Football Outsiders recently released their 2021 Football Almanac, a resource choc full of statistics, analysis, and predictions for the upcoming NFL season. Mike Tanier of Football Outsiders was kind enough to sit down with Hogs Haven for an interview to discuss their Almanac and some of the analysis and predictions most relevant to fans of the Washington Football Team. This interview will be published in two parts, look for the second part coming soon.
HH: How did you first get involved with Football Outsiders, what are your responsibilities there, and how did you contribute to the 2021 Almanac?
I have been part-time with Football Outsiders since 2004. If you find a 2004 Almanac in a bin somewhere and want to read things about Drew Bledsoe, I’m in there somewhere. But I was only recently promoted [to full time], so it was 17 years of job interview. This is a tough company to get ahead in (laughs). And of course I was working part-time for Bleacher Report, The New York Times, etc, back then. I am now a Senior Analyst here at Football Outsiders. We’re expanding, we’re growing, we’re adding more fantasy services, we’re adding Twitch streams and podcasts and I’m here now full time. I wrote 4 chapters for this year’s Almanac and I think that brings the grand total to 775 chapters for the Almanac over the last decade plus.
HH: You wrote the chapters for the Cowboys and the Giants, right?
Yes I did, and also the Bengals and the Ravens, but for Washington fans’ [interest], Cowboys and Giants. A positive one and a not-so-positive one.
HH: A lot of people mistakenly believe that football analysts just look at numbers in a dark room somewhere, but the Almanac points out that your analysts actually spend a lot of time watching game tape. How much tape do you study and how does that inform the analysis given in the Almanac and on the FO website?
You know, it’s funny, I don’t know where they think the numbers are coming from if we’re not watching the games, you know? Oh they just download stat sheets or something? Obviously we’re watching. When you look at the statistics we have, things like blown blocks and dropped passes and a cornerback’s rate in coverage (completions allowed). I’m not personally checking off each of those stats, but we’ve got an agreement with Sports Info Solutions and some others for that. During the course of the season, I probably watch about 90% of the games, give or take a Texans-Bengals game in week 16 - eh, I can live without that.
While writing, for example with the Giants chapter, I watched a total of 5 or 6 Giants games intensely, really paying attention to the linebackers and the offensive line. Part of that is to corroborate the statistics we have and answer “why is this?” If there’s a 7-sack game, we want to understand who was the offensive linemen who blew 3 blocks when he was blocking Chase Young and see why he did that. Everyone who does scouting is doing analytics as well, because the scout is going to the coach or the GM to tell them “that guy blew blocks 23% of the time” and he does math to get there. Everyone who’s doing analytics is doing some level of scouting, even if it’s just to get the data that we’re then going to crunch.
HH: DVOA is one of the most important stats developed by Football Outsiders. What is the correct way to use DVOA and how do you see it misunderstood or misapplied by others?
It makes a very good simple [team] power rating. It’s good at spotting undervalued or overvalued teams, usually overvalued. Overvalued team DVOA will usually scream its head off at us percentage-wise. A 5-1 team with a bunch of close wins against bad opponents where they kicked five 56-yard field goals and that’s how they won, DVOA will spot it.
At the same time, it’s not just a power ranking. The best way to use it is we have offense ranking, we have defense ranking, we have run defense and pass defense ranking, and special teams ranking. When you get our premium package at FO plus, you can get receiver rankings by types of pass, things like that. So it isn’t just used to tell you “this team is 10th best.” It can tell you they’re 5th best run defense, but 18th best pass defense. You can get those breakdowns and use them to make more informed wagering decisions, if you’re a fantasy person [you can better determine how your players should do], or if you just want to know the teams better, that’s the best way to use it.
The main misunderstanding we get is when that 5-1 team is ranked 14th in DVOA, when everyone sees them 3rd in the simple power rankings, that’s when the hate mail comes in.
HH: Did DVOA spot the Steelers as being overrated when they were 11-0 last year?
Yes, that was a good example of it. In fairness, a lot of people were looking at that team and thinking “that was a really ugly, yucky win.” Roethlisberger’s passes were going like this [makes wobbling duck hand gesture], but yes, it was sitting there that this was a team primed for decline. Over the years we’ve been best at spotting the overrated team. People send us hatemail, but I hope the smart [readers] are taking the under because very often it is correct. It’s harder to spot the underrated team because if they’re playing well, statistically they’re going to get some wins.
HH: What are the most predictive stats for which team will win a given game?
During the course of the season, great teams pass well and stop the pass. Now that’s the opposite of what coaches say, they say that you win by running and stopping the run. If you look at raw stats you get confirmation bias that’s the case. Look at the Chiefs and how much they run the ball, but every casual fan can see that it’s mostly in the 4th quarter, they ran the ball 17 times and the opponent was passing to catch up, and that’s why it looks like they have a great run defense and a great run offense. But we all watch the games and they throw for a million yards in the first quarter and then they take the foot off the gas, [so gameflow explains the stats.]
You also have to be super careful with takeaways. At the end of a game, you can look and say this team fumbled 4 times and this team didn’t turn the ball over, you know who won. But you have to watch it when you’re trying to predict the next game forward. Some takeaways are very predictive, like if you’re getting lots of interceptions because the QB is on his back with Chase Young and Montez Sweat on top of him, that carries over from week to week. If you’re just recovering a lot of fumbles, you [drill down] and there were 10 fumbles this year and we recovered all 10 of them, that’s not because your defenders are great at bouncing on balls, that’s because you got lucky and the balls bounced into them. So you have to watch things like takeaways, sometimes it’s super predictive and sometimes it’s actually lying to you and it’s the reason the Steelers started 11-0.
HH: What are the most predictive stats for a team’s win-loss record in the upcoming season and how do you calculate projected wins for a team in the Almanac?
The projected wins come from 10,000 simulations. We put in a lot of data and there are tons of variables and you get a spread. The number for Washington is 8.1 wins, but that’s part of a spread because [the simulation finds] a bunch of 10-7 seasons, a bunch of 9-8 seasons, there are some 13-4 seasons, and there might be a few 4-13 seasons.
From a predictive standpoint (and it’s bad news for Washington) the offense last season is more predictive than the defense. Last season’s offense is likely to carry over, whereas defense is more subject to the QBs you faced and things like that. You can count on a QB’s performance, it’s more likely to follow a gentle curve, whereas defenses don’t do that.
There are other indicators. A team that’s lights out on 3rd down, but not good on 1st and 2nd down [is a red flag.] Chances are they are the same team on 1st and 2nd down, but on 3rd down they got lucky and someone slipped or they have a great scrambler who made a couple of miracle plays. You can find under- and overvalued teams like that.
HH: The Almanac includes a “pregame” section with many lessons learned over the years of Football Outsiders analysis. What are one or two of the most important lessons learned you think football fans should take away from this section?
One of them that has permeated the league is the value of the running back. When we started writing the Almanac back in 2004, people were like “what do you mean RBs aren’t valuable, I just watched LaDainian Tomlinson do XYZ?” Now it’s still debated by Ezekiel Elliott’s agent and some father-in-law’s on Facebook, but it’s not that RBs don’t matter, it’s that if you’re trying to build around the RB you’ve got 1 or 2 years that you know they’re going to be spectacular, then they’re probably going to get injured and their play will fall off. They should not be paid a great deal of money and you’re better off having a 2, 3, or 4-person committee. I think that’s the big one that has permeated conversation.
The other one is the one I just described: one of the first things I’ll look at when I have a 4-0, 5-1 team (that doesn’t have Aaron Rodgers going lights out) is I want to look at their fumbles. The total amount of fumbles isn’t random, Chase Young is going to get you more strip sacks than the next guy. It’s the recoveries that are random, because he’s going to strip sack Dak and Dak’s going to fall on it, or he’s going to strip sack Dak and Payne or somebody is going to fall on, that’s the random part. You have to watch for that, because that’s going to sometimes tell you a lie one way or the other about the team.
HH: When analyzing offensive line play, have you seen indication of differences in the positional value of the different positions along the offensive line? Is there any truth to the belief that centers are more important to run blocking and tackles are more important to pass blocking, or that left tackles are more important than right tackles?
Left tackle is a little more important than right tackle, but it gets really overstated. The tackles are important for pass protection, that’s true. Most of what we’re doing analytically, with the recognition that the center has a special job, [we treat] the interior line as the interior line. It’s almost like you have to put them together as one box of people, with the acknowledgement that Washington has a very strong center and guard that we all know about, so that means a very strong interior line. You guys have questions at the tackle positions, there are some good solutions there, but those are potential question marks in pass protection.
The thing about all of it is that the defense will find the weak link. This is what Aaron Schatz (our president) says all the time: “the weakest link of the offensive line is the problem, no matter what position it’s at.” You can move Aaron Donald anywhere you want and line him up over anyone you want [Editor’s note - this triggers memories of Aaron Donald lined up opposite Wes Martin that I’ve tried hard to suppress]. So if that RT isn’t good, the edge rusher’s going to be there and the blitz is going to come from his side. If you’re weak up the middle, they’re going to shuffle the new NASCAR package and they’ll put the DE on the inside. That’s the biggest takeaway, that the weakest link of the offensive line is the weakness of the offensive line no matter where it is.
Other than that, the conventional wisdom there is... conventional. You want a good LT. You had a good one there for years, now you brought somebody in Leno that is potentially a replacement.
HH: If you are building a defense, what do the stats say about which positions you should prioritize and which you should devote fewer resources towards?
This is such a topic, especially in the analytics world. When you look at the stats one way, they say put it all in the secondary. The other way says put it all in the pass rushers and the defensive line. Older-school people (and I’ve talked to NFL people who do analytics) lean on the side of the pass rushers. It really doesn’t matter, it’s potato-potahtoe, because we’re still at the “stone tools and axes” stage of teasing out if the pass rusher caused an interception in our statistics.
If you dig deep into it, it looks like the secondary is the place to put your resources, except that your CBs have way up-and-down kind of play and your edge rushers tend to be pretty consistent year to year. A Washington example is Josh Norman, who you guys brought in a few years ago and looked like an All Pro, Revis-level player and was very fluctuating and dependent on the scheme and had some bad games as well as good games. At edge rusher, you know Chase Young is either going to get the sacks or get triple teamed and someone else like Sweat gets the sacks. So while the secondary is more important, it might be more trustworthy to put draft picks, money, etc into the edge rushers, try to kill the QB, and the CBs will be okay because they only have to cover for a few seconds.
HH: Switching gears to the Washington Football Team, they are projected to have 8.1 wins this year, which will be a losing record with the expanded season. How much variance does that projection have and what are the most important factors in determining it?
There’s a lot of variance this year, particularly for Washington. More variance than I’d expect for a defense-oriented team, because a lot of times for a defensive team you're projecting the final score’s going to be 20-17, something like that.
HH: But if it’s lower-scoring games, then any score changes the outcome a lot more.
Right. Now your chances of being what we call Super Bowl Contenders (defined as 12+ wins) is an 11% chance. An 11% chance of a 12+ win season is better than the Giants and Eagles. Your chances of 0-5 wins is 18%, so that’s kind of worrisome.
One source of volatility is your QB. He’s a new QB [to the offense] and he’s a legendarily volatile QB. Another source is the division itself. In other divisions... the chances of the Broncos overcoming the Chiefs are not high - in the AFC North you know the Bengals are at the bottom of that division. We’re pretty certain the Cowboys are going to win this division, though there are questions there, and we think the Giants and Eagles could be all over Creation and so could Washington. That creates situations in the simulation where a 9-8 or 8-9 team wins the division and you have to account for all that.
HH: In general, what do you feel very confident in forecasting for Washington’s upcoming season and what would you be very hesitant to predict based on the unknowns?
Defense, defense, defense, we really feel confident in them. We said a few minutes ago that defense is volatile and changes year to year, but there’s not much evidence that there’s going to be a big change here. I believe we have them ranked 7th overall.
HH: They were 3rd last year, right?
Yeah, we have a decline, but a decline to 7, so a modest decline. Sometimes when you’re doing statistics, the rule of central tendency says teams that are great fall back, teams that are bad pop up.
HH: And sometimes other teams just got a lot better, right?
Right, you didn’t have the resources to make those [defensive] improvements because you devoted those resources to your offense.
I hesitate to make lots of predictions about the passing game because we’ve spent the last 13 years wondering which Ryan Fitzmagic shows up. Factor in that we don’t know what this offense really looks like because last year it looked like an emergency situation, and I think our projection is 28th, it’s real low for offense. But if you tell me a veteran QB distributing the ball to these weapons takes them up to 15th, I could definitely see that as a possibility.
Mike was generous enough to give me 30 minutes of his time, so I broke this interview up into two halves. This is the first half and I’ll publish the second as soon as I am able (hopefully tomorrow). I want to thank Mike Tanier of Football Outsiders for taking the time to do this interview with me and to thank Football Outsiders for setting it up.
I’d also like to apologize for taking so long to publish this interview. Mike and I originally spoke towards the end of July. I started writing this article afterwards and just needed to add a few graphics before publishing, when (a couple of weeks early) my wife went into labor with our first child (a baby girl). It’s been a busy couple of weeks since then and I only now feel like I’m in a position to return to publishing articles (which I can work on while the baby and wife sleep, whatever time of day that may be).
Poll
What do you think is most predictive stat in determining which team will win a game?
This poll is closed
-
15%
Passing DVOA
-
11%
Pass defense DVOA
-
2%
Rushing DVOA
-
4%
Rush defense DVOA
-
39%
Turnover ratio
-
23%
3rd down conversion efficiency
-
3%
Something else (describe in comments)
Poll
Which position group would you invest the most resources into when building a defense?
This poll is closed
-
54%
Edge rusher
-
14%
Interior defensive line
-
7%
Off-ball linebacker
-
21%
Cornerback
-
1%
Safety