It's interception rate, INT%, scaled to a deviation from the league's average. 100 is exactly league average. Anything below 100 is worse than league average, meanwhile anything above 100 is better.
I could've just used INT rate for this, but it's much easier to understand when I say Justin Herbert being 32 percent better than league average (132 INT%+) at avoiding interceptions on just a 1.9 CPOE is unsustainable, because these are both stats scaled to be deviations from NFL average. I could've also said that a 0.6 percent INT rate is unsustainable on accuracy this close to league average. It would've been just as truthful, but people tend to understand comparisons to league average better.
So yes, INT%+ is just interception rate, expressed as the percent deviation of how much better or worse a player is than league average. 132 INT%+ means 32 percent better than league average, meanwhile 100 - 32 = 68 INT%+ means 32 percent worse than league average.
I think Tiers 2 and 3 are the most interesting and potentially controversial tiers on the list. It is where you really see the difference between actual performance and popular narratives. For example, I have heard 3 or 4 different people in the media predict that this is the last year Kyle Murray plays for the Cardinals. While that could be true if he has a bad season, based on this year it would be insane to do so. As we have spoken about before, he provided above league average value in every meaningful area (Accuracy, sack avoidance, turnovers, rushing). As long as he can stay healthy he should be a good QB for at least a couple more years depending on what is mobility looks like in his 30s
Another one is Sam Darnold. I'm not sure what the consensus is but I heard a fairly data driven guy call him a bad QB. I know they can be large discrepancies between skill and results but I don't think you can be the 7th most valuable results in a season and be bad.
This leads me to two questions. The first is about your ranking system. I understand the reasoning behind it but am curious about the application. Do you use a formula? If so what weight do you put on different inputs (I'm guessing they are the ones you put in the stats column) and how has that different pre CPOE era and pre EPA era?
The second question is who is the worst QB to be on the best team in the league (in terms of your Exp Wins model)? I know you and David have talked about the worst QBs to make and win SBs but I seems to be harder to be the best team in the league with a below average QB then to make the SB with one.
First, yes this list is ordered by a mathematical formula. It's actually a public opinion list, fitting the players' results to composites combining every best QBs in the NFL list I could find on the internet in the last 20 years or so. It's not objectively a measure of anything. It's taking the values that fans have implicitly placed on different factors over this period of time, and applying them the same way to every player, regardless of narrative. This is what results in the following order of importance:
1) Total touches. To be fair, the internet has kept their word on this. By far the most important factor in this equation is total touches, because the best ability is availability, and placements on tier lists around the internet tend to back up that tough talk. It's also the case that volume tempers the importance of rate stats, which is why it's included as its own separate factor.
2) ANY/A. The most public facing advanced QB stat. It makes sense this would be valued second.
3) Total EPA. Mainly for the purpose of the Josh McCown test. How many plays of elite value does it take before a player is actually high on the list? It did shock me though how the public values total EPA more than EPA/Play.
4) EPA/Play
5) An extremely small adjustment for skill.
These are the only five inputs. Everything else is approximated too well by this combination of five, and just gets in the way if included individually. That's why I didn't feel the need to include TD passes, yards, or anything like that. With a combination of total touches, and the foremost rate numbers, you get a clearer view at who's good at moving the ball without any of the other stuff.
The thing that jarred me most when I was making this formula several years ago is just how little the public cares about skill. It's not nothing, but it's little. In general, ratings in the public eye are done based on the real results, once it's taken into account how many touches a player has. That's why this list ends up so real results dependent, because this is just how the public values players. As such, when I replaced the placeholder values for CPOE that I had going back to 1970 with the fitted ones I made just recently, very few rankings throughout history changed. Maybe a few, but not very many.
Moving into years older than 1999, things have to be done on the fly, but for most of them, the same fitting is done with football outsiders DYAR, DVOA, total touches, and ANY/A, with the same extremely small adjustment for skill. DVOA is not a replacement for EPA, but it my formula, it works well at serving the same purpose.
Moving onto your final question, it's actually the obvious answer. I have not checked this thoroughly, but I believe Joe Flacco in 2008 is the only rookie ever to play QB on the best team in the NFL by my Expected Win formulation.
With Joe's 0.015 EPA/Play, -1 CPOE, 93 sk%+, 5.29 ANY/A on a significantly deleveraged 517 total touches, he finishes just 22nd on my 2008 tier list, the same spot CJ Stroud occupies on this one, which is the worst QB the best team in the NFL has ever had, in my opinion. In terms of xEPA, I believe there are worse, but I haven't cross checked that. In terms of real results, there has been no worse QB on the best team in the league than 2008 Joe Flacco.
I also think tiers two and three are the most juicy every year, as it's where both the underrated and overrated guys end up. I myself said last year that 2024 stood a significant chance of being Kyler's final season with the Cardinals, if he could not turn it around from what had been consecutive seasons finishing 26th and 27th. To Kyler's credit, he did turn that around. He's right back to being what he was before the injuries, either tenth or 11th.
I went over this on my QB contracts note a little while back. Being stuck in tenth or 11th means Kyler is being overpaid. You can get more for less money easily, but does it mean he's a bad player? In no way shape or form. If what you're looking for is a chance to win football games, Kyler is difficult to upgrade from. They were already the best Expected Win team in the NFC West in 2024, although that's not big praise. If the Cardinals are looking to do a Giants impression (i.e. find ways to lose football games), getting rid of Kyler Murray will be a good way to do that.
I have a hard time believing anybody called Sam Darnold a bad QB in 2024. I discussed how the Kevin O'Connell offence has never had a QB significantly overperform his xEPA/Play in it, no matter how much people pretend it has, so in my opinion, you have to try hard to get to the result that Sam Darnold was not very good in 2024. It requires a cocktail of being enamoured with the O'Connell offence, understanding that sacks are a QB stat, but somehow failing to understand that accuracy is the stat that a QB has the second most individual control over, and also falling victim to the stigma that QBs who change teams are somehow lesser than those who don't, or as I like to call it, the Andrew Luck/Jeff George fallacy.
There are reasons to be afraid when it comes to Sam. A lot of them are predicated around the areas on the field in which he is accurate, and the areas of the field where he is not, but in his accuracy profile, he looks very similar to Josh Allen (accurate short, accurate long, not very accurate in the middle, basically the opposite of Lamar Jackson). Josh's profile has not sent him off the cliff, so I don't believe it will happen to Sam either, although I do find it difficult to forecast him back in the top ten next year, for Seahawks reasons. This does not mean he's a bad player though.
That’s really interesting that this is based on an objective formula—albeit one trying to predict more subjective rankings—rather than your own subjective weighting (which is how I read the framing of it). Obviously, I knew you would put a premium on stats, but thought it might be a bit more ad hoc when it came to breaking ties, etc. How does the grouping into tiers come about once you have the results from the formula?
One thing that stood out to me (before I knew it was calculated by a formula) was how true you were to the principle that volume matters by ranking Russell Wilson ahead of Justin Fields, despite your position that switching was a mistake (even if not necessarily a huge one). Even though you’d have preferred Fields over the rest of the season, the real result of the change took that out of your hands and so Wilson ends up a few spots higher than Fields despite Fields being otherwise “available.”
Purdy/Mahomes’s positioning also illustrates that to some extent. The CPOE & sk% aren’t that different, but Purdy’s EPA/Play and ANY/A are noticeably better than Mahomes. Under the guise of “a full season of play/how many plays do I have to replace this QB for,” I don’t think I’d necessarily take Mahomes over Purdy. Purdy’s total touches represent basically a full season within the context of the SF offense, so it isn’t as though I need to replace many of his touches with replacement level play. But since the metric doesn’t actually account for that, the real difference in volume makes up for the poorer results.
I’d also be interested (since the formula is derived from public opinion) in who the formula doesn’t get right. We’ve talked several times about Matthew Stanford, and the difference between his public perception and his quantifiable skill/results. I imagine the formula may have consistently missed on his ranking and predicted him lower than the actual accumulated rankings of the internet. Were there any quarterbacks the formula consistently overvalued relative to the public perception it was designed to predict?
Tiers come about when there is a clear jump in terms of the score. The formula gives each player a score that's mapped to a ranking list (1 means good, 32 means bad), and tiers are made whenever a substantial difference between two people is found in the rankings, to mark that we've moved into a different class of player.
For instance, to break tier six into tier five, the break point is obvious, because here are the scores:
25) Bryce Young - 23.49
24) Drake Maye - 23.46
23) Russell Wilson - 20.98
22) CJ Stroud - 20 flat
You can see the break point there. Despite being only one position apart, Russell Wilson is a fundamentally different class of player than Drake Maye, which is why the tiers come into it. There doesn't have to be tiers. I could mention this in the course of the write up, but the tiers are a natural way that people understand to delineate the different classes of players. It's actually just a crafty way to save myself some words, and not really anything more. As if there weren't already enough words in this list.
Nothing is ad hoc with me LOL. All my lists are mathematical formulas. Either subjective ones that I designed myself, like for the one offs and the upsets, or more objective ones like this, fitted to a very specific parameter, which I believe to be the most objective way to do it, due to reasons I just explained in a reply to Marc. I'll refer you to that, so I don't type them again here.
The Justin Fields/Russell Wilson example is exactly right. That's why this is the results-based tier list. Do I think Russell Wilson in 2024 was a better player than Justin Fields? No, but were his 2024 results better? Yes they were, because he touched the ball more than 150 times more. Volume is the key to all this, because high volume makes putting up good rate stats more difficult. Rate stats mean nothing without volume, which is why volume is the most important factor in the fitted formula. It would be the most important factor that myself or anybody would subjectively design too, and my position is that you either accrued the volume or you didn't.
We can't make special exemptions for guys that were benched and not injured like Justin Fields or benched for non-football reasons like Daniel Jones or were just injured like Tua Tagovailoa or whatever may have happened. You either played, or you did not. We're ruthlessly objective here, and we can get away with that, because the interpretation of the list is, looking back at 2024 with the benefit of hindsight, who would we have picked? This means we know Justin Fields is getting benched after 251 plays and not a single one more, and we can make our choice accordingly.
This carries over into the Mahomes/Purdy example. It's not strictly a matter of how many plays you need to replace a guy for. The assumption of the model is that if a QB touches the ball more total times, each touch is more difficult on average than QBs who touched the ball fewer times, because it's more likely a defence would be counting on them to touch it. This is not always true by the way. For instance, if a rush offence is horrendous, taking touches away from it might actually make a QB's life easier, despite the defence keying in on him more, due to less bad (often second and long) situations, but every model has to make an assumption somewhere. You cannot account for everything.
You've actually found the key assumption underlying my whole process here David. I'm impressed. High five. My position is that football is not baseball. In this sport, there is no such thing as a stat accumulator. Each time a QB touches the ball makes his team more and more pass heavy. As a QB's team gets more and more pass heavy, the opposing defence (think of it as one defence for the purpose of interpretation) starts to sell out more and more to stop the pass. As they do this, it gets more and more difficult for the QB to put up good rate numbers through the air.
The boost is not just a boost for real volume. It's a bonus intended to make up for the fact that the difference between Patrick Mahomes' 695 plays and Brock Purdy's 570 likely also means Patrick's average play was more difficult than Brock's. Patrick Mahomes put up 0.165 EPA/Play on 695 touches, compared to Brock's 0.198 on 570. In volume terms, this is basically equal. 114 for Patrick. 113 for Brock. The difference maker is the belief that Patrick's touches were fundamentally more difficult than Brock's, because the defence is more likely to be selling out to stop the pass against Patrick.
All of this is just interpretation. It's easy to disagree with, but it's the methodology that built this list. This is what I meant when I said my methodology questions why the 49ers did not throw more passes. The unasked but implicitly defined question was 'can Brock Purdy handle the per play difficulty that the guys in front of him, each with more total touches, had to deal with?'
We don't know whether he can or not, because this cycles back to what I was talking about above. In 2024, he didn't. Patrick did. That's why Patrick goes ahead. That's why Jordan Love is so low, and it's why Joe Burrow is so high. It's also why Caleb Williams is so high. Having to drop back to pass on every play makes each one harder, so he can get away with worse results than CJ Stroud can get away with. That would be the interpretation of my methodology anyways. That's why he's above him on this list.
I dislike your connotation that this list is trying to predict public opinion. It's not. It's fitted using the public's opinion, to implicitly place values on different aspects of play. If I were trying to predict the public's opinion, I would need career Super Bowl championships and etcetera in the formula somewhere. This formula strictly strives to weigh every aspect of play the same for everybody, but since I do have the rankings on my list, and several public opinion polls, I can answer your questions anyway.
Matthew Stafford's ranking on my list did not significantly differ until he won his championship. Through basically all of the 2010s, my tier lists have him in the mid teens, and public opinion also has him in the mid teens. I have Matthew fourth on my 2011 tier list, a season that is underrated even to this day, and that's coming from Matthew Stafford's biggest hater, but that is the only significant deviation, until we reach 2021, whereupon people seem to think Matthew Stafford is a top ten guy for some reason, while his results have not budged from the mid teens.
In terms of types of players that are over/underrated by the public relative to their actual results, the general formula is that turnover avoidance is overrated, and sack avoidance is underrated by the public. These are both true in separate, but super effective when combined. This means that guys who are great turnover avoiders but not good sack avoiders (CJ Stroud, Jalen Hurts) tend to wildly overshoot in the public's eye where they would be located on any objective, results based measure like mine, and vice versa.
Guys who are great sack avoiders, but have turnover troubles, often end up really underrated on a list like this. There are few examples of this in 2024. Bo Nix is probably the best, but that's likely more just his sack avoidance being underrated. Josh Allen also used to fall into this in the past.
Stick to the TEP principles. A turnover is 2.4 times as bad as a sack, and that's all. Do not mix up the values of these things. This is the area of QB analysis the public seems to have the biggest trouble with right now. Fans in general have gotten pretty good at weeding through everything else.
Thanks for the clarification! I apologize for my poor wording that the model is intended to predict the aggregate public opinion, rather than just being fitted to the public opinion data. I get your point: the end goal isn’t to match the public opinion data, that’s just the method you used to fix the weights. In that sense, it fits the public opinion data as best it can via the stats you selected, but if you were truly trying to predict the public opinion, you would have included other variables.
What was the reasoning behind crafting the formula so that if fitted public opinion (or more accurately public values) rather than a more objective approach? For instances, you and the analytics community at large value EPA more than ANY/A yet it is more important in your formula.
Do you think there is a greater divides amongst player perception know than before? The reason I say is because the more information we have access to, the more divergent opinions can occur. In the past, you maybe only knew a guy's basic stats and how much his team won. Now, you can watch more of a player's game than ever before and have way more and better data and metrics.
Take Brock Purdy in 2023. If you look at EPA/play, CPOE, and ANYA you would be inclined to say he should have won the MVP. However, if you are an "eye test" guy, you wouldn't have him in the top 10 because he doesn't make a lot of wow throws, plays in a Shanahan offense, doesn't have pretty throwing motion, and is slow. You could also be an (eye test) guy and say he doesn't a great job of running the offense and has great anticipation. If you look at volume numbers (both traditional and advanced) and win %, he would be in the top 5 but probably not a your pick.
I'm not really sure what my point is in all of this, but is interesting how one could come to and argue radically different opinions on a player placed on what you want to believe, who you listen to, and what you care about in a QB.
Have you ever used something like rateyourmusic Marc? A website where people vote on how good something is? This is almost exactly that, but with me stealing hundreds of rankings from all over the place and collecting them all together. As for the contrast between this and an objective alternative, I asked myself this question years ago.
The question and issue becomes, what is a more objective alternative than an averaged out ranking from all kinds of different people? Is it for me to decide which stats are important and which are not? That doesn't make any sense. Is it for me to fit this individual ranking to team stats, like points scored or winning? That doesn't work for me either. To me, the way I have done it is the very most objective way to rank players, without bending to the whims of teams' stats, and without slathering my own biases (I am a bigger fan of small sample greatness than most, we all know this) all over everything.
In the end, the fitting mechanism is only a way to assign weights to things, at which point it becomes a level playing field. Everybody is playing under the same rules. In that way, it's as objective as can be.
It's also not the case that ANY/A is weighted heavier than EPA. I should've explained that better. While ANY/A is worth more on a rate basis, EPA/Play obviously contributes to total EPA, and when you add the contributions of (3) and (4) together, it outweighs number two. It's only as individual factors that ANY/A outweighs them both.
In general, I do not think divergences between my list and the public opinion are getting bigger, as it's always been kind of heavy on splitting the difference. The divergences within public opinion are another matter, because there are different types of divergences. Some have existed forever. Some are new. For instance, the divergence between those of us who think Terry Bradshaw was a league average QB for most of his career and those who think he was a top five guy of the 70s because of all the winning he did were happening even in the midst of the 70s. The modern allegory is obviously Patrick Mahomes, and the disagreement that goes on about his positioning in the hierarchy is a tale as old as the hills in the NFL. That disagreement is no bigger or smaller than it used to be.
When it comes to somebody like Brock Purdy, I think you're right. The disagreement is bigger now because people have the means to disagree now. The obvious case is Jim Everett. I have him top four in my tier list four times, including three in a row, and including a number one position, but in real life, he made one Pro Bowl, which was not in the season I rank him first. It's because all of the credit in his day was stolen by Zampese, not entirely different to what Brock is going through now with Shanahan, and the stats that really love Jim Everett did not exist in 1988. It was all about passer rating and, if you were an advanced metric guy of the era, ANY/A, all of which WAY overdiscriminate against turnovers, meaning all it took was Jim's career INT%+ being only 100 to sink him, because the stats of his day did not see him as being too good to deny in the way the stats of today do.
If Brock Purdy were playing in 1988, he and Jim Everett would swap places. I would be angling that Brock Purdy was a Hall of Fame guy, with the benefit of modern metrics, but modern fans wouldn't have even heard of him. They would have heard the narrative that he was in a great offence, like they've heard about Jim Everett and Trent Green, and leave it at that.
More metrics being available today does create more grounds for disagreement, and naturally, more disagreement. I think this is what you were getting at, and I think you're right.
This is a great list, and there's so much that could be said here. Obviously there's no debate about the top spot. The only knock on Jackson is that he hasn't made it to the Super Bowl yet. And I'd say it's certain he will, except that the the next two QB's on the list also play in the AFC. I honestly can't say which of those two I'd rather have on my team, though if pressed I'd probably choose Allen.
It seems odd that #4 on the list is in the tier of quarterbacks who can carry a team to the Super Bowl but also is described as a Game Manager. Maybe it's not a contradiction, but it seems like it should be.
Speaking of game managers, I wonder how much of Patrick Mahomes apparent decline is really Travis Kelce declining? What kept Mahomes from being just a game manager in previous years was his ability to pull a rabbit out of a hat and keep drives going even when you thought he was beat, and most of the times that happened it was Kelce who made the catch. I'm certainly not going to complain if Mahomes has lost that magic touch, but it'll take more than a down season or two for me to feel comfortable with rating him as merely above average.
Moving on to Bo Nix, there's plenty to say. His season overall was good, obviously, but what I was happiest about was his growth over the season. His first few games were pretty rough, with just enough glimmers of promise to keep it going. Lots of inaccurate throws. But he improved. A lot. By the end of the regular season he was the best rookie in terms of completions, yards, touchdown passes, and total touchdowns, and he did that with a weak rushing offense, mediocre (at best) tight ends, and only one elite receiver. Yes, I do think Sutton can be elite. He's capable of making some unbelievable catches, and seems to have a good rapport with Nix. Given the improvements Denver has made in the off-season this year, I think both will have a great season. I'm confident he'll be solidly in tier 3 at worst, and most likely in tier 2.
There's almost no way Lamar will fail to make the Super Bowl at some point. By xEPA/Play, he's getting close to being the best QB of his generation as he rises and Patrick falls. There's never been a QB who was the best (or even second best) of their time that never made it, except for Dan Fouts. Ken Anderson took ages, but he made it, and he's the only one that even came close to being shut out for a whole career that isn't Dan Fouts.
Take Peyton Manning as the example. We can all admit that he was not a playoff performer, but through sheer talent, even far below his best, he managed to make the Super Bowl four times. Is Lamar Jackson at the talent level of Peyton Manning? Generally not, but in 2024 yes. Like I said, if he can keep this 117 sk%+ going, he will make a Super Bowl imminently, almost certainly multiple. If he goes back to the old no pocket presence Lamar, it's debatable whether he's even a Hall of Fame guy. This is what puts me on edge about Lamar Jackson. In his two seasons with a sack rate less than six percent, he was number one (but no tier zero) in 2019, and tier zero this time around. The problem is there's only two seasons with a sack rate less than six percent, and he's only got three career top ten finishes in what is now seven qualifying seasons. Not quite Dak Prescott, but a very inconsistent player.
The game manager thing comes as a result in a slight difference of opinion on what exactly the term 'game manager' means. To me, game manager is a play style, consisting of short throws, easy throws, EPA/Play values that underperform very high success rates, and in the modern day, NFL NGS CPOE hating you, as a result of only throwing to wide open targets, with the addendum that running QBs who fit all the other characteristics (Jayden Daniels, Kyler Murray) get a pass. Some people take the term game manager to be a reflection of the quality of play. I do not.
Taking the above to be the meaning of the term game manager, there have been plenty of game managers in the upper echelons in plenty of different years, including a tier zero season that I feel absolutely fits in the game manager parameters (Patrick Mahomes 2022). I don't think the terms 'game manager' and 'elite QB' are mutually exclusive. You can be both at one time. It's tough to pull off, which is why it doesn't happen often, but it can be done.
As far as Patrick and Travis Kelce go, it's very possible that what you're saying is correct. I did say in the article that even a single top 40 receiver would help Patrick out a lot, but I think 2023 raises a mild argument to this. In 2023, Travis had declined, but not to the point of no longer being a top 40 receiver, like he declined this year, and there was also Rashee Rice. Nevertheless, Patrick ranked tenth in EPA/Play last year, and tenth in EPA/Play this year (I use ranks because Patrick's raw numbers rose with the tide, just like everybody's did).
I do want to make clear my position that Patrick Mahomes was likely playing over his head the whole time, which would be an agreement with you. This is not an assertion to make lightly, but it must also be said that even Alex Smith put up a top three results season with Tyreek Hill, Travis Kelce, and this Andy Reid offence. Now that Tyreek is gone, and Travis (as an elite player) is gone, we're going to see what Patrick Mahomes is really made of, and so far, it's not been a great showing.
Patrick is an odd football player. His combination of fantastic physical talents with a game manager's playstyle is not something the league has ever really seen before. In my opinion, this is why a lot of the advanced metrics don't like him. Even my baby, xEPA/Play, has Patrick as the fifth best QB of all time, but even that is an underestimation compared to his results, which are better than fifth as of right now.
Every advanced metric out there thought Patrick was due for a come down. The issue was only how far. There are ones like NFL NGS CPOE, who think Patrick just sucks, and can't throw the ball accurately at all, but there are also others like my xEPA/Play, which thinks that while top one may be out of reach for Patrick at this point, top five is still possible. I think this is much more reasonable, although with a sack rate in the high five percent bracket, I don't see Patrick Mahomes as an elite QB. I'm not changing my stance on that. Merely predicting that he won't end up this bad at sack avoidance in a full season again.
I still think I put Patrick pretty high. It took everything in the world going right (multiple different kinds of unsustainable luck) for Justin Herbert (a player who some people think is a top five guy, although I vehemently disagree) to sneak ahead of Patrick into eighth place, a miracle that will not happen twice, but even if it is eighth place, I do think Patrick has lost a little bit of the magic. It's not like 2022 Tom Brady, but just for the record, 2022 Tom Brady finished ninth on that tier list...
Onto Bo Nix, we are almost entirely in agreement. Bo is a lot like Bryce Young, in that a few bad games at the beginning of the year weigh his season-long numbers down a bunch. In the same seven game stretch in the second half of the season where I discussed Jayden Daniels going through the falloff (a falloff that is quite customary in the second half of a rookie season, by the way), Bo Nix was doing the exact opposite, generating 0.255 EPA/Play in the back half of the year, sixth in this span, behind only the ironclad top four, and the late season surge of Baker Mayfield. Name a QB outside the top five, and I'll tell you that Bo Nix was better than them, in the second half of the season.
This list looks backwards, so the weak first four games that almost dropped a great Denver team out of the playoffs altogether have to count against him, but if it were looking forwards, picking guys for 2025, Bo would be higher than 14th. There's no question there. In the final 12 weeks, his CPOE was a much better looking (especially considering the arm isn't the appeal) 2.1. If you substitute this 2.1 in place of Bo's 0.6 he got in real life, all of a sudden Bo Nix is the league's seventh placed xEPA/Play finisher, and one of the guys above him just retired.
It's a stretch to say he can be a top five guy next year, because we will see about the 121 sk%+. He only has to repeat this level of sack avoidance one time to convince me it's for real, but if he can repeat it, than a 2.1 CPOE will be good enough to put him into the top five, skill-wise. He played like that in the final 12 games of the 2024 season. We will see about the real results, but good news for a Broncos fan is that Bo is one of the very most underrated players by his real results in 2024, and his real results were still enough to get him to 14th place as a rookie. That is mightily encouraging.
Absolutely incredible read, I don't know how you manage to write this much but it's an absolute pleasure to read. I've always been more of a stat guy over a results guy, so these kinds of reads feel great to read. Amazing job Robbie, as always :)
Like I said, this was a mammoth project. If anybody was wondering about my radio silence over on Notes for most of the last week, this list is why. It took everything I had to get this list to an acceptable standard, as they require an even higher level of quality to overcome the 'lazy list' stigma, but I think I managed to pull it off. Thank you very much again for your kind words. There is not enough objective QB analysis out there, and I sincerely hope that I'm able to provide some.
this article was clearly well researched and well-written, but can you explain what INT%+ means? I googled it and couldn’t find anything
It's interception rate, INT%, scaled to a deviation from the league's average. 100 is exactly league average. Anything below 100 is worse than league average, meanwhile anything above 100 is better.
I could've just used INT rate for this, but it's much easier to understand when I say Justin Herbert being 32 percent better than league average (132 INT%+) at avoiding interceptions on just a 1.9 CPOE is unsustainable, because these are both stats scaled to be deviations from NFL average. I could've also said that a 0.6 percent INT rate is unsustainable on accuracy this close to league average. It would've been just as truthful, but people tend to understand comparisons to league average better.
So yes, INT%+ is just interception rate, expressed as the percent deviation of how much better or worse a player is than league average. 132 INT%+ means 32 percent better than league average, meanwhile 100 - 32 = 68 INT%+ means 32 percent worse than league average.
ah, ok! That is a better statistic
I think Tiers 2 and 3 are the most interesting and potentially controversial tiers on the list. It is where you really see the difference between actual performance and popular narratives. For example, I have heard 3 or 4 different people in the media predict that this is the last year Kyle Murray plays for the Cardinals. While that could be true if he has a bad season, based on this year it would be insane to do so. As we have spoken about before, he provided above league average value in every meaningful area (Accuracy, sack avoidance, turnovers, rushing). As long as he can stay healthy he should be a good QB for at least a couple more years depending on what is mobility looks like in his 30s
Another one is Sam Darnold. I'm not sure what the consensus is but I heard a fairly data driven guy call him a bad QB. I know they can be large discrepancies between skill and results but I don't think you can be the 7th most valuable results in a season and be bad.
This leads me to two questions. The first is about your ranking system. I understand the reasoning behind it but am curious about the application. Do you use a formula? If so what weight do you put on different inputs (I'm guessing they are the ones you put in the stats column) and how has that different pre CPOE era and pre EPA era?
The second question is who is the worst QB to be on the best team in the league (in terms of your Exp Wins model)? I know you and David have talked about the worst QBs to make and win SBs but I seems to be harder to be the best team in the league with a below average QB then to make the SB with one.
First, yes this list is ordered by a mathematical formula. It's actually a public opinion list, fitting the players' results to composites combining every best QBs in the NFL list I could find on the internet in the last 20 years or so. It's not objectively a measure of anything. It's taking the values that fans have implicitly placed on different factors over this period of time, and applying them the same way to every player, regardless of narrative. This is what results in the following order of importance:
1) Total touches. To be fair, the internet has kept their word on this. By far the most important factor in this equation is total touches, because the best ability is availability, and placements on tier lists around the internet tend to back up that tough talk. It's also the case that volume tempers the importance of rate stats, which is why it's included as its own separate factor.
2) ANY/A. The most public facing advanced QB stat. It makes sense this would be valued second.
3) Total EPA. Mainly for the purpose of the Josh McCown test. How many plays of elite value does it take before a player is actually high on the list? It did shock me though how the public values total EPA more than EPA/Play.
4) EPA/Play
5) An extremely small adjustment for skill.
These are the only five inputs. Everything else is approximated too well by this combination of five, and just gets in the way if included individually. That's why I didn't feel the need to include TD passes, yards, or anything like that. With a combination of total touches, and the foremost rate numbers, you get a clearer view at who's good at moving the ball without any of the other stuff.
The thing that jarred me most when I was making this formula several years ago is just how little the public cares about skill. It's not nothing, but it's little. In general, ratings in the public eye are done based on the real results, once it's taken into account how many touches a player has. That's why this list ends up so real results dependent, because this is just how the public values players. As such, when I replaced the placeholder values for CPOE that I had going back to 1970 with the fitted ones I made just recently, very few rankings throughout history changed. Maybe a few, but not very many.
Moving into years older than 1999, things have to be done on the fly, but for most of them, the same fitting is done with football outsiders DYAR, DVOA, total touches, and ANY/A, with the same extremely small adjustment for skill. DVOA is not a replacement for EPA, but it my formula, it works well at serving the same purpose.
Moving onto your final question, it's actually the obvious answer. I have not checked this thoroughly, but I believe Joe Flacco in 2008 is the only rookie ever to play QB on the best team in the NFL by my Expected Win formulation.
With Joe's 0.015 EPA/Play, -1 CPOE, 93 sk%+, 5.29 ANY/A on a significantly deleveraged 517 total touches, he finishes just 22nd on my 2008 tier list, the same spot CJ Stroud occupies on this one, which is the worst QB the best team in the NFL has ever had, in my opinion. In terms of xEPA, I believe there are worse, but I haven't cross checked that. In terms of real results, there has been no worse QB on the best team in the league than 2008 Joe Flacco.
I also think tiers two and three are the most juicy every year, as it's where both the underrated and overrated guys end up. I myself said last year that 2024 stood a significant chance of being Kyler's final season with the Cardinals, if he could not turn it around from what had been consecutive seasons finishing 26th and 27th. To Kyler's credit, he did turn that around. He's right back to being what he was before the injuries, either tenth or 11th.
I went over this on my QB contracts note a little while back. Being stuck in tenth or 11th means Kyler is being overpaid. You can get more for less money easily, but does it mean he's a bad player? In no way shape or form. If what you're looking for is a chance to win football games, Kyler is difficult to upgrade from. They were already the best Expected Win team in the NFC West in 2024, although that's not big praise. If the Cardinals are looking to do a Giants impression (i.e. find ways to lose football games), getting rid of Kyler Murray will be a good way to do that.
I have a hard time believing anybody called Sam Darnold a bad QB in 2024. I discussed how the Kevin O'Connell offence has never had a QB significantly overperform his xEPA/Play in it, no matter how much people pretend it has, so in my opinion, you have to try hard to get to the result that Sam Darnold was not very good in 2024. It requires a cocktail of being enamoured with the O'Connell offence, understanding that sacks are a QB stat, but somehow failing to understand that accuracy is the stat that a QB has the second most individual control over, and also falling victim to the stigma that QBs who change teams are somehow lesser than those who don't, or as I like to call it, the Andrew Luck/Jeff George fallacy.
There are reasons to be afraid when it comes to Sam. A lot of them are predicated around the areas on the field in which he is accurate, and the areas of the field where he is not, but in his accuracy profile, he looks very similar to Josh Allen (accurate short, accurate long, not very accurate in the middle, basically the opposite of Lamar Jackson). Josh's profile has not sent him off the cliff, so I don't believe it will happen to Sam either, although I do find it difficult to forecast him back in the top ten next year, for Seahawks reasons. This does not mean he's a bad player though.
That’s really interesting that this is based on an objective formula—albeit one trying to predict more subjective rankings—rather than your own subjective weighting (which is how I read the framing of it). Obviously, I knew you would put a premium on stats, but thought it might be a bit more ad hoc when it came to breaking ties, etc. How does the grouping into tiers come about once you have the results from the formula?
One thing that stood out to me (before I knew it was calculated by a formula) was how true you were to the principle that volume matters by ranking Russell Wilson ahead of Justin Fields, despite your position that switching was a mistake (even if not necessarily a huge one). Even though you’d have preferred Fields over the rest of the season, the real result of the change took that out of your hands and so Wilson ends up a few spots higher than Fields despite Fields being otherwise “available.”
Purdy/Mahomes’s positioning also illustrates that to some extent. The CPOE & sk% aren’t that different, but Purdy’s EPA/Play and ANY/A are noticeably better than Mahomes. Under the guise of “a full season of play/how many plays do I have to replace this QB for,” I don’t think I’d necessarily take Mahomes over Purdy. Purdy’s total touches represent basically a full season within the context of the SF offense, so it isn’t as though I need to replace many of his touches with replacement level play. But since the metric doesn’t actually account for that, the real difference in volume makes up for the poorer results.
I’d also be interested (since the formula is derived from public opinion) in who the formula doesn’t get right. We’ve talked several times about Matthew Stanford, and the difference between his public perception and his quantifiable skill/results. I imagine the formula may have consistently missed on his ranking and predicted him lower than the actual accumulated rankings of the internet. Were there any quarterbacks the formula consistently overvalued relative to the public perception it was designed to predict?
Tiers come about when there is a clear jump in terms of the score. The formula gives each player a score that's mapped to a ranking list (1 means good, 32 means bad), and tiers are made whenever a substantial difference between two people is found in the rankings, to mark that we've moved into a different class of player.
For instance, to break tier six into tier five, the break point is obvious, because here are the scores:
25) Bryce Young - 23.49
24) Drake Maye - 23.46
23) Russell Wilson - 20.98
22) CJ Stroud - 20 flat
You can see the break point there. Despite being only one position apart, Russell Wilson is a fundamentally different class of player than Drake Maye, which is why the tiers come into it. There doesn't have to be tiers. I could mention this in the course of the write up, but the tiers are a natural way that people understand to delineate the different classes of players. It's actually just a crafty way to save myself some words, and not really anything more. As if there weren't already enough words in this list.
Nothing is ad hoc with me LOL. All my lists are mathematical formulas. Either subjective ones that I designed myself, like for the one offs and the upsets, or more objective ones like this, fitted to a very specific parameter, which I believe to be the most objective way to do it, due to reasons I just explained in a reply to Marc. I'll refer you to that, so I don't type them again here.
The Justin Fields/Russell Wilson example is exactly right. That's why this is the results-based tier list. Do I think Russell Wilson in 2024 was a better player than Justin Fields? No, but were his 2024 results better? Yes they were, because he touched the ball more than 150 times more. Volume is the key to all this, because high volume makes putting up good rate stats more difficult. Rate stats mean nothing without volume, which is why volume is the most important factor in the fitted formula. It would be the most important factor that myself or anybody would subjectively design too, and my position is that you either accrued the volume or you didn't.
We can't make special exemptions for guys that were benched and not injured like Justin Fields or benched for non-football reasons like Daniel Jones or were just injured like Tua Tagovailoa or whatever may have happened. You either played, or you did not. We're ruthlessly objective here, and we can get away with that, because the interpretation of the list is, looking back at 2024 with the benefit of hindsight, who would we have picked? This means we know Justin Fields is getting benched after 251 plays and not a single one more, and we can make our choice accordingly.
This carries over into the Mahomes/Purdy example. It's not strictly a matter of how many plays you need to replace a guy for. The assumption of the model is that if a QB touches the ball more total times, each touch is more difficult on average than QBs who touched the ball fewer times, because it's more likely a defence would be counting on them to touch it. This is not always true by the way. For instance, if a rush offence is horrendous, taking touches away from it might actually make a QB's life easier, despite the defence keying in on him more, due to less bad (often second and long) situations, but every model has to make an assumption somewhere. You cannot account for everything.
You've actually found the key assumption underlying my whole process here David. I'm impressed. High five. My position is that football is not baseball. In this sport, there is no such thing as a stat accumulator. Each time a QB touches the ball makes his team more and more pass heavy. As a QB's team gets more and more pass heavy, the opposing defence (think of it as one defence for the purpose of interpretation) starts to sell out more and more to stop the pass. As they do this, it gets more and more difficult for the QB to put up good rate numbers through the air.
The boost is not just a boost for real volume. It's a bonus intended to make up for the fact that the difference between Patrick Mahomes' 695 plays and Brock Purdy's 570 likely also means Patrick's average play was more difficult than Brock's. Patrick Mahomes put up 0.165 EPA/Play on 695 touches, compared to Brock's 0.198 on 570. In volume terms, this is basically equal. 114 for Patrick. 113 for Brock. The difference maker is the belief that Patrick's touches were fundamentally more difficult than Brock's, because the defence is more likely to be selling out to stop the pass against Patrick.
All of this is just interpretation. It's easy to disagree with, but it's the methodology that built this list. This is what I meant when I said my methodology questions why the 49ers did not throw more passes. The unasked but implicitly defined question was 'can Brock Purdy handle the per play difficulty that the guys in front of him, each with more total touches, had to deal with?'
We don't know whether he can or not, because this cycles back to what I was talking about above. In 2024, he didn't. Patrick did. That's why Patrick goes ahead. That's why Jordan Love is so low, and it's why Joe Burrow is so high. It's also why Caleb Williams is so high. Having to drop back to pass on every play makes each one harder, so he can get away with worse results than CJ Stroud can get away with. That would be the interpretation of my methodology anyways. That's why he's above him on this list.
I dislike your connotation that this list is trying to predict public opinion. It's not. It's fitted using the public's opinion, to implicitly place values on different aspects of play. If I were trying to predict the public's opinion, I would need career Super Bowl championships and etcetera in the formula somewhere. This formula strictly strives to weigh every aspect of play the same for everybody, but since I do have the rankings on my list, and several public opinion polls, I can answer your questions anyway.
Matthew Stafford's ranking on my list did not significantly differ until he won his championship. Through basically all of the 2010s, my tier lists have him in the mid teens, and public opinion also has him in the mid teens. I have Matthew fourth on my 2011 tier list, a season that is underrated even to this day, and that's coming from Matthew Stafford's biggest hater, but that is the only significant deviation, until we reach 2021, whereupon people seem to think Matthew Stafford is a top ten guy for some reason, while his results have not budged from the mid teens.
In terms of types of players that are over/underrated by the public relative to their actual results, the general formula is that turnover avoidance is overrated, and sack avoidance is underrated by the public. These are both true in separate, but super effective when combined. This means that guys who are great turnover avoiders but not good sack avoiders (CJ Stroud, Jalen Hurts) tend to wildly overshoot in the public's eye where they would be located on any objective, results based measure like mine, and vice versa.
Guys who are great sack avoiders, but have turnover troubles, often end up really underrated on a list like this. There are few examples of this in 2024. Bo Nix is probably the best, but that's likely more just his sack avoidance being underrated. Josh Allen also used to fall into this in the past.
Stick to the TEP principles. A turnover is 2.4 times as bad as a sack, and that's all. Do not mix up the values of these things. This is the area of QB analysis the public seems to have the biggest trouble with right now. Fans in general have gotten pretty good at weeding through everything else.
Thanks for the clarification! I apologize for my poor wording that the model is intended to predict the aggregate public opinion, rather than just being fitted to the public opinion data. I get your point: the end goal isn’t to match the public opinion data, that’s just the method you used to fix the weights. In that sense, it fits the public opinion data as best it can via the stats you selected, but if you were truly trying to predict the public opinion, you would have included other variables.
What was the reasoning behind crafting the formula so that if fitted public opinion (or more accurately public values) rather than a more objective approach? For instances, you and the analytics community at large value EPA more than ANY/A yet it is more important in your formula.
Do you think there is a greater divides amongst player perception know than before? The reason I say is because the more information we have access to, the more divergent opinions can occur. In the past, you maybe only knew a guy's basic stats and how much his team won. Now, you can watch more of a player's game than ever before and have way more and better data and metrics.
Take Brock Purdy in 2023. If you look at EPA/play, CPOE, and ANYA you would be inclined to say he should have won the MVP. However, if you are an "eye test" guy, you wouldn't have him in the top 10 because he doesn't make a lot of wow throws, plays in a Shanahan offense, doesn't have pretty throwing motion, and is slow. You could also be an (eye test) guy and say he doesn't a great job of running the offense and has great anticipation. If you look at volume numbers (both traditional and advanced) and win %, he would be in the top 5 but probably not a your pick.
I'm not really sure what my point is in all of this, but is interesting how one could come to and argue radically different opinions on a player placed on what you want to believe, who you listen to, and what you care about in a QB.
Have you ever used something like rateyourmusic Marc? A website where people vote on how good something is? This is almost exactly that, but with me stealing hundreds of rankings from all over the place and collecting them all together. As for the contrast between this and an objective alternative, I asked myself this question years ago.
The question and issue becomes, what is a more objective alternative than an averaged out ranking from all kinds of different people? Is it for me to decide which stats are important and which are not? That doesn't make any sense. Is it for me to fit this individual ranking to team stats, like points scored or winning? That doesn't work for me either. To me, the way I have done it is the very most objective way to rank players, without bending to the whims of teams' stats, and without slathering my own biases (I am a bigger fan of small sample greatness than most, we all know this) all over everything.
In the end, the fitting mechanism is only a way to assign weights to things, at which point it becomes a level playing field. Everybody is playing under the same rules. In that way, it's as objective as can be.
It's also not the case that ANY/A is weighted heavier than EPA. I should've explained that better. While ANY/A is worth more on a rate basis, EPA/Play obviously contributes to total EPA, and when you add the contributions of (3) and (4) together, it outweighs number two. It's only as individual factors that ANY/A outweighs them both.
In general, I do not think divergences between my list and the public opinion are getting bigger, as it's always been kind of heavy on splitting the difference. The divergences within public opinion are another matter, because there are different types of divergences. Some have existed forever. Some are new. For instance, the divergence between those of us who think Terry Bradshaw was a league average QB for most of his career and those who think he was a top five guy of the 70s because of all the winning he did were happening even in the midst of the 70s. The modern allegory is obviously Patrick Mahomes, and the disagreement that goes on about his positioning in the hierarchy is a tale as old as the hills in the NFL. That disagreement is no bigger or smaller than it used to be.
When it comes to somebody like Brock Purdy, I think you're right. The disagreement is bigger now because people have the means to disagree now. The obvious case is Jim Everett. I have him top four in my tier list four times, including three in a row, and including a number one position, but in real life, he made one Pro Bowl, which was not in the season I rank him first. It's because all of the credit in his day was stolen by Zampese, not entirely different to what Brock is going through now with Shanahan, and the stats that really love Jim Everett did not exist in 1988. It was all about passer rating and, if you were an advanced metric guy of the era, ANY/A, all of which WAY overdiscriminate against turnovers, meaning all it took was Jim's career INT%+ being only 100 to sink him, because the stats of his day did not see him as being too good to deny in the way the stats of today do.
If Brock Purdy were playing in 1988, he and Jim Everett would swap places. I would be angling that Brock Purdy was a Hall of Fame guy, with the benefit of modern metrics, but modern fans wouldn't have even heard of him. They would have heard the narrative that he was in a great offence, like they've heard about Jim Everett and Trent Green, and leave it at that.
More metrics being available today does create more grounds for disagreement, and naturally, more disagreement. I think this is what you were getting at, and I think you're right.
This is a great list, and there's so much that could be said here. Obviously there's no debate about the top spot. The only knock on Jackson is that he hasn't made it to the Super Bowl yet. And I'd say it's certain he will, except that the the next two QB's on the list also play in the AFC. I honestly can't say which of those two I'd rather have on my team, though if pressed I'd probably choose Allen.
It seems odd that #4 on the list is in the tier of quarterbacks who can carry a team to the Super Bowl but also is described as a Game Manager. Maybe it's not a contradiction, but it seems like it should be.
Speaking of game managers, I wonder how much of Patrick Mahomes apparent decline is really Travis Kelce declining? What kept Mahomes from being just a game manager in previous years was his ability to pull a rabbit out of a hat and keep drives going even when you thought he was beat, and most of the times that happened it was Kelce who made the catch. I'm certainly not going to complain if Mahomes has lost that magic touch, but it'll take more than a down season or two for me to feel comfortable with rating him as merely above average.
Moving on to Bo Nix, there's plenty to say. His season overall was good, obviously, but what I was happiest about was his growth over the season. His first few games were pretty rough, with just enough glimmers of promise to keep it going. Lots of inaccurate throws. But he improved. A lot. By the end of the regular season he was the best rookie in terms of completions, yards, touchdown passes, and total touchdowns, and he did that with a weak rushing offense, mediocre (at best) tight ends, and only one elite receiver. Yes, I do think Sutton can be elite. He's capable of making some unbelievable catches, and seems to have a good rapport with Nix. Given the improvements Denver has made in the off-season this year, I think both will have a great season. I'm confident he'll be solidly in tier 3 at worst, and most likely in tier 2.
There's almost no way Lamar will fail to make the Super Bowl at some point. By xEPA/Play, he's getting close to being the best QB of his generation as he rises and Patrick falls. There's never been a QB who was the best (or even second best) of their time that never made it, except for Dan Fouts. Ken Anderson took ages, but he made it, and he's the only one that even came close to being shut out for a whole career that isn't Dan Fouts.
Take Peyton Manning as the example. We can all admit that he was not a playoff performer, but through sheer talent, even far below his best, he managed to make the Super Bowl four times. Is Lamar Jackson at the talent level of Peyton Manning? Generally not, but in 2024 yes. Like I said, if he can keep this 117 sk%+ going, he will make a Super Bowl imminently, almost certainly multiple. If he goes back to the old no pocket presence Lamar, it's debatable whether he's even a Hall of Fame guy. This is what puts me on edge about Lamar Jackson. In his two seasons with a sack rate less than six percent, he was number one (but no tier zero) in 2019, and tier zero this time around. The problem is there's only two seasons with a sack rate less than six percent, and he's only got three career top ten finishes in what is now seven qualifying seasons. Not quite Dak Prescott, but a very inconsistent player.
The game manager thing comes as a result in a slight difference of opinion on what exactly the term 'game manager' means. To me, game manager is a play style, consisting of short throws, easy throws, EPA/Play values that underperform very high success rates, and in the modern day, NFL NGS CPOE hating you, as a result of only throwing to wide open targets, with the addendum that running QBs who fit all the other characteristics (Jayden Daniels, Kyler Murray) get a pass. Some people take the term game manager to be a reflection of the quality of play. I do not.
Taking the above to be the meaning of the term game manager, there have been plenty of game managers in the upper echelons in plenty of different years, including a tier zero season that I feel absolutely fits in the game manager parameters (Patrick Mahomes 2022). I don't think the terms 'game manager' and 'elite QB' are mutually exclusive. You can be both at one time. It's tough to pull off, which is why it doesn't happen often, but it can be done.
As far as Patrick and Travis Kelce go, it's very possible that what you're saying is correct. I did say in the article that even a single top 40 receiver would help Patrick out a lot, but I think 2023 raises a mild argument to this. In 2023, Travis had declined, but not to the point of no longer being a top 40 receiver, like he declined this year, and there was also Rashee Rice. Nevertheless, Patrick ranked tenth in EPA/Play last year, and tenth in EPA/Play this year (I use ranks because Patrick's raw numbers rose with the tide, just like everybody's did).
I do want to make clear my position that Patrick Mahomes was likely playing over his head the whole time, which would be an agreement with you. This is not an assertion to make lightly, but it must also be said that even Alex Smith put up a top three results season with Tyreek Hill, Travis Kelce, and this Andy Reid offence. Now that Tyreek is gone, and Travis (as an elite player) is gone, we're going to see what Patrick Mahomes is really made of, and so far, it's not been a great showing.
Patrick is an odd football player. His combination of fantastic physical talents with a game manager's playstyle is not something the league has ever really seen before. In my opinion, this is why a lot of the advanced metrics don't like him. Even my baby, xEPA/Play, has Patrick as the fifth best QB of all time, but even that is an underestimation compared to his results, which are better than fifth as of right now.
Every advanced metric out there thought Patrick was due for a come down. The issue was only how far. There are ones like NFL NGS CPOE, who think Patrick just sucks, and can't throw the ball accurately at all, but there are also others like my xEPA/Play, which thinks that while top one may be out of reach for Patrick at this point, top five is still possible. I think this is much more reasonable, although with a sack rate in the high five percent bracket, I don't see Patrick Mahomes as an elite QB. I'm not changing my stance on that. Merely predicting that he won't end up this bad at sack avoidance in a full season again.
I still think I put Patrick pretty high. It took everything in the world going right (multiple different kinds of unsustainable luck) for Justin Herbert (a player who some people think is a top five guy, although I vehemently disagree) to sneak ahead of Patrick into eighth place, a miracle that will not happen twice, but even if it is eighth place, I do think Patrick has lost a little bit of the magic. It's not like 2022 Tom Brady, but just for the record, 2022 Tom Brady finished ninth on that tier list...
Onto Bo Nix, we are almost entirely in agreement. Bo is a lot like Bryce Young, in that a few bad games at the beginning of the year weigh his season-long numbers down a bunch. In the same seven game stretch in the second half of the season where I discussed Jayden Daniels going through the falloff (a falloff that is quite customary in the second half of a rookie season, by the way), Bo Nix was doing the exact opposite, generating 0.255 EPA/Play in the back half of the year, sixth in this span, behind only the ironclad top four, and the late season surge of Baker Mayfield. Name a QB outside the top five, and I'll tell you that Bo Nix was better than them, in the second half of the season.
This list looks backwards, so the weak first four games that almost dropped a great Denver team out of the playoffs altogether have to count against him, but if it were looking forwards, picking guys for 2025, Bo would be higher than 14th. There's no question there. In the final 12 weeks, his CPOE was a much better looking (especially considering the arm isn't the appeal) 2.1. If you substitute this 2.1 in place of Bo's 0.6 he got in real life, all of a sudden Bo Nix is the league's seventh placed xEPA/Play finisher, and one of the guys above him just retired.
It's a stretch to say he can be a top five guy next year, because we will see about the 121 sk%+. He only has to repeat this level of sack avoidance one time to convince me it's for real, but if he can repeat it, than a 2.1 CPOE will be good enough to put him into the top five, skill-wise. He played like that in the final 12 games of the 2024 season. We will see about the real results, but good news for a Broncos fan is that Bo is one of the very most underrated players by his real results in 2024, and his real results were still enough to get him to 14th place as a rookie. That is mightily encouraging.
Absolutely incredible read, I don't know how you manage to write this much but it's an absolute pleasure to read. I've always been more of a stat guy over a results guy, so these kinds of reads feel great to read. Amazing job Robbie, as always :)
Thank you very much my friend!
Like I said, this was a mammoth project. If anybody was wondering about my radio silence over on Notes for most of the last week, this list is why. It took everything I had to get this list to an acceptable standard, as they require an even higher level of quality to overcome the 'lazy list' stigma, but I think I managed to pull it off. Thank you very much again for your kind words. There is not enough objective QB analysis out there, and I sincerely hope that I'm able to provide some.