Time to punch the numbers
Wednesday, October 3, 2007
John Hollinger's revised ratings are out and the bloggers are talking. Happy: Fans of the Rockets (projected NBA champions), Bulls (Eastern Conference champs), and Kelenna "Sambuca" Azubuike (as Hollinger projects Marco Bellinelli to be so bad, the Warriors will have no choice but to play him).Not so thrilled: Wizards fans, who are trying to understand how a team with
While there are worse concerns for the Wiz these days than a columnist's predictions...still, yikes. Doctor Dribbles may publicly trust in PER, but most of the We Rite Goode riters are closet Hollinger-heads; for our purposes, his words carry scary weight. The estimable Pradamaster is over at Bullets Forever furiously debunking Hollinger's rationale, which includes allegations that Gilbert Arenas's quickness-dependent game will slide and Antonio Daniels is due for a fall. Not entirely idle speculation, but Prada's click-worthy argument strongly rebuts these concerns.
But one Hollinger point--that Antawn Jamison is slipping--is an observation that we'd also long, secretly feared.
True, Jamison's PER of 18.4 was his highest in three seasons in DC; he certainly seemed to fill it up down the stretch and in the playoffs. But, we worried, might his performance have been skewed by an end-of-season flurry, with Arenas and Caron Butler sidelined? And if the Big Three stay healthy this coming year, and Arenas and Butler continue to dominate possessions, doesn't it seem likely that Jamison's numbers will take a dip?
Granted, this may be an effort to rationalize Hollinger's prediction (it's horrible--in some twisted way, we'd prefer Jamison to actually be declining than to have Hollinger be wrong), but curiosity got the better of us and prompted some game log investigation and back-of-the-envelope math courtesy of Excel. We've saved our very basic spreadsheet here, for those who are curious.
The findings:
* In 70 games, Jamison's season averages were 19.8 ppg, 8.0 rpg, 45% FG shooting, and 16.3 FGA per game. (Granted, it didn't take Excel to uncover this data, just ESPN.com. But a crucial piece of baseline info.)
* Jamison actually played 14 games without Butler and/or Arenas; 15, counting the game where Butler played 13 minutes and took 2 shots, before removing himself because of stiffness; and 16, where Gilbert played two minutes before hurting his knee. So, operatively, 16 games in the "not-so-Big-Three" (NSBT) games--that's nearly 23% of his 70-game season. Importantly, Jamison averaged 37.9 mpg in those games versus 38.0 mpg on the season, which we can call that a draw.
* In the NSBT games, Jamison averaged 22.75 ppg, 8.37 rpg, and 18.3 FGA per game.
* Accounting for the NSBT games, Jamison's averages in the 54 games with Butler and Arenas fall to 18.92 ppg, 7.83 rpg, and 15.9 FGA per game. A huge difference? No. But demonstrable? Yes, and probably enough to measurably effect PER, which rewards for field goal attempts, in addition to other performance gains in the same MPG.
* An important caveat: Jamison's shooting percentage shrunk when other teams increasingly keyed on him, shooting only 40.9% in the NSBT games, compared to 46.4% in those 54 other games. And who knows--that FG% decline may be enough of a counter to the other performance gains. (Honestly, we don't know, and will defer to Ty Keenan and Carter Blanchard to load all of this data into the PER formula.)
* However, Wizards fans may remember a six-game stretch where the Big Three all played, but Butler was burdened by the knee stiffness that had kept him out in late February and would force him back out in March--as a result, although Butler was on the court and drawing defensive attention, he was much less active in the offense. In those six games (arguably, a tiny sample size), Jamison averaged 21.8 ppg, 7.6 rpg, and shot 48.6%.
So what does this all mean? That Jamison's PER was affected by his performance in Butler- and Arenas-less games? (Maybe.) But is this a significant observation, since every NBA player will have games where he carries more of the burden, as teammates are out with injuries/clashing with coaches/hiding from process servers? (Probably not.)
Getting to Hollinger's point, does it mean Jamison's declining? If the crack WRG basketball team had to read into the numbers, we say no; more likely, the data indicates that Antawn can still be a reliable scoring option, but at this point in his career, directing too much of the offense through him comes at a cost to his efficiency. But can Jamison step up his game in context of the Big Three? Note that when Arenas and an injured Butler were on the court and drawing the defense, Jamison may have enjoyed one of his more efficient and effective stretches of the season. His days of carrying a team are over (if they ever existed), but Jamison remains a key third wheel.
(And, in the most important finding, does it mean that we learned how to upload a spreadsheet to the Internet? Oh yes. And you can too.)
Labels: Antawn Jamison, Bullets Forever, Caron Butler, Gilbert Arenas, John Hollinger, NBA, Washington Wizards
posted by Crucifictorious @ 21:09,
,
Paul Millsap better have a heck of a year
Wednesday, September 12, 2007
Sequels almost never live up to the original, but The Empire Strikes Back, Dr. No, and High School Musical 2 just got company: Tom Ziller is back at Ballhype with "Love and Mathematics, Pt. 2," and it's a doozy. The post is notable not for what it proves but what it disproves: The fallacy that more minutes equals decreased production.
With data support from Ballhype founder Jason Gurney, Ziller goes deeper on what he calls "The Paul Millsap quandry," in honor of the Utah Jazz rookie (the theoretical test case we keep returning to) who put up sterling production in about 18 mpg last year.
But would Millsap be just as good a starter as he was a sub? Or would playing more minutes against first-team competition expose his weaknesses? It's that sort of question that's been floating around since Plissken at the Buzzer first questioned PER, John Hollinger's kitchen-sink stat on NBA players' offensive performance, over the weekend. (See recap here). FreeDarko's Silverbird5000 and Bethlehem Shoals then advanced the debate by discussing how the inequalities of NBA play muck up the data in a way that makes Hollinger-style statistical analysis misleading.
However, Ziller limits himself to a data set (251 role players whose role increased the subsequent year) that made sense to this blogger. To summarize his analysis, 70% of the players incurred a slight-to-significant increase in production, while 28% of the players declined. Citing the standby correlation vs. causation rule, Ziller's cautious about the significance of these findings:
I want to emphasize this again: We don't know [whether Millsap's per-minute production would suffer, improve or be maintained given more playing time]. We cannot look at any of this data and say "Increased minutes leads to increased per-minute production (aka efficiency)" just as we cannot and should not say "Increased minutes leads to decreased per-minute production." But this data does indicate a positive relationship between minutes and efficiency.(In this corner, I'm guessing that more playing time tends to correlate with an NBA player's progression, having always thought of mpg as somewhat of a bell curve. Thus, players are getting better as they earn more minutes. But you know what...I'm more than happy to let Ziller, Jason, and co. do the math on that).
Still, Ziller hammers home a key insight: That more playing time doesn't mean a player's PER will necessarily go down. And yes, this conclusion may be intuitive to some of us, but given the fray that's gone on the past few days, sometimes it takes a big stick of data to truly get the attention of non-believers.
Whew! When does Professor Dean Oliver stroll into town?
Now, to you--the reader who stumbled across this.
Look, I'm not going to kid you--I'm glad you're reading this, but this debate is moving well out of the scope of We Rite Goode at this point. We can follow the argument and will continue to happily
However, if you want to read about mediocre athletes and cheese, and usually not in that order, stick with us.
Note: There are some who point out--and perhaps rightfully so--that a conversation so focused on PER leaves out the other school of thought on basketball statistics put forth by Dave Berri. I can't speak for the model--as it's been nearly a year since I looked at Wages of Wins, and remember only that I found it lacking--but FreeDarko was clinically dismissive of the work.
Labels: John Hollinger, NBA, Paul Millsap
posted by Anonymous @ 00:27,
,
No stupid questions, just confused bloggers
Sunday, September 9, 2007
I don't carry a calculator, have never read Bill James, and barely survived high school math. That said, I'm a total stathead...at least for certain metrics. The good ones. Like PER.
A quick background on PER, for readers who've avoided ESPN.com the past year or Magic GM Otis Smith: It's John Hollinger's stat that combines "the good (made shots, steals, assists, rebounds, blocked shots, free throws), and subtracts the bad (missed shots, turnovers, fouls) by assigning a point value to each item." The figure is then adjusted on a per-minute basis and by the team's pace. It doesn't account for defensive presence, but it's a good barometer for offensive ability.
For example, when Raja Bell moved from the turgid Utah Jazz to the up-tempo Phoenix Suns, of course his scoring average went to nearly 15 ppg--there were plenty of possessions to go around and he played 37+ mpg. Writing for ESPN, Hollinger was less-impressed by Bell's "breakout season" and rightfully so; Raja didn't do anything different from his past performance, he just had more chances to do it. (amazingly, Raja's closest comparison, according to the database, is actually an old friend of ours).
(And yes, I love Raja as a defender and wish he played for the Wizards, but offensively, he's one of the weaker starting-2 guards in the league).
Similarly, Tim Duncan may average "only" 20 ppg (and you'd hear this argument last year, from stupid commentators who didn't think that was an MVP figure) but he's playing for slow-it-down San Antonio offense and just 34 mpg. His PER last year was 4th best in the league, which sounds far more representative of his value than his ranking on conventional lists (20th in ppg, 7th in rpg).
Look, PER isn't the greatest. If taken out of context, a bench player--who has a mini-scoring spree over a few minutes of garbage time--can look like Michael Jordan. But over the course of the season, players who rack up enough meaningful minutes can be accurately judged.
Anyway, what's this all about? Why question PER, a stat increasingly trusted by basketball fandom, in the middle of a football weekend? Well, Plissken at the Buzzer begged to differ with the formula. Here's the gist:
Plissken: Want to write about Lamar Odom. Lamar Odom=good. So why only 16.2 PER? No one understands PER. Pape Sow and Dajuan Wagner=better PER than Odom. PER lies. Give me conventional stats.
(Having seen the post, I immediately wrote a witty, clever rebuttal, as is my style)
Me: PERperfect, but very good. Pape Sow and Dajuan Wagner misleading examples. Still best single stat in basketball, take with grain of salt.
There were pronouns and verbs too, but I'm lazy and want to get back to watching football.
Anyway, Tom Ziller--who pens the incomparable Sactown Royalty blog, which has the unfortunate distinction of being a Sacramento Kings blog--went and did me a lot better. Writing on BallHype, Ziller goes all "your argument, not so much" to Plissken, with convincing prose, anecdotes, and even a few graphics to defend PER's honor. Good stuff. More interestingly, it's drawn out a bunch of well-read sports bloggers, who reveal where they come down on PER (Jason Gurney: Pro. Miss Gossip: A little skeptical.) Much more interesting wedge issue than ethanol.
Anyway, if Hollinger, PER, or the whole idea of a
(Update: Cruc added the Plissken link earlier; here are a few others worthy of checking out, too. Thank You Isiah has a short, similar (dare I say, more incisive?) critique of Plissken, with a longer discussion in the comments. And juggernaut FreeDarko has gotten in on the conversation with an intriguing statistical analysis of PER inflation and deflation).
(Update on the update: Ziller's back with a sequel of sorts, breaking down what happens to role players' PER when they get more time. If clever stats and easy-to-read graphics aren't your thing, though, you can read our simplified summary here.)
Labels: John Hollinger, NBA
posted by Anonymous @ 17:08,
,
Around the Web...
Tuesday, June 26, 2007
From ESPN.com: The Maestro of Mathmatics, the Professor of PER, John Hollinger finally delivers a system to project college players' NBA success.From Gilbertology: Two more Arenas/Durant commercials that you'll be tired of by Friday, but amused by today.
From The FanHouse: Why interest in Kevin Garnett is going nuclear.
(And From the BasketballJones via WithLeather: Why you shouldn't bat an eye.)
From the Washington Post: The glamorous, glitzy life of the college basketball scrub-turned-Ukranian professional basketball player.
Labels: Around the Web, Gilbert Arenas, it's getting drafty, John Hollinger, Kevin Garnett
posted by Crucifictorious @ 23:38,
,
