Dear Bill:
I am one of your millions of fans. I love your columns and podcasts. You inform and entertain me several times a week. We share a passionate love affair with Friday Night Lights and Mad Men. I was amazed when our school PR manager Lura Forcum informed me that you had mentioned me in your column. Since you do not seem to think much of my basketball statistical analysis I thought I should write a response to your column. So here goes.
Should Belichick have gone for the first down?
You clearly think this was a bad decision. I think the decision is not clear cut, but math can shed a lot of light on this controversy. You quote several people on advanced topics like WIN PROBABILITY. Actually, things are not that complex. I think you agree with me that the decision depends on your estimate of three numbers.
MAKE = Chance Pats make first down
LONG = Chance Colts score TD after a punt.
SHORT = Chance Colts score TD after a failed first down attempt.
Intelligent people can differ in their estimates of these parameters. Whatever your parameter estimates, however, the math makes it pretty clear that the correct decision is to go for it if and only if
MAKE + LONG/SHORT>1.
See my post on this earlier this week. I estimate MAKE = .5, LONG = .5 and SHORT =.8 so the formula says to go for it. I polled 10 college professors on this issue and based on their estimates going for it was the right decision for 70% of the professors.
You point out that less than 40% of two point conversions attempted with passes were successful, so you clearly think that MAKE<.50. I would argue, however, that the short field makes it easy to defend a pass on a two point conversion, so this is not a relevant data point. During the last three years, teams have converted 4th and 2 passing attempts more than 55% of the time. Clearly the Pats have a great passing game, so I think a MAKE = .50 is reasonable, but people can differ on this. Clearly, however, Belichick’s decision to go for it was not as crazy or irrational as many people think.
Next you point out that only around one time a year does a team score 3 TD’s in the 4th quarter and win a game. This statistic has no relevance to the question at hand. What is relevant is the chance that a team which trailed by 18-20 points in the 4th quarter that has scored 2 TD’s in a row will score a TD if they get the ball with 2 minutes left. I bet there isn’t much data on that so you have to go with your gut feel here.
In summary, reasonable people can come down either way on the question of whether the Pats should have gone for it.
Let’s move on to basketball. Love your book by the way. Its status as the #1 best seller is well deserved. I totally agree with you that stats are much more useful in baseball than basketball. This is because baseball is mostly a two man game between the pitcher and hitter (fielding does matter, but it now can be well measured). I also agree with your comments in the book that Box Score metrics like PER miss a lot of the game. I love your discussion of why we need a stat that shows that Wes Unseld was a great player(“he made his teammates better in so many ways”). While over short spans of time Adjusted +/- has a lot of noise in it, over a large sample, Adjusted +/- gets at how a player adds (or subtracts) to a team’s success. Essentially Adjusted +/- looks at every minute of every game and uses how the score moves, together with the 10 players on the court, to tease out the influence of each player on the score of the game So now let’s look at the two assertions you made on your blog:
· Anyone who thinks KD was a below average NBA player must be pretty stupid.
· Anybody who thinks Tim Thomas is underrated must be pretty stupid.
As Edwards Demings, the great American statistician used to say “In God we trust, all others need data.” So let’s look at your two assertions.
How did Kevin Durant perform during his First Two Seasons?
KD had a PER of 24 last year, which indicates to box score followers that he was a great player. I do not think this was the case. His Adjusted +/- was -7 points for his first two seasons (other people get just about the same number). This indicates that after we adjust for who KD played with and against; our best estimate is that he reduced the Thunder’s performance by 7 points a game. After factoring in the noise, there is less than a 5% chance that KD’s Adjusted +/- for his first two seasons exceeded 0 (an average NBA player). By the way ask your good friend Daryl Morey if he thinks it is “nonsense” to say KD hurt the Thunder during his first two seasons. To substantiate the fact that KD hurt rather than helped the team during his first two years look at the following numbers (standard deviation measures the “noise”).
Let’s break down all Thunder minutes during 2008-2009 into 3 lineup combinations and look at how (adjusting for strength of opposition) the Thunder played. The standard deviation of these estimates (rounded off) is also given
· Collison Westbrook Green and KD in +.4 points (std dev 4 points)
· All other KD minutes -11.2 points (st dev 2 points)
· All minutes with KD out -2.6 points (std dev 3 points
I hope you can enlighten me on how these numbers show that KD helped the Thunder win a lot of games.
By the way right before KD was injured last year the Thunder was on a 7 game losing streak. As soon as he got hurt they went 5-2. I am sure you think that was a coincidence.
Now the good news is that KD (and the Thunder) are playing great this year. Our numbers indicate that KD is an all star caliber player this year. Could this be because Henry Abbott in his gutsy TRUE HOOP column pointed out KD’s shortcomings (failing to play the pick and roll correctly and shooting when doubled and tripled teamed)? Maybe KD corrected these flaws in his game and this led to his improvement. If you can ask KD if this is the case, I think it might make a great story. By the way to improve a team’s performance by 20 points a game, you need only score one more basket or give up one less basket on around one out of 20 possessions (assuming 200 possessions per game).
On to Tim Thomas. Both you and Kevin Pelton (his Basketball Prospectus is also a great book) have written that Tim Thomas “defines a replacement (read bad) player.” I do not dispute your comments that Thomas did not hustle in many Clipper games. He may be a bad off court influence; I just do not know. But in over 20,000 minutes of NBA play his adjusted +/- rating is +0.78, so he grades out over his career as a slightly above average NBA player. This is an awful lot of data. Basically, for all of Tim Thomas’ minutes we know who he played with and against and how the score of the game changed during each of his minutes on the court. Given this data, there is virtually no chance that Tim Thomas is as bad as you say he is (against ask Daryl Morey his opinion on this). Of course, if he hustled more he would have been better. One final stat on Tim Thomas’ performance with the Knicks last year:
In 105 minutes with Gallinari, Harrington and Thomas in the Knicks played 22 points better than an average NBA team. In 140 minutes with Gallinari and Harrington in and Thomas out the Knicks only played 3 points better than average.
I think it should clear to a basketball expert like yourself that Tim Thomas let’s you spread the floor (he also did this with the Suns) and when he is on the court with guys who can shoot and drive he can make a team awfully hard to defend. This data certainly flies in the face of your assertion that Tim Thomas is worthless.
Well, thanks for reading this. I emailed you last year when you dissed our rating of Jason Kidd. I wanted to show you why we had him rated as a good player, but I can appreciate how busy you are. Hope to hear from you sometime ([email protected] or 812-322-4270).
Sincerely yours,
Wayne Winston
