Fantasy Schedule Projections

by Aaron Schatz

This question showed up in the mailbag a few days ago:

Andy: I've been a subscriber the last couple of years with Football Outsiders and while the end of season info and stat analysts is spot on for my playoffs, one thing I'd love to see you guys do is provide better preseason defensive rankings.

By rankings what I mean, I'd love to see DVOA projections for each team for BOTH rushing and passing defense. The combined/overall defensive DVOA is not as helpful. Sometimes you need to plan your RBs, WRs or QBs by committee, so you need to know what weeks are best to get which team's players to cover the holes you have with existing players in your draft.

Do you anticipate breaking down Defensive DVOA into separate rush and pass defensive numbers? It would really be helpful to many of us out there.

Andy is in luck. I actually did this project for a recent issue of ESPN the Magazine, the one with David Beckham on the cover. That project was to get the easiest and hardest fantasy schedules. From what I gather, you want straight out the projected fantasy points allowed per game for each team, yes? I can provide that too.

I took the same colossal sheet filled with variables that I use for the defensive DVOA projections, and then ran regressions and played with ideas that would project fantasy rushing points allowed and fantasy passing points allowed instead of team defensive DVOA.

First, here are the projections in terms of points allowed per game, ranked from the best defense to the worst:

The thing I want to drive home about this table is that offense is more consistent from year to year than defense is, and pass defense is more consistent from year to year than run defense is. Remember how impossible it was to run on Minnesota last year? They aren't even projected among the top 10 fantasy run defenses for 2007, and I don't believe they were one of the top 10 fantasy run defenses in 2005 either.

I have absolutely no idea how Atlanta ends up with such a high projection as a fantasy run defense, considering that they'll probably be losing late in most games and nobody really considers their front seven to be that special. But hey, that's how the projection ended up. It's probably a mistake, but you never know -- remember, that's what we thought about the Chargers business back in 2004.

Now, let's look at each team's schedule. I'm doing this two ways. First, the average fantasy points per game against the team's 15 opponents in Weeks 1-16. We're leaving out Week 17, since most fantasy leagues don't include the last week of the year. The second number is the same, except it counts the opponent in Week 14 double, the opponent in Week 15 triple, and the opponent in Week 16 quadruple, giving more importance to the teams your players will face during your fantasy football playoffs. In case you were wondering one reason why Mike Tanier and I took both Maurice Jones-Drew and Fred Taylor for our team in that experts draft...

Oakland is interesting. They play Indianapolis in Week 15, but Jacksonville in Week 16. If you want to make it all the way to the fantasy football finals and then lose, draft LaMont Jordan. Here's passing:

Alex Smith fever! Do I have any takers? It's strange how the teams with the easiest passing schedules don't actually have quarterbacks anyone would want to draft in their fantasy leagues.

Let it be noted that these projections were done fairly quickly, and are not perfect. No projections are perfect, of course. On that note, while I'm talking about projections, I wanted to point out something I wrote on the new fantasy football open discussion board [1]. Normally, I don't plan on responding to stuff in there -- I would prefer to be e-mailed directly -- but I thought this was an important point.

The issue was the high projection for Donovan McNabb. If this projection is wrong, does it mean the tool is flawed? Here was my response:

OF COURSE the tool is flawed.

All tools are flawed. All of these statistical projection systems are flawed. Ours, and everyone else's. When I throw 25 variables in there to project a quarterback's performance, I can guarantee you there are at least 25 variables that would improve the projection but either a) I have not even thought of them yet, or b) we have no way of collecting the data in an orderly, objective fashion.

Every year, we seem to project the St. Louis offense too low. I've tried a ton of variables to fix that. Nothing has worked yet. It's possible St. Louis has just enjoyed a couple of lucky years, and there's nothing wrong with the projection systems. It is also possible we just haven't figured out what is the missing "X" variable that would solve our St. Louis offensive issue.

(Note: This doesn't change the fact that the Rams defense sucks.)

The team projections and the KUBIAK projections both improve each year as we gather new data and consider new ideas. That's also why you will see FO people, including me, say things like "yeah, that number is weird, we don't quite agree with that one." A good example of this: I know the KUBIAK projections have Frank Gore ahead of LaDainian Tomlinson, but if I had the first pick in a fantasy draft, I would take LT, just like everyone else.

It's something to remember when you see a projection that doesn't seem to make sense -- like the projection above for the Atlanta defense vs. fantasy running backs. It's possible that the projection system is picking up something most people don't understand. It's also possible that we're missing a variable that would improve things. If you think there's a pattern of weird projections and you can figure out what variable they all have in common, feel free to e-mail me and suggest something for me to try in the projection systems. I'm all ears.