4 Comments
Jul 12, 2021Liked by Matthew Coller

Very interesting stuff.. I saw somewhere that stated the gain that offenses get from going to a bad OL to an average OL is much bigger than going from an average OL to a good OL... Obviously it is not obvious that the youth movement will pay huge dividends this year (though maybe by 2022) but according to PFF the Vikings were #26 last year....Even a jump to 18-20 would be pretty helpful....

Expand full comment

I think that Klint is an underdiscussed variable when it comes to how this season plays out. We all know that how well our OL performs, whether Dalvin stays healthy, how often GoodKirk comes out to play, and how often BadKirk stays banished to the netherworld will have substantial impacts on how this season goes. That said, as Matthew suggests, each of these may be notably impacted by how Klint manages playcalling. The more that our playcalling is nuanced and keeps our offense out of third and long situations, the better our OL should look. The more we have a light touch with Dalvin, the healthier he should be. The more we use Jefferson short and give Kirk unpredictable playcalling, the better (and more consistently) he should perform. I think that Klint could legit be the difference between, say, so bad as 7 wins this season (if he is a noticeably worse version of his old man that further strains our young OL) or as good as 13 wins this season (if he is a significantly better version of Stefanski), to give the absolute outlier projections. Probably more likely that his playcalling is the difference between 9 and 11 wins, all things equal, but the variance could be even greater.

Also, I'm pretty sure that you have discussed this before (or it might have been a different podcast), but one reason why offenses might run more on 2nd and short is just to give their OL a breather. Teams are passing more than ever before, and every offensive lineman will tell you that it is harder and more taxing and less fun to pass block than it is to run block. If a team has been passing their way down the field, an OC may take the occasion of a 2nd and 2 or whatever to run just to give their OL a break. After all, it makes sense to take advantage of a greater chance to convert (which your stats seemed to show exist) when that playcall also has the advantage of making your OL happy and, potentially, better over the course of the full drive (and/or full game) by giving them a bit of a break amidst a historically-unprecedented passing game.

Beyond that, I agree that all of the things you discussed above would help efficiency, but also points scored and % of drives that end with points don't really measure efficiency. These two just measure, well, points, which are basically by definition flukey and not really static or predictive stat over time. For example, it is basically guaranteed that the Vikings would have extremely poor points scored and % of drives that end with points being as they have the worst starting field position in the league when paired with the worst field goal percentage of the league, both of which are not strongly related to their starting 11 on offense, and both of which they did by wide margins. Specifically, the difference between the Vikings and the 31st place team in the league for FG% was bigger than the difference than the 31st place team and the 28th place team, and the difference between the Vikes and the 31st place team for average starting field position was bigger than the 31st place team and the 26th(!!!) place team.

Also, EPA doesn't really measure efficiency either, but rather EPA is (again) tied directly to the actual points of the game, as the total EPA in any given game is always equal to the actual points scored in the game. Put differently, the total EPA of a team within a game will always be the exact same as the actual score between two teams in a game, as at the end of each game they use history to mathematically apportion the relative effect of each play to the final score. EPA is very valuable and interesting and will tell you a lot, and it is fantastically insightful on a micro level (as it divides out how individual plays can contribute to scoring when compared to the history of the game), but it frankly somewhat less valuable on a macro level, as it is explicitly handcuffed to the score.

If anyone wants to discuss efficiency, they should use DVOA. Incidentally, DVOA still supports all of Matthew's conclusions, which makes sense as they are good conclusions. Specifically, each of the teams in the NFC championship game were top 5 in offensive DVOA last year. Over the last 7 years, only 3 teams that have appeared in the championships game were not top 8 in offensive DVOA (Jags in 2017, Denver in 2015, and Colts in 2014, each of which were clear outliers for various reasons), and the superbowl champ was top 6 in each year but with Peyton's Broncos in 2015 (another objective outlier), as well.

To give an example of how DVOA is better than EPA by comparing two actual plays from last year, let's say that one team scores a touchdown on 1st and goal from the 2 against last year's Steelers (arguably the best defense in the league last year, and one of the best against the run) to go up 14-3 in the first quarter as the Texans did last year, and let's say that another team scores a touchdown on 1st and goal from the 3 against last year's Vikings for a Xmas day present (by then the Vikings were easily one of the worst rush defenses in the league, down basically everyone important in their front 7) to go up 52-33 with less than two minutes to go in the game. According to EPA, it is more impressive and more meaningful for Kamara to score his TD to go up 52-33 with 110 second remaining than it is to run in a TD from 2 yards out to go up 11 against the Steelers in the 1st quarter, because the Kamara TD was one yard further away - in other words, EPA completely ignores your competition and it completely ignores game situation. There are a host of other things that EPA ignores that are meaningful, but this is just one very quick way to illustrate how EPA is a very interesting but objectively pretty blunt tool that lacks nuance, especially when compared to something like DVOA.

If I recall, Coller said in a recent podcast that he was ignoring DVOA because it said that the Vikings had a slightly above-average defense last year before injuries really ravaged the team (they got as high as 9th after week 13 when they had a record of 6-6). Which, I mean that's fine, I guess, but to me it seems like dismissing the entirety of a widely respected stat that is purely analytical because it doesn't match up with your eyes in one instance without even trying to understand why it did so is a curiously anti-analytics viewpoint. I would say it is akin to dismissing PFF grades for all time as dumb and irrelevant because they didn't say that Xavier Rhodes was a top 14 cornerback in the 2017 season, as the eye test clearly said that he was, particularly given his competition that year.

I have mainly stopped trying to get Matthew to reject Football Outsiders and to accept PFF as The Ultimate Truth, but I can't help myself now and again from taking a potshot, if only because I want my three football besties (Vikings, Purple Insider, and Football Outsiders) to play nice. Love you, Matthew!

Expand full comment
author

Love you too but... There’s definitely some misinterpretation in here. What I’m saying in using scoring %, EPA and points is that those measures tell you how much you were scoring and in order to score more, you have to be more efficient by making the right decisions situationally and using players to their maximum, which is why I mentioned Jefferson’s yards per target, Cook’s yards per carry and Cousins’s impressive stat line on 2nd and short. I only used EPA to demonstrate that the results were closer to a Tier 3 offense, so.. how can they better those results?

The Vikings were 27th in pass attempts last year, I’m sure their linemen can handle a handful more drop backs per game.

Also I do not hate DVOA. I think it’s used poorly sometimes to make points that do not pass the smell test, particularly early in the year when people try to argue that a team is actually good or actually bad because of DVOA when, IMO, you need a much bigger sample for it to tell you what you’re looking for.

Also, I’ve never ever said PFF is the ultimate truth. I use PFF’s data and conclusions as a tool in conjunction with my reporting. I think they do an excellent job of studying the game and presenting new ideas i.e. coverage being more valuable than pass rush.. but if you listen to all the podcasts with PFF folks, they always include deep dives into what their numbers are saying.

Expand full comment

I think we are both misinterpreting each other a bit here, in the least. As I said, I agree with literally each of your conclusions in how to improve efficiency. You don't need to convince me that efficiency can be improved, and that what you suggest is the way to improve it, and that if it is improved that scoring will improve. What I *AM* saying is that, if you are arguing about efficiency, it is very weird to not use DVOA. It would be like if you had a big article about ball security, and then used scoring stats to make your arguments, and then never once gave any stats on turnover differential. If the article is about efficiency, you should use weighted DVOA, as that is the efficiency stat.

Moreover, it isn't even as if DVOA disagrees with you; weighted DVOA had the Vikings as the 11th best offense on the year, packed in tight right behind Cleveland, Baltimore, and Indy (and substantially behind the top 5 of GB, TB, KC, the Titans, and Buffalo). That feels fair to me - whether or not the Vikings are in a third tier, they are clearly well below the 1st tier.

Along those lines, of course the Vikings can pass more, and their OL can handle more drop backs (well, maybe they can, or maybe they'll be garbage at pass pro, but I understand and agree with your greater point). As I said, I am not debating a single one of your conclusions when it comes to the Vikings. However, the article was trying to determine "reason[s] teams are running so much" on 2nd and short across the league, not JUST for the Vikings, upon which you speculated that this was just to get a 1st down. I was just trying to point out that, in some circumstances and for some teams, the reason might be to just throw the OL a bone. To be clear, I am not even saying that this is the correct call (I don't know one way or another, and haven't seen a study as to whether or not OL performance tends to lag when throwing % goes over some threshold that the Vikings would obviously be below), I am just trying to provide a potential reason why any team would ever run on a 2nd and short, as there is a very obvious reason that the article didn't state despite the article raising the question.

Otherwise, I guess maybe you just were using hyperbole in the podcast when you recently said that you were "over" DVOA because of how they rated the Vikings high midseason? That is what I was reacting to. Certainly you do get a bit more bombastic and hyperbolic in the podcasts as compared to the substack, which delights me and of course helps make the podcasts more amusing. Also, whether or not you were arguing as much, I would strongly argue that 13 weeks of weighted DVOA data is a sufficiently large sample size, just to be clear on that point... I just think that the defense last year wasn't quite as trash the whole way through the way that some were indicating - I mean on the year it still somehow had the 3rd best 3rd down defense in the NFL and the 2nd best red zone defense. It was objectively a bad defense when it came to the results and the talent, but overall it was still weirdly efficient until Kendricks went down.

As such, I guess my point is that whether or not you literally "hate" DVOA, you never ever use it, which is very curious to me. It is a great stat, and, when used in conjunction with your excellent reporting, would make purple insider even more delightful. I have always found the absence of DVOA data within your columns curious. Maybe you know something that Bill Barnwell et al. do not?

Otherwise, yes, I did my own bit of hyperbolizing in stating that your stated opinion is that PFF is the ultimate truth. I mean, clearly you are a bit of an evangelist when it comes to PFF - you only need to look at the purple insider podcast feed to see that. You also have regularly made points such as "PFF grades are clearly useful, proof of this is that they use them on the NBC games," ignoring that they are on NBC because Chris Collinsworth is the majority owner of PFF (such that he is trying to build his own product by demanding that they be on there). I like PFF, and like that you use them and have good relationships with their legit football experts. I just like needling you about how much you obviously trust their analysis, which is a surprising dichotomy when paired with how much you apparently distrust (or at least never use or reference) the analysis of anyone at football outsiders. *shrug*

Expand full comment