Hello all! Welcome back to another analysis piece!

This blog post kicks off my 2023 position group review series. I intended to keep this version shorter than the 2022 one, as I will merge clustering analysis into these pieces. Additionally, this work will be simpler than some of the blogs I've published recently, so it will have analysis in the same vein as the blogs recapping the 2022 season. Modeling and metric work will continue after this series - hopefully by then I will have the upgraded compute ready to go.

This specific piece will focus on fantasy point production and clustering based off the main sources of quarterback fantasy production. Follow up parts will narrow in on rushing and passing stats.

I should also note that fantasy points may look slightly different than your league's scoring, as the data is coming from an NFL data python library.

With all that out of the way, let's dive into the review for quarterbacks!

We'll start by taking a look at fantasy points production in 2023.

I want to note, before discussing the plots below, for a quarterback to qualify for any of the analysis in this piece, they must have played at least 5 games in 2023 - this goes for both season totals and per game numbers. The graph below plots the top 30 quarterbacks by fantasy points:

First, I want to point out that Josh Allen is QB1 overall, and by a pretty decent margin. The gap between him and, the QB2, Hurts is about 37.82. That's a larger gap between Hurts and, the QB5, Jordan Love (margin was about 37.76). Allen outscored the average QB1 - who is at 307.71 fantasy points, including Allen's numbers - by 86.93 points. Divide that by 17, and you get a 5.11 points per game advantage with Allen, against the average tier one starter.

The common motif in the 2022 review was that the best players in a position group are worth every penny, as their dominance provides so much value over other stars and starting value fantasy players. On the first graph of the 2023 review, Josh Allen exemplifies this perfectly. Sure, there are other QB1s that are exciting and talented, but there's only one Josh Allen in fantasy. He is the Travis Kelce (perhaps of yesteryear) of quarterbacks.

For the Mahomes fans, I think it's time to admit that Josh Allen is head and shoulders above everyone in the realm of fantasy. If it wasn't for the tragedy that happened Damar Hamlin, there's a good chance Allen is QB1 for four straight years! If you own Mahomes in a league and could trade him for Allen and change, I would pull the trigger in a heart beat. I realize Mahomes is the better real-life quarterback, but finishing as QB8 surely has to bump him down off the Allen tier.

Before turning this into a Josh Allen stan article, I want to pivot here and focus on some other guys.

It was good to see Lamar and the Ravens, mostly, stay healthy for a season. His ceiling is high, but I find it difficult to trust him to be a top 3-5 quarterback consistently, simply because of his injury history and how the Ravens handle injuries across the board.

Prescott was also back to his best this year finishing at QB3. I think he'll stay in the QB1 conversation, but replicating this season would be tough, as he had one of the easiest schedules from a fantasy perspective this past year:

The bar plot above shows fantasy strength of schedule (aka Score on the legend) and fantasy points for the 30 easiest schedules for quarterbacks. Quarterbacks towards the right of the graph had the easiest schedule. Those towards the left had the hardest from this list.

The difficulty here is predominantly based on passing stats for both quarterbacks and defenses. For more info on how the score was calculated, please check out the appendix.

Back to looking at Prescott - he had the fourth easiest schedule, which is why I see him dropping down the fantasy rankings in 2024.

In a similar vein, Purdy also had a pretty easy schedule as well, but I like his stock more as Purdy is surrounded by elite talent and will have Shanahan around for a while. If the 49ers upgrade their interior oline, I wouldn't be surprised to see Purdy improve again next season.

I also want to highlight both Stroud and Love. Both are young and talented quarterbacks, in seemingly good future environments. Finishing as QB1s and the way in which they did it is impressive. Stroud has flashed a much higher ceiling with less help, but Love has a supporting cast that is flourishing with young talent. I'm not sure the gap is quite as large as KTC makes it out to be - as of this writing, Love and AJ Brown are considered fair value for Stroud.

There's a lot to like across the board so far, but there's also a fair share of disappointment here.

Outside of Mahomes' very off fantasy year, the main three that came as a surprise were Tua, Trevor Lawrence (TLaw), and Fields.

Tua just barely finished as a QB1 despite being in an electric offense. At this point, it seems like he's destined to be a lower tier QB1, who may have a hot year here or there. Unfortunate, considering how many points Miami puts up.

TLaw also just finished as a QB1. He had some injuries and the rest of the offense wasn't quite clicking, but after two years of Pederson it doesn't look like TLaw will be living up to his original hype.

Lastly, Fields finished in QB2, but that was mostly due to injuries. If we see the per game version of fantasy points, the outlook on him is a little different:

Fields finished as a QB1 on a per game basis, which appears exciting. The drawback on Fields in future years is that he 1) might get replaced by Caleb Williams in 2024 and 2) didn't improve enough as a passer to where his future might be safe on a different franchise. His anticipation and reading of a defense was better, but not great:

So, Fields is great for fantasy because of a rushing upside, if he's a starter. The question we don't know the answer to, is how long will he be a starter in the NFL?

I'd like to point out that TLaw and Tua are both firmly in QB2 range on a points per game basis - not ideal. Moreover, as a bit of a fun fact, Mahomes is behind Fields... at QB13... still think he's better than Allen in fantasy?

Two other quarterbacks that deserve some attention are Burrow and Watson.

While Burrow finished as the last QB2, he suffered from average-to-poor offensive line play, injuries, and a middle-of-the-road schedule. With all those factors, I would expect a healthy Burrow to return to form next season. That being said, I'm not sure he is a buy low, as Burrow is ranked at QB 5 on Keep Trade Cut's (KTC) rankings, somehow above Hurts - whose had two back-to-back top 3 years, go figure.

Watson finished just behind Burrow on the points per game scale. However, he left quite early in one of the six games he played. If you were to average his five other games together, Watson would be sitting at about 18.3 points per game. That's mid-to-low end QB1 territory. I understand the dislike surrounding the player, but his value is far below his fantasy performance. Watson is currently QB19 on KTC, which is spectacular value for a low-end QB1.

This section uses K-Means Clustering to place quarterbacks into tiers based on the four main sources of fantasy points: passing yards, passing touchdowns, rushing yards, and rushing touchdowns. I also threw in fantasy points for cluster creation, for good measure. For each of the season total and per game numbers, I created one cluster graph. I used six tiers, as that offered quite a bit of nuance, while barely sacrificing optimal cluster size.

For each of the sub-graphs in the cluster plots, the upper right of the sub-graph is the most ideal position for a player to be in.

Thanks to those who've made it this far - I'm excited to breakdown the rest of the 2023 fantasy season with y'all!

In the next one, I'll dive more into passing and rushing stats for quarterbacks, which should wrap up the 2023 review on QBs.

One last time, the recap:

• Allen is in a tier of his own. Mahomes is probably the only guy just beneath him. Again, this is for fantasy purposes.
• Burrow and Hurts - more so the latter - are solid buy lows for elite QBs this offseason.
• Stroud and Love have a large disparity in value between them, but I'm not sure the numbers justify that. Stroud does appear to have a higher ceiling, but Love seems to be more consistent with higher quality games (i.e. Stroud can drop a 50 bomb, but Love seems to be in the 20s more consistently). Would love to have either.
• Prescott and Lamar are my sell high candidates. Prescott had a really easy fantasy schedule this year, so I expect regression. I'm not sure Lamar will put up another year of this caliber next year. He has shown an MVP season can be repeated, but it feels like a lot has to go right for him and the Ravens for that to happen.
• Cousins, Goff, and Stafford are all decent vets, who would be awesome as QB2s. Cousins might still be a QB1 and is somehow about the same value as Stafford on KTC.
• Purdy is an interesting case. Regression could very well be on the cards due to how easy his schedule was for fantasy. However, he's still young, has a lot of weapons around him, and I expect the 49ers to improve their interior oline this offeason, so for that reason I think he could improve or maintain his current production.

Cheers,
Alex

Term definitions below:

• passing_a_y: How far the football moves on a pass from the line of scrimmage to the catch point (does not matter if the pass is incomplete).
• pacr: Passing (yards) Air (yards) Conversion Ratio - the number of passing yards per air yards thrown per game.
• dakota: Adjusted EPA + CPOE composite based on coefficients which best predict adjusted EPA/play in the following year.

These definitions come from the R Data Dictionary.

The calculation used to strength of schedule first starts with calculating defensive rating for each defense, for every matchup of the season. This is done via the following formula (for stat definitions refer to the glossary above):

                    
                        defensive_rating = (completions * -0.02) + (attempts * -0.01) + (passing_yards * -0.04) + (passing_tds * -5) \
                        + (ints * 2) + (sacks * 1) + (fumbles_recovered * 2) + (fumbles_forced * 0.3) + (passing_a_y * -0.01) \
                        + (pacr * -10) + (dakota * -5) 
                    
                

Then, these ratings are grouped up by quarterback, so that we have the all of the defensive ratings for each match-up a QB played in. These are then averaged to get the average defensive rating a quarterback played against during a season. 40 is then added to each average so that they are all positive.

So, the more positive a strength of schedule score for a QB, the tougher the match-ups were, as a less negative defensive rating implies the defense played better.