Joined: 03 Jan 2005 Posts: 460 Location: Washington, DC
Posted: Wed Nov 22, 2006 4:54 pm Post subject:
What's kind of amusing about Gladwell's parting shot at...well...us, is that most of us have been interested in using stats to learn truths about the game. I think that some of the comments and criticisms Berri has received fall into the "defensive" category Gladwell describes, but others think that Berri has relied on flawed assumptions about the game, and has therefore gotten some stuff wrong. _________________ If you can't explain it simply, you don't understand it well enough.
The league adjustment came to mind when I first read Berri's post. The one thing I would point out is, if I'm understanding PER correctly, it's possible to be league average in everything else, shoot slightly below league average and still have a 15.0 PER thanks to your usage rate. My assumption is that the "break-even point" is a little more complicated to figure out.
John, if you're reading, correct me if I'm wrong on that. (Or anyone else.)
PER = (0.440228848*TS%)+(0.154933814*AstR)+(-0.306420306*ToR)+(0.605165609*Usg)+(0.537938584*RebR)-24.01548294
It works to within +/- 0.97 PER. Not sure if that will help you answer your question, Kevin, but it does tell you how many units of Usage are needed to offset a drop in TS%, for instance.
Interesting stuff. What might be even better is to see the standardized numbers so we know we're comparing like units.
My PER spreadsheet is definitely off, so I'd love for someone else to look at this, but I found that increasing two-point attempts increased PER even at startlingly low two-point percentages - even lower than the break-even point Berri referenced.
I think that it is a good thing that different statistical approaches and the people who made them decide to go to "battle". With words that is, in a civilized way. (I must be carefull when I use the words going to battle on an American board, before I know it you guys invade Belgium ) It increases the chances that further progress will be made.
I don't know if Hollinger his post on Truehoop will still be visible in a couple of days/weeks. So perhaps I can quote it: "Berri leads off with a huge misunderstanding of PER -- that the credits and debits it gives for making and missing shots equate to a "break-even" shooting mark of 30.4% on 2-point shots. He made this assumption because he forgot that PER is calibrated against the rest of the league at the end of the formula. Actually, if we took a player was completely average in every other respect for the 2006-07 season -- rebounds, free throws, assists, turnovers, etc. -- and gave him a league-average rate of shots, and all
of them were 2-pointers, and he shot 30.4%, he'd end up with a PER of 7.18. As long-time PER fans know, that would make him considerably worse than nearly every player in the league.To end up with a league-aveage PER of 15,00, the actual break-even mark in this case is 48.5%, which is exactly what the league average is on 2-point shots this season. Secondly, regarding Gladwell's comments on Berri's ranking of players like Ben Wallace and Dennis Rodman (as well as several other less famous players who also got piles of rebounds), the issue isn't the discovery of some new Holy Grail of basketball analysis, but rather the result of an assumption in Berri's formula that distorts their value. Namely, that his formula equated a rebound as being just as valuable as a steal, and a missed shot just as damaging as a turnover. While this made his sums work in the aggregate, for individuals it threw things way out of kilter, as it overrewarded rebounders and overpenalized guys who miss shots.And it seems clearly mistaken, both because missed shots can be rebounded while turnovers can't, and because a defensive rebound is merely the completing piece of a sequence that began by forcing a missed shot
Last edited by Analyze This on Thu Nov 23, 2006 2:27 am; edited 2 times in total
It seems that Hollinger didn't address the main crux of Berri's critique of PER. A player who is very inefficient but still above the "break even" point (as Berri defines it), could increase his PER by taking more shots with shooting efficiency staying constant.
For instance, suppose there's a 35% shooter who shoots 15 times a game. If If he can hold that 35% shooting constant, he can increase his PER just by increasing his shot attempts. In reality, if such a player did that he'd only be hurting his team more as he takes more shots. But according to PER, he'd be a better player.
Correct me if I'm wrong, but I believe that's what Berri is getting at. It seems to me to be a very good point.
Joined: 14 Jan 2005 Posts: 971 Location: Delphi, Indiana
Posted: Thu Nov 23, 2006 8:33 am Post subject:
When a player takes a 30% shot, there's about a .29*.70 = 20% chance of an offensive rebound on the play. So the odds are about 50/50 that it's a 'good' play.
When the typical 30% shot is taking place, it's often because there isn't a better shot in the allotted time. Passing up the low-% shot for a possibly-slightly-better shot doesn't necessarily make sense. Factor in the possibility that the pass doesn't get thru, or the clock expires, etc.
Many low-% shots are with the hope that a foul will be called. When Paul Pierce goes 5-19 from the floor and 13-15 from the line, some may ask, "Why foul a guy who shoots 26% ?" But if you saw the game, you'd see that many of his FGA were in heavy traffic -- as were most of his fouls received.
I'd like to see players % of missed FG which become OffReb. Maybe 82games already has that. There's a world of difference between firing a quick outside shot with nobody underneath, and doing the same with rebounders poised. Or driving and getting fouls a good bit of the time.
In game-time, only the coach can act on whether a player's low FG% is due to poor shot selection or execution of a strategy. Obviously, there are often better options than a 30% shot; but often the alternative is no shot. _________________ 40% of all statistics are wrong.
Am I alone in being confused by Berri's continued insistence that the only legitimate criticism of his model is to create a superior one? From his response to John's response today:
Quote:
And what makes a model “better”? A model would be preferred if it explains wins better and provides better forecasts of future player and team performance. Explanatory and predictive power are not the only criteria we have in evaluating models. But they are on the list. And this list does not include whether or not the model conforms to your prior beliefs.
While I have no doubt that Wins Produced would show as "better" than PER in such a comparison, has Berri ever provided this evidence?
Above and beyond that, if Berri had any evidence that demonstrates Wins Produced is superior to DeanO's Win-Loss records -- which dynamically value rebounds, as opposed to the strict approach taken by Wins Produced -- I would be very surprised, given that DeanO's ratings explain a team's wins or losses just as well as Wins Produced does.
Does this strike anyone else as hypocritical? Or have I just missed something along the way?
Joined: 03 Jan 2005 Posts: 413 Location: Greensboro, North Carolina
Posted: Fri Nov 24, 2006 4:54 pm Post subject:
I have never programmed PER up myself, but using PER values that Justin Kubatko gave me a couple years ago, I can regress PER onto true shot attempts, points, offensive rebounds, defensive rebounds, etc. And when I do, it seems to me that the break-even point for true shooting percentage is around 30 percent. So it appears to me that Dave Berri is right on this point in his discussion of PER.
I agree with much of what Hollinger writes in his e-mail, but his thought experiment just does not address the point that Dave Berri brings up. Suppose a player was average in every other respect, but shot 10 more shots per game than the average player and his true shooting percentage was 48 percent (which is below average). With Wins Produced this player would be rated below average (and farther below average with every additional shot). With PER I think he would be rated higher than average (and higher with every additional shot).
Joined: 14 Jan 2005 Posts: 971 Location: Delphi, Indiana
Posted: Fri Nov 24, 2006 5:56 pm Post subject:
Dan Rosenbaum wrote:
... With PER I think he would be rated higher than average (and higher with every additional shot).
Taking 48%-makeable shots (or 30% shots) is frequently justifiable. Of the 52% which are missed, about 29% are OffReb. Add the probablilities: (.52*.29 = .15) .15 + .48 = .63
When taking a 52% shot, the odds of a non-negative outcome are 14%+52% = 66%
Even a 30% shot leaves you with a 50% OK result. It isn't 22% lower than the 52% shot; it's only 13% lower.
30/52 = 58% ; but 50/66 = 76%
So the 'weak' shot could be considered 3/4 (or more) as good as the 'average' shot. It may depend on whether your team is ahead, time left in the quarter, etc. If more FT are generated, it could even be better. In any case, it's effectively a lot closer to the better shot than it is to No shot. _________________ 40% of all statistics are wrong.
Joined: 03 Jan 2005 Posts: 413 Location: Greensboro, North Carolina
Posted: Fri Nov 24, 2006 6:05 pm Post subject:
admin wrote:
Am I alone in being confused by Berri's continued insistence that the only legitimate criticism of his model is to create a superior one? From his response to John's response today:
Quote:
And what makes a model “better”? A model would be preferred if it explains wins better and provides better forecasts of future player and team performance. Explanatory and predictive power are not the only criteria we have in evaluating models. But they are on the list. And this list does not include whether or not the model conforms to your prior beliefs.
While I have no doubt that Wins Produced would show as "better" than PER in such a comparison, has Berri ever provided this evidence?
Above and beyond that, if Berri had any evidence that demonstrates Wins Produced is superior to DeanO's Win-Loss records -- which dynamically value rebounds, as opposed to the strict approach taken by Wins Produced -- I would be very surprised, given that DeanO's ratings explain a team's wins or losses just as well as Wins Produced does.
Does this strike anyone else as hypocritical? Or have I just missed something along the way?
It is really hard to chalk this up as anything other than hypocritical.
Remember I could use points per game as a metric for player value, and as long as I add in a team adjustment, I would explain team wins just as well as Wins Produced.
So explaining team wins cannot be used to differentiate those two metrics. So why don't we move to something else like how a team does when a particular line-up is on the court? Well, it turns out that points per game does just as good of a job as Wins Produced explaining how teams do when a particular line-up is on the court.
OK, that's not really true. Wins Produced does microscopically better. But points per game is soooo much easier to compute. I don't even need a position adjustment or to divide by minutes played.
So why didn't the authors of Wages of Wins write a book using points per game as their metric? I don't know; you'll have to ask them. Maybe that is the threatened sequel.
So let's review. Wins Produced is a metric that (a) professes to be regression-based, but is only marginally so, (b) misapplies its own logic when it derives (rather than estimates) its linear weights, (c) proposes explaining team wins as a barometer when ANY metric with a team adjustment (no matter how bizarre) would explain team wins just as well, and (d) only performs microscopically better than points per game at explaining how teams do when particular line-ups are on the floor.
Quote:
Commenting on someone else’s work
Hollinger states at the onset of this e-mail “I’ve been trying real hard not to say anything about Wages of Wins, because it would only come across as self-serving to knock the work of another person in the field.”
This statement is disingenuous. In The Wages of Wins we quote extensively from Hollinger’s critique of the plus-minus approach. So Hollinger has addressed issues he has had with other approaches.
And in fact, if you are going to utilize a metric of performance you have to at least address why the available alternatives were not chosen. This is something we discuss in the book when we review NBA Efficiency and plus-minus. We did not offer a discussion of PERs, though, hence the need for the recent post.
Maybe, just maybe the argument above is why John Hollinger felt he could go about his business without explaining to his audience why he has not switched to using Wins Produced as his preferred metric.
Last edited by Dan Rosenbaum on Sat Nov 25, 2006 1:35 am; edited 3 times in total
Joined: 03 Jan 2005 Posts: 413 Location: Greensboro, North Carolina
Posted: Fri Nov 24, 2006 6:11 pm Post subject:
Mike G wrote:
Taking 48%-makeable shots (or 30% shots) is frequently justifiable.
I agree. I think often the alternative to Allen Iverson taking another shot is an attempt to create a shot that runs a high risk of a turnover or really low percentage shot.
I agree with Berri that Hollinger did not really respond to his point, but that is in no way an endorsement of the view that shot creation has no value.
All times are GMT - 5 Hours Goto page Previous1, 2, 3 ... 11, 12, 13
Page 13 of 13
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
)