Natural talent matters more than we'd like to admit.
Is Tim Tebow 'clutch'?
Keep it simple--the secret to Jim Harbaugh's success.
Analytics is the future.
What could'a been--pro basketball in a lost season.
This is exactly how I feel about the game probability model.
Why our brains make us laugh.
Checking in on Total QBR.
This is one of the problems I have with studies like the 'Lewin Career Forecast' for predicting which college QBs would do well in the pros. If you dump a laundry list of 20 variables into a regression model and find that one is significant, chances are it appears significant only by chance (type I error). Plus, there's a really good chance a truly significant variable appears non-significant due to chance (type II error). When you think of it, if this is all your research is based on, you've almost guaranteed misleading results.
On a related note, here's a good discussion of the effect of sample size. The Gates Foundation has literally wasted over a billion of dollars and disrupted the education of thousands of students on an effort to reduce school size. The stats showed that a large number of small schools were very high-performing. The foundation poured enormous resources into breaking up larger high schools into smaller ones before anyone realized the effects of small size were a statistical illusion. The performance of schools with small populations of students will naturally vary more widely than those with large populations. While it was true that a number small schools over-performed, an equally large number under-performed!
SR and Cowboy RBs.
Boy, the Saints really have lost their momentum. Maybe after their bye, they can get it back. It's all about who wants it more.
- Home
- Unlabelled
- Roundup 11/26/11
Roundup 11/26/11
By
Brian Burke
published on 11/26/2011
Subscribe to:
Post Comments (Atom)
Brian, could you do an analysis of when, if ever, it would have made sense for the Dolphins to have let the Cowboys score a TD on their final drive? It seems to me that getting kicked off to down 5 with 1:52 remaining and 2 timeouts would be better than allowing the other team to run out the clock and attempt a short FG as time expires. However, my best efforts with the WP calculator seem to suggest otherwise. This might be a case where the timeouts available is a significant factor which I can't account for on the WP calculator.
It was foolish of the Denver coaching staff to ever let Tebow start, because now they've created a monster. Now the fans are behind him, and so are his teammates. I suspect that they started him in the first place in the hope that he would fall on his face, thus shutting up those who were clamoring for him to play. Now they're going to take a PR hit when they bring in someone else, and if his replacement struggles, things will get even worse.
Tarr: There was a similar situation in a game last year between the Packers and the Bears. I believe the Bears won the game, helped by the fact that the Packers kept stopping them at the goal line near the end of the game instead of just letting them score so they'd have more time to score themselves.
http://content.usatoday.com/communities/thehuddle/post/2010/09/ray-lewis-packers-should-have-let-bears-score-in-final-minutes-to-get-ball-back-on-offense/1
In that game, the Bears ended up scoring the go-ahead FG to make the game 20-17 with just 8 seconds left. Had the Packers let them score a TD, the score would have been 24-17 but the Packers would have gotten the ball back with something like 1:30 left.
In endgame situations like that, I'll be watching the clock tick down and in my mind I'll be imagining WP ticking down at the same time. Sometimes time is worth more than points.
Yeah, the GB-CHI game was an interesting one. GB definitely should have let them in the end zone.
Tarr-I completely missed that. Too busy eating and worrying about the Ravens game, I guess. Glad you caught it.
2 TOs is the typical number remaining in that situation so I wouldn't make any adjustments to what the calculator says. Besides, 1:45 with even a single TO is plenty to get a TD these days.
I would have let them score at the 2-1 from the 20, but not from the 1-10 from from the 29.
Having a 2nd and 1 from the 20 gives DAL a 0.88 WP, and MIA a 0.12 WP. Being down 5 from your own 22 with 1:45 to play is a 0.18 WP. It's a big difference as these things go. The problem is, if Murray were smart, he'd take a knee at the 1 and blow the plan.
Brian, if it were up to you, what of your work would go into the NY Times?
If “those projections don’t have any scientific validity whatsoever" is "exactly" how you feel about your own projections, would it be possible to persuade the NY Times to run material you think has more validity to it?
The Handbooks started in '93, which is about a decade after the Abstracts started making their mark in the mainstream (I bought the '82 Abstract in a book store within the University Co-op on the drag at UT). James had some leeway to have fun with the numbers at that point.
Football analytics is in the early stages of working into the mainstream within these past few seasons, with token presences at ESPN and the NY Times finally coming more than 25 years after I bought my first Abstract. Yet, the "no validity" stuff is what goes into the Times?
What would you prefer most go into that slot? And, what happens when the Times realizes they can just use the "sum of all models" moneylines from the betting markets to calculate win likelihoods? Eventually somebody's going to realize that. (Seems like Bill Simmons is the only guy with a byline who knows how pointspreads and moneylines work at the moment).
Is there a future for football analytics in the media if a prominent current name doesn't believe there's validity to his own projections?
Wow!! A really great roundup this week.
I'd like to comment on several points. On the issue of publishing predictions I can understand your reticence. I believe that this site mentioned Mandelbrot's popular book The (Mis)behaviour of the Markets. It's an excellent read and speaks to the many pitfalls of trying to predict behaviour. Mandelbrot focuses on the stock market but his comments can be applied to performance in any arena. In essence nothing works consistently as a predictor.
On the questions of the Gates Foundation effecting education. The problem is not as simple as school size. It's timely that there is an article about how researchers collect data and massage it. I'm certain that the same research led to the conclusion that changing school size will improve performance. The problem with the educational system is larger and more complex. Changing one variable will do little to improve performance. Teaching methodologies and curriculum are huge factors which would contribute to performance.
On the issue of NO losing coin tosses. Several years ago (perhaps 30) a team called the Oakland Raiders had a designated coin toss caller. Unfortunately I don't remember his name but I do remember that that he had an incredible streak, at least for a while.
Brian - thanks for that. I must have mis-entered the numbers in the calculator, because the 2nd and 1 was exactly the play I was looking at. More specifically, I was thinking that once Murray had made, say, six yards or so on first down, the Dolphin defenders should have let him run the rest of the way uncontested.
So Vince Young was the Tebow of 2006? Making Tebow the next Vince Young ...
How exactly do you "let the other team score"? There is no good strategy that will not backfire. The other team also knows that time is very important.
if your defenders sit down and refuse to play the down, then the offense slowly walks up to the 1 inch line, and just waits while the clock ticks down.
or do you envision some academy award winning performances where defenders dive about 3 feet away from the ball runner in a fake tackle attempt?
An better strategy would maybe be to make some ridiculous risky defenses, like go with a full out blitz and try to get a fumble, make them lose lots of yards, or panic and throw and interception.