Garbage Time Challenge
Dr. Z looked over my Stat Power Index at the draft and opined with some suggestions.
Most important, he thinks garbage time stats have to be taken out of the equation. Sounds easy, right? In theory, maybe. But what's the universal point differential that signifies garbage time? And at what point in the game?
Z thought that down three scores in the fourth quarter would work. But there are teams that have come back from down three scores in the fourth quarter to win. Forget the Miracle at the Meadowlands. The other Monday Night Miracle, when the Colts visited the Bucs, saw Indy win despite being down 21 points with five minutes and change left in the game.
We're going to get into this more extensively later, when I get my thoughts together and send my opinion on the matter to the good Doctor. But our friend Bill Krasker at footballcommentary.com has a dynamic programming model that calculates the probability of winning during all game situations in order to assess coaching strategy.
He wrote to me in an e-mail, "If you fall behind by 21 points with 13:40 to go in the game, your probability of winning is just 0.005. If you fall behind by 21 points with 3:20 to go in the 3rd quarter, your probability of winning is 0.01. I can of course produce other probabilities; you just have to be specific about what you would like to see.
So then the question is, what win probability corresponds to 'garbage time.' Is it when the team brings in the subs and stops trying to win? Or is it when they are so far behind that, even though they're still trying to win, their best chance is to implement a drastic departure from their normal strategy (e.g. by passing on every down)?"
By the way, Krasker's model said the Colts probability of beating the Bucs in Tampa that night down 21 points with 5:09 left was .0002. He does think the model understates the benefits of the hurry up offense, so the number, he says, was probably closer to .001 (in other words, one team out of a thousand down that amount with that much time left will win).
Z says you have to find a number that no team could possibly come back from. But if minus-21 with five minutes left doesn't make the cut, what possibly could? If we're searching for loss certainty, garbage time will never exist. I think .01 is good enough. I mean, if it's good enough for birth control companies, it should be good enough for the SPI.
Again, more on this to come. END.
Most important, he thinks garbage time stats have to be taken out of the equation. Sounds easy, right? In theory, maybe. But what's the universal point differential that signifies garbage time? And at what point in the game?
Z thought that down three scores in the fourth quarter would work. But there are teams that have come back from down three scores in the fourth quarter to win. Forget the Miracle at the Meadowlands. The other Monday Night Miracle, when the Colts visited the Bucs, saw Indy win despite being down 21 points with five minutes and change left in the game.
We're going to get into this more extensively later, when I get my thoughts together and send my opinion on the matter to the good Doctor. But our friend Bill Krasker at footballcommentary.com has a dynamic programming model that calculates the probability of winning during all game situations in order to assess coaching strategy.
He wrote to me in an e-mail, "If you fall behind by 21 points with 13:40 to go in the game, your probability of winning is just 0.005. If you fall behind by 21 points with 3:20 to go in the 3rd quarter, your probability of winning is 0.01. I can of course produce other probabilities; you just have to be specific about what you would like to see.
So then the question is, what win probability corresponds to 'garbage time.' Is it when the team brings in the subs and stops trying to win? Or is it when they are so far behind that, even though they're still trying to win, their best chance is to implement a drastic departure from their normal strategy (e.g. by passing on every down)?"
By the way, Krasker's model said the Colts probability of beating the Bucs in Tampa that night down 21 points with 5:09 left was .0002. He does think the model understates the benefits of the hurry up offense, so the number, he says, was probably closer to .001 (in other words, one team out of a thousand down that amount with that much time left will win).
Z says you have to find a number that no team could possibly come back from. But if minus-21 with five minutes left doesn't make the cut, what possibly could? If we're searching for loss certainty, garbage time will never exist. I think .01 is good enough. I mean, if it's good enough for birth control companies, it should be good enough for the SPI.
Again, more on this to come. END.
Entry #111552114295829532 posted by Catherine Salfino at 10:40 PM
<< Home