Roundup 2/6

How much of a game's three hours is actual football action? David Biderman tells us, and you'll be surprised. Hat tip - PFR.

Jonathan Adler of Harvard Sports Analysis finds that outlawing "the wedge" has made little to no impact on kick returns.

Kurt Warner's total passing yards by season. A real roller coaster. Nothing earth-shattering here; I just love graphs. Give me the choice between a 3,000-word essay or a clear 500 x 500 pixel graph, and I'll take the graph any day.

The Fantasy Football Librarian grades the fantasy ranking sites. Here's my take-away: None of these sites are any better than the others. It's luck. None of them are consistently good across multiple positions, and none of them rank near the top in multiple years. Nothing against stat-based web sites (I happen to be very fond of them), but they've all got the same fantasy information.

Advanced stats for Golf.

Chase Stuart points out just how pass-oriented the Colts offense is.

A great discussion of the NFL's All-Decade Team by Chase. Maybe I'll do an 'Decade All-WPA Team' post at some point. (Gosh, I wonder who the QB would be?)

Carl Bialik shows how regression to the mean operates at the team-level in college football. I wouldn't expect regression to be that strong, at least among the top teams. Because college teams recruit their players, and the best players tend to choose schools that are already big winners, it creates a 'rich-get-richer' system. There are none of the constraints that pro teams have, such as salary caps and free agency. The feedback loop is strongly positive (a team wins, gets better players, wins even more). Regression and randomness play a role in mitigating that effect, but I'd guess it's relatively small in college. That's why we keep seeing the same teams at the top of the rankings for several-year stretches. It's also why we get a power law distribution of BCS bowl appearances.

'ZEUS' diagreed with my analysis on Norv Turner's onside kick attempt at the end of the Chargers' loss to the Jets. ZEUS says there was a small advantage, and I found that kicking away would be the percentage play by the slightest of margins. I'm struck by how infrequently the various methods for modeling football disagree on controversial decisions.

A great new feature at PFR--league totals. It's like they make this stuff just for me.

Bill Krasker of asks whether we can assume conversion rates on 3rd down are valid replacements for 4th down rates. (Caution, very mathy.) The answer is no, we can't assume it. But to me, the question is whether they are equivalent enough for practical purposes. I encourage everyone to peruse Bill's articles. There's a lot of great stuff.

Football Outsiders has been posting a 'stat of the day' for the past couple weeks, and some of them are pretty interesting. Defensive lineman stop rate vs. the run. Quarterbacks who get hit the most. Teams with bad luck on 3rd down. Defenders with the most QB hurries. The most penalties for players and teams of 2009.

Unfortunately, there's also this write-up on Adjusted Games Lost. It seems like a clever, useful idea at first--tallying teams' reported injuries weighted by how bad the injury is (probable, questionable, etc.). But if I taught an intro applied statistics course, I'd use this as the poster-child of how not to construct research. I'd say, "Ok class, let's list all the errors, fallacies, and other problems with this on the board." Data-mining for type I errors, small-sample fallacy, widely varying injury-reporting doctrine by teams, correlation-causation fallacy...Some of their comments have additional criticisms.

Could it be that some teams regularly report half their players as probable or questionable regardless of true injury status? And that some teams don't disclose injuries that they should? Also, once it becomes apparent a team is not in the playoff hunt, many players will suddenly be held out or put on IR. So there is a "causation back-flow" where being a bad team creates injuries. Adjusted Games Lost is like a bad research pinata. It's fine to do something like AGL if you address its shortcomings up front, but claiming they've discovered that some teams are tangibly better at keeping players healthy is an overreach.

A great discussion by Phil Birnbaum on why baseball fans tolerate a single team that spends 2 1/2 times the average payroll on talent. In my mind, part of the answer is survivorship. The people who it does bother aren't big baseball fans anymore, or they never became fans in the first place. That includes this former die-hard MLB fan.

Somewhat on the same topic, Neil Paine of Sports-Reference asks, what if the NBA had as short a season as the NFL? Hat tip: Daily Fix.

Ian Stanczyk and David Gassko look at how likely a cow is to make it to the Super Bowl. Hat tip: Freakonomics.

Site note: Don't forget about the live Win Probability graph for Sunday's game. I've added a live comment/chat feature, so we can discuss strategy decisions, blown calls, and everything else in real-time. If the server can handle the traffic, it should be cool.

  • Spread The Love
  • Digg This Post
  • Tweet This Post
  • Stumble This Post
  • Submit This Post To Delicious
  • Submit This Post To Reddit
  • Submit This Post To Mixx

3 Responses to “Roundup 2/6”

  1. Brian says:

    Interesting note from the PFR league totals. . . it looks like the number of interceptions per game has dropped substantially, while the number of pass attempts has increased. Could this help explain the current over-reliance on the running game? Perhaps a holdover from an era in which passing was rightfully considered to be a much riskier option?

  2. John Morgan says:

    Speaking of interceptions, have you run a comparison of runs and passes using your new baseline for the value of an interception? I noticed you're still using 45 instead of 60.

  3. bytebodger says:

    "I wouldn't expect regression to be that strong, at least among the top teams."

    I guess that depends on what you mean by "that strong". If you're thinking that top college teams don't become bottom-dwellers and vice versa, then yes, I agree with you. But if we accept that the top teams are essentially working from their own baseline, then I think it is pretty clear to see regression-to-the-mean taking place amongst them.

    The top teams probably have a mean of ~10 wins per season and, when they're "up" they get 12-13 wins and when they're "down" they get 7-8. After all, a 7-5 season is considered quite disastrous at a football powerhouse like USC or Florida. But the reinforcing feedback loop of college football's recruiting system ensures that these top teams rarely ever fall completely out of that top tier.

    I've also noticed that college football's regression tends to follow longer time cycles. A team like Alabama that won it all this year will almost certainly NOT regress back to, say, 8-4 next season. But it's not hard at all to imagine that they could be in such a situation in 3-4 years.

    A recent example that comes to mind is LSU. They had several several consecutive championship-caliber seasons. Now they are, by LSU standards, in the doldrums. In 3-4 seasons they could be right back at the championship level. But they never really get "bad". They never fall out of that top tier of college teams.

    This probably makes sense because most impact players in college are starting for 2-4 years and a program's surge could be defined by having 2-or-more consecutive superior recruiting classes. This means that top programs probably experience windows of 2-4 years when they have their best shot at winning a title.

Leave a Reply

Note: Only a member of this blog may post a comment.