Weekly Game Probabilities

Weekly game probabilities are available now at the nytimes.com Fifth Down. I take a detailed look at the NE-NYJ match-up. The Patriots are actually a much better running team than most think, and the Jets are better at passing than they are running so far this season.

  • 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

7 Responses to “Weekly Game Probabilities”

  1. James Sinclair says:

    First of all, let me state for the record that I have a ton of respect for your game probability model, and I'm not trying to poke holes in it, or accuse the objective data of being somehow biased against my favorite team, or anything like that—I just think this is worth pointing out.

    Here are the Falcons results since probabilities were first posted in week 4, with their WP for that game in parentheses:

    Week 4: at Seattle, 30-28 win (.41)
    Week 5: vs. Green Bay, 14-25 loss (.33)
    Week 6: vs. Carolina, 31-17 win (.44)
    Week 7: at Detroit, 23-16 win (.23)
    Week 9: at Indianapolis, 31-7 win (.45)

    They've gone 4-1, despite being the underdog in every game (as they are again this week, .45 WP vs. New Orleans). Combining this with last year's results (including the playoff game, as much as I'd rather pretend it never happened), since week 4 of 2010 the Falcons have gone 6-1 as favorites and, more interestingly, 9-3 as underdogs.

    I know you've written before about the Falcons' tendency to defy the odds, and I've seen it discussed elsewhere, and the prevailing opinion seems to be that they've had a lot of good luck and are probably not as good as their record indicates. I'm sure that's at least a little bit true, but the longer this "luck" continues, the less likely it is that there isn't some other explanation.

  2. Anonymous says:

    I was going to post a similar comment a few weeks ago too. Is there some reason the Falcons are performing so well despite being low in the rankings? They're climbing their way up now, but it wasn't long ago that they were ranked dead last yet continued to pull out wins.

    I'm wondering if there's some hidden factor that isn't included in the WPA model where the Falcons are somehow excelling. Or maybe it's just random chance. In any case, it seems totally insane for any team to be listed as underdogs (however slight) against the Colts.

  3. Anonymous says:

    In Atlanta's case I think the sample size is still too small to say it can't be chance. Its probably a combination of Atlanta having great luck with there not being enough data in the earlier weeks for their strength to be accurately quantified. It could also be they've actually gotten better every week, but I don't know how often that really happens.

  4. Eric says:

    I have a theory about the Falcons (and other teams like them), but the Falcons are the most extreme example.

    The Falcons tend to not pick up huge chunks of yardage at a time. Instead, they move slowly and methodically down the field running the ball and using the short passing game. As a result, their yards/play is low compared to other teams.

    I have my own model that differs in some ways from Brian's model. I use strength of schedule also and yards/play. Yards/play accounting for strength of schedule for Atlanta has them ranked 26th offensively, which is very poor and also I'd guess a reason they are ranked pretty low in the model.

    Another stat I use is a stat I call first down efficiency. It's different than 3rd down efficiency. It measure the percent of time that a team is able to pickup a first down (or touchdown) starting from 1st and 10 or 1st and goal. It's a measurement of how successful a team is at moving the chains (or stopping the other team from moving the chains if it's the defense). In this category (again accounting for strength of schedule), Atlanta's offense ranks 9th, way better than their 26th ranking in yards per play.

    Most of the time, if a team ranks high in yards per play, they rank high in 1st down efficiency or vice versa. Atlanta is an exception. Cleveland is also like this ranking 21st in 1st down efficiency and 31st in yards/play.

    Other teams, such as Dallas, are the opposite. They pick up big chunks of yardage, but have a tougher time picking up first downs.

    I added this stat as a proxy for guessing how good a team would be at scoring touchdowns once they get down near the goalline. 3rd down efficiency and red-zone efficiency just have too small of sample sizes.

    Not sure if Brian uses this type of stat, and that could explain the issue with the Atlanta Falcons. My model currently has them ranked 12th (#14 on offense and #10 on defense).

  5. Steve Freeman says:

    The sample size is *not* too small. Even if a team’s expected win probability is 50%, the likelihood of winning 15 out of 19 games is less than 1%. The scientific standard for saying something is not luck or chance is any likelihood up to 5%.

    Given that they were underdogs in 12 out of the 19 games, I’m figuring that Brian’s estimated win probability for Atlanta averaged only about 0.45 in those games. And given that assumption, the likelihood of winning 15 out of 19 games is only .0025 (¼ of 1%).

    I am astounded by the intelligence and attention to detail that has gone into Brian’s models and explanations, but having seen the first half of the Atl-Ind game, I feel it safe to say that any model predicting that Atlanta would likely lose has a major hole in it.

    The only Atlanta games I’ve seen over the past two years was this one and the playoff dismantling by Green Bay, so I couldn’t even begin to explain what that hole is, but perhaps, James, you have some thoughts on it?

  6. James Sinclair says:

    My best guess is that the Falcons' play-calling is geared toward getting a lot of first downs, but not necessarily a lot of big plays (someone else posted a comment earlier that made this point better than I could, but it disappeared—not sure what happened there). The typical series in the Mike Smith era goes something like this:
    1st down: 3-yard run by Michael Turner.
    2nd down: incomplete pass.
    3rd down: 8-yard pass to Tony Gonzalez (who's supernaturally good at getting open just past the first down line).

    I'm being facetious, of course, but there really is a lot of that, and it seems to work (though I'm not the only fan who wouldn't mind seeing things open up a little). The Falcons are below average in yards per play, but among the league leaders in time of possession, offensive plays per game, first downs per game, and third down conversion percentage, all of which seems consistent with a team that takes its time moving down the field, but does so effectively nonetheless.

  7. Anonymous says:

    Steve, So there is a 1 in 400 chance that Atlanta is not doing something special that is being ignored by the model? The other way to look at it is to see how many times this has happened to the model overall. That is, you can find 400 sets of 19 games and check how many times such a model-defying outcome has been observed. My guess is that the answer is "one." I.e., if you are hunting through enough observations, you are likely to see a 1 in 400 event occur by chance alone.

Leave a Reply

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