Per Minute

How much should we think about a player's per minute stats versus his per game stats? Dave Berri argues:

Let me close by noting that I don’t think that people should solely look at WP48 or just per-minute stats. If you did that, Jerome James - who posted a 1.341 WP48 - would have been the first half MVP. James, though, only played five minutes in the first half of the season, so his WP48 doesn’t really mean much.

Although I do think people need to look at more than per-minute numbers, I also think people need to stop focusing solely on the per-game stats. Specifically, when we are looking at players who played at least 30 minutes a contest, we shouldn’t penalize players whose minutes are closer to 30 than to 40. Such penalties — as we see in the case of KG — can easily cause us to miss the obvious.

I think the players who are playing at least 30 mpg are exactly the players we should penalize for lower minutes. After all, a great player who offers you 32 minutes per game is genuinely less valuable than a great player who offers you 40 minutes per game. Things like stamina, injury resistance, and ability to avoid foul trouble are all part of what makes for a useful player. It's the players who play less than that who we shouldn't penalize. Of course you don't want to rely on tiny samples like in the Jerome James example, but a guy who's playing well in 15 minutes per game is probably limited to 15 mpg by coaching decisions -- decisions that might be wrong, or might indicate a jam-up of good players at the same position one of whom should be traded -- rather than fatigue.

Comments

Matt, are you going to fix the italics on the previous post?

Posted by Italics Whiner | April 6, 2008 2:25 PM

But Matt, look at the Celtics this year. They've now won 22 games by at least 20 points. Garnett isn't averaging 33 minutes a game because of stamina, injuries, or foul trouble. After all, he's averaged 38 minutes per game in each of the past 11 years. The reason Garnett is averaging 33 minutes a game is because his team is doing so well, there's no reason for him to play more than 33 minutes.

I'm not arguing against the points you made as much as I'm arguing against their being some sort of proper black-and-white view of the situation. As with all stats, you've gotta understand what you're looking at for them to mean anything.

Posted by Will | April 6, 2008 2:39 PM

but what good is a 40 mpg player if he is gassed in the playoffs? A coach has to balance game-by-game interests with seasonal interests as well.

Posted by Andrew | April 6, 2008 3:24 PM

"The reason Garnett is averaging 33 minutes a game is because his team is doing so well, there's no reason for him to play more than 33 minutes."

That's a good point, but I really don't think it's at all common for that to occur.

Posted by Christopher Colaninno | April 6, 2008 3:27 PM

Garnett isn't averaging 33 minutes a game because of stamina, injuries, or foul trouble.

Exactly right.

Why on earth does Matthew think that players who are limited to 33 MPG rather than 40 MPG are limited by "stamina, injury resistance, and ability to avoid foul trouble" rather than coaching?

Posted by Al | April 6, 2008 4:12 PM

I don't understand what aspect of per-minute data would prevent one from taking into account MPG. In baseball, we can say with confidence that a .330 hitter is a good hitter. If in the next moment we also notice that the player misses 40 games a year, that's a data point worth knowing too. I don't see why you can't do the same thing with per-minute statistics. Having awesome per-minute stats is an argument in favor of a player, and playing 48 consistently is an argument in favor too. Combining the two is best of all.

Posted by Martin | April 6, 2008 5:09 PM

Of course you penalize players for playing fewer minutes!

This is relatively simple, and an analogy to baseball stats makes the case. Let's take Pedro's last super-duper-awesome season of 2003 and compare it to Tim Hudson's 2003. They were the league leaders in ERA+, Pedro at 202 and Hudson at 164. But Pedro threw just 186.7 innings, while Hudson threw 240. Those extra 73.3 innings matter, since Hudson was in effect saving them from that many league average innings. So a more accurate comparison is (202*186.7 + 7,333)/(164*240).

With this weighting factored in, Pedro's relative value is 1.14, instead of the 1.23 it initially appears to be. There are other ways to measure, of course; perhaps we factor in an extra 73.3 innings of pitching at the Boston average rather than the league average, or maybe it's 73.3 innings at 107 ERA+, which is what one should get at Hudson's 2003 salary relative to league average.

Either way, though, the player's value is a relative measure, not an absolute; what are they giving you compared to what you would get instead? The result is that any fair comparison discounts the per-minute averages relative to the total number of minutes played. Certainly, one shouldn't just look at them alone or dump them entirely. The penalty on playing fewer minutes isn't absolute; it's relative. But the question is what you're discounting against, not whether to discount at all.

The defense of "his team is doing really well!" doesn't work, either. When a player plays more minutes and his minutes lift his team from also-ran to barely-the-best, that's actually more valuable than lifting a team from barely-the-best to obviously-the-best. If he's getting extra rest because his team is just that awesome, the chances are they're still pretty awesome without him.

Garnett's fewer minutes are nice for him, but what it really means is his coach is attempting to save the player's value for later rather than expend it now. And if you don't think current value trades off with future (playoff) value, then what justifies a coach ever resting his players? Doc Rivers has intentionally withheld some of Garnett's regular season value to use later; so be it, but it should also factor in to considering the league's Most Valuable Player.

Posted by jhupp | April 6, 2008 5:15 PM

THIS IS EXCELLENT NEWS!! FOR HILLARY!!!

Posted by idiotic | April 6, 2008 5:23 PM

Except that players who play fewer minutes are already penalized-- in their per game statistics.

Those who play fewer minutes for the reasons Matt states are most likely hurting their team. But someone like Garnett (as well as the guys on the Spurs and Pistons, who generally play 35 mpg or less) is not hurting his team by playing 33 mpg-- but he is hurting his own stats, getting fewer points, rebounds, etc, than he would by playing 40 mpg.

For this reason-- and for his defense and leadership-- KG is a perfectly defensible MVP choice (although I would give it to Paul, and put LeBron second, if I could vote; KG I'd probably put third, just ahead of Dwight Howard and Kobe).

Posted by PJ | April 6, 2008 5:44 PM

jhupp's argument is pretty good, too.

Posted by PJ | April 6, 2008 5:48 PM

As a Celtic fan, I have to say that Kevin Garnett has been a revelation this year. I knew he was good, but he's a really rare guy: a superstar who really plays like a "glue" role player. He gets a lot of talk about being a superstar, so you tend to look at his numbers in the context of a guy who is the dominant first option of the team. But KG really doesn't play like that. It's the most overused cliche in sports, but the numbers *really* don't tell the story for KG (and I agree with others that Matt's argument about minutes misses the point completely; KG is never not on the court when the team needs him. The Celtics have just had a huge number of blowouts)

You need to watch him consistently (and know basketball) to see what he does for the team. He is absolutely the glue of the team on both ends of the court, he passes up shots to keep the team in rhythm on offense, he forces other players to communicate on defense, and he is a ferocious rebounder, one of the best I've seen. When the game is close, he can completely dominate the defensive boards.

I've been watching the Celtics since the early 80s, and KG absolutely belongs in the conversation of the best player ever to play for the Celtics. That's saying a lot, I know, but KG is that good.

Posted by Brian | April 6, 2008 6:51 PM

Adding ... I'd like to respond to this directly:

The defense of "his team is doing really well!" doesn't work, either. When a player plays more minutes and his minutes lift his team from also-ran to barely-the-best, that's actually more valuable than lifting a team from barely-the-best to obviously-the-best. If he's getting extra rest because his team is just that awesome, the chances are they're still pretty awesome without him.

I don't think this makes all that much sense. You could make the same argument to discount Michael Jordan when the Bulls won over 70. If they were so awesome, they'd be pretty awesome without him, too. Obviously, that's not true. He was just that awesome.

Now, KG isn't Jordan. Jordan may have been the most dominant player in any American team sport ever. But I'm just pointing out that just because a team has a great record doesn't necessarily tell us much about what that team would be doing without its best player.

And as for Doc sacrificing some of his value for the playoffs, that really isn't true. How much value is being lost by him not playing in so many fourth quarters of blowouts? The comparison to baseball isn't between an oft-injured starter and a workhorse. A better comparison would be in the "leverage" stats used for relievers. KG is playing in all the hi-lev situations; he's not just pulling mop-up duty.

Posted by Brian | April 6, 2008 7:08 PM

While a player MPG are interesting data, you should also take into accounts all possible causes, some of which may have nothing to do with the player himself. A very good backup or a coach that likes to use his bench a lot can cause a starter to have relatively low MPG but aren't really his fault.

Posted by Charrua | April 6, 2008 8:30 PM

Brian, well, first off, in 1996 Michael Jordan was 5th out of the top 10 MVP candidates in minutes per game, so it's not as if he was slacking off. Second, the team really might have been pretty awesome without him; remember they also added Rodman that year to a team that had 55 and 47 wins the previous two years, and they got a much better performance from Ron Harper (some of which is probably attributable to Jordan).

But that aside, I wouldn't disagree that even Jordan minus a couple minutes a game should have been MVP because, as you say, he was just that awesome. His per minute contributions were so much better than everyone else's that it's just not close. But Jordan at 40 minutes a game is still offering more value than Jordan at 35 minutes a game, which is yet still more than Jordan at 30. It's probably true that at a certain point, each additional minute offers a diminishing rate of return, but it's still some additional return.

Still, I think the disagreement is less on Jordan than on a slightly less great player like KG. I guess I would just say, don't overemphasize the word "discount"; I'm using it in the economics sense where it means "loses some value," not the oft-used sense of "loses all but negligible value." Taking away most of the value from a guy playing about 15% fewer minutes wouldn't make any sense.

KG sitting does mean that someone who isn't as good is playing those minutes. So maybe you take the league average PER and sub it in, or maybe you take the Celtics average PER and sub it in. Either way, on average every minute he sits really is a minute the Celtics are getting lesser production from someone else, which just can't be thought of as additional value (except in the player development for future seasons sense, which seems a silly basis for MVP).

Now, the flip side to that is what you say at the end: you can play other people there because you don't need him anymore if you're blowing people out. Okay, but then again, you don't need him then, which inherently means his value to your team shrinks dramatically for the duration of those minutes, playing or not. But even if those additional minutes are less important than the first 30, they're still a bit of extra value that you're letting remain potential rather than actual.

Finally, and then I swear I'll leave the topic alone: the idea (brought up by Charrua and others) that fewer minutes aren't the player's fault seems irrelevant to me; no matter who caused it, he's still playing less. (Thus I don't care why Jerome James is barely playing if he's barely playing; ditto for Thaddeus Young, whose +/- is outstanding but is playing just 37% of Philly's minutes.) It does make a little difference if it's because the team's way out front, but otherwise, so what? Most Valuable Player If Only the Coach Wasn't Being Stupid isn't an award. (Probably should be, though; Young would win in a walk.)

Posted by jhupp | April 6, 2008 9:09 PM

Yawn. There's no point discussing which stats are better without saying what question you're trying to answer with them.

Posted by Boring Commenter | April 6, 2008 10:20 PM

Man, what happened to the suns tonight? they scored NINE points in the 4th quarter.

Posted by Suns Guy | April 7, 2008 1:26 AM

I continue to be mystified why Dave Berri is treated as an authority. Even when he makes what seems a sensible point, per minute production matters, its still in service of his completely nonsensical analytic model in which the defensive rebound is the height of basketball accomplishment, and the very BASIS of basketball skill, shot creation, counts for naught...

Posted by Pooh | April 7, 2008 5:57 AM

Adrian Dantley is going to the Hall of Fame? I was too young during his playing days so maybe someone can enlighten me to the contrary, but that really seems a stretch.

Posted by Antid Oto | April 7, 2008 12:25 PM

Matt,
Every time you write about hoops it reminds me that there's no possible way you could have ever played (at least not beyond HS). No offense though. :) It'd be like me trying to write about politics I'm sure.
Anyways, I just pinpointed my source of frustration with your writing about basketball. You seem to take a baseball analytical approach which doesn't really fit on the hardwood. A basketball team is the sum of it's moving parts, so if you want to worry about something like a player's per minute stats vs per game stats, you have to take into account what the team stats were during those same time frame. My college coach used to say High School teams are made at the dinner table, College teams are made in the dorm rooms. Which means to say a player's value to a team is also determined by the camaraderie component that only those inside the team can know / understand, which is why teams like Davidson can make a run in march.

Posted by Brian Holmes | April 7, 2008 1:38 PM

Adrian Dantley is going to the Hall of Fame?

Statistically one of the best pure scorers ever.

23rd in career points scored, with a career TS% of 61.7% (by comparison, MJ had a career TS% of 56.9% and Kareem was 59.2%).

I think he deserves it, but I think Rodman should be in the HoF as well -- I'm sympathetic to players that are arguably the best ever in a key area of the game.

Posted by NickS | April 7, 2008 2:05 PM