Fitzpatrick, and Kyle Orton. Other than Harrington this is the worst performing cluster.
• Cluster 10 is Chad Henne, Drew Brees, Sam Darnold, Josh Freeman, and Eli Manning. Low completion percentages and yards/attempt for this crew.
• Cluster 11 is Matt Ryan, Philip Rivers, Matt Cassel, Kyler Murray, Teddy Bridgewater, Cam Newton, Jared Goff, Trevor Siemian, and Andy Dalton. This group didn't throw a lot of TDs. Again, give the people who believed in Siemian a little slack.
• Cluster 12 is Jacoby Brissett, Trent Edwards, Jake Locker, Ryan Tannehill, Case Keenum, Colt McCoy, Sam Bradford, and JaMarcus Russell. Interestingly, this group is actually slighter better than cluster 8 when it comes to passer rating. Josh Allen aside, it's hard to argue that the end results aren't better too.
• Cluster 13 is Marc Bulger, Baker Mayfield, Carson Palmer, Kirk Cousins, Derek Anderson, and Matthew Stafford. Anderson and Stafford are interesting inclusions as their rating is much lower than the rest. It is entirely an artifact of completion percentage, every other number is very similar. Anderson of course never improved (actually he regressed badly) while Stafford did.
• Cluster 14 is Gardner Minshew, Marcus Mariota, Jay Cutler, Mitchell Trubisky, Daniel Jones, Joe Flacco, and Carson Wentz. An interesting collection, completion percentage in the low 60s and except for Minshew, TD/INT ratios of high 1s.