@tH Talks to Kevin Pelton (Part 1)

by Rohan on Nov 24, 2010 9:15 AM CST in Interviews

There's a lot of high quality basketball writing out there these days. We have multiple networks of team blogs, excellent general NBA sites, humor blogs... really everything and then some. Included in that list is the rise of statistically oriented analysis, merging excellent observational analysis with factual data.

Basketball Prospectus' Kevin Pelton has been at the forefront of the NBA statistical movement for many years now. For the second straight NBA season, he (and Bradford Doolittle) have published the Pro Basketball Prospectus. For those not familiar with it, I'd describe it as the ultimate handbook for the basketball junkie. Every team and every player in the league is broken down statistically and analytically. The amount of depth and research found in the Prospectus is just incredible. If you're not a stats guy, you'll still become infinitely more knowledgeable about the game simply by reading the team breakdowns. If you love stats, this is the holy grail. It's simply a perfect mix of analytics and top-notch writing. The 2010-2011 Pro Basketball Prospectus is available as a PDF download for $9.98 or in print for $19.95.

Last week, Kevin Pelton graciously agreed to answer ~~some~~ many, many questions for At the Hive about the Prospectus and the 2010-2011 New Orleans Hornets season. Since the interview runs rather long, I've decided to split it up into two parts, one today and one Friday. Part 1 is after the jump.

Update: Click here for Part 2.

Rohan: I'd like to begin more generally before getting into specifics about PBP '11 or the Hornets. The most general question is surely also the most complex: how does one go about building a projection system? How would you explain the inputs and functionality of the SCHOENE system to a layperson?

Kevin Pelton: You can trace SCHOENE back to when I first read about Nate Silver's PECOTA projection system, which was introduced in Baseball Prospectus 2003. That provided an ideal template, but getting there took a long time. The first step was coming up with similarity scores to match players with peers from their past. I came up with those by the summer of 2003 and took a first stab at projecting statistics for a handful of players. By the start of the 2004-05 season, I had developed it to the point where I projected stats for everyone in the league and used them to come up with team projections.

Then everything sat idle for four years, until I started devoting more time to Basketball Prospects in the fall of 2008. I tweaked the similarity scores and player projections, but the key difference was generating team statistics and creating this complete imaginary league where everything balanced out against each other. That took the better part of October 2008. Basketball-Reference.com pointed out that its projections outperformed mine, which offered the impetus to incorporate three years of player statistics instead of just one, which helped improve SCHOENE. Along the way, I've also added translations for college, Euroleague and D-League statistics so we have a projection for just about everyone in the league.

The way I try to sum up what SCHOENE does is like this: We take three years' worth of player stats, then estimate the aging process by looking at how similar players at the same age have developed in the past. Using these player statistics, projections for games missed due to injury, subjective guesses at playing time and a handful of team indicators on defense, we come up with projections for team performance.

R: How do you measure the success of a projection system?

KP: I think it is pretty straightforward in terms of comparing the projections to actual results, but there are a couple of complications. The first is, what's the baseline? Basketball-Reference.com's Simple Projection System provided a pretty good baseline at the individual level, since it's much, well, simpler, than SCHOENE. It outperformed us in 2008-09, but using three years of statistics allowed SCHOENE to leap ahead last year. At the team level, we tracked a variety of statistical projection systems last year and SCHOENE finished second of these. The other complication is there is a lot of noise in one-year team projections, so it will probably take a few seasons to really make a definitive comparison between various projection systems.

R: Staying on the subject of player comparisons, you mentioned that Chris Paul "broke" the SCHOENE system (either in '08 or '09, can't remember) in terms of very few players being comparable to him in NBA history. Can you explain which stats were particularly unique and how often you see such a thing?

KP: In Paul's case, I think the biggest factor is how good he was--and he was coming off an MVP-caliber season at this point--at such a young age. If you opened up similarity to all ages, there were a couple of comparable players--Isiah Thomas, most notably--but generally those performances came much older. We see situations like this, if not as extreme, with unique star players (Steve Nash, for one), specialists (like Steve Novak) and especially young players (Jrue Holiday being an example this year).

R: How long and through what age ranges do great point guards sustain elite levels of play, historically?

KP: I think that might be getting older. At one point, the conventional wisdom was somewhere around 32; Thomas, for one, was retired by age 34. Nowadays, however, it seems rare for elite point guards not to keep playing at an All-Star level through at least 35. Two factors seem to work in favor of point guards as they age: Size and shooting ability. Paul's height won't help him, but I think by that point in his career he should be established enough as a shooter to help cover for declining quickness.

R: Is there a position that tends to age better than others? Or is aging more a function of player skill-set or reliance on athleticism?

Big guys do tend to hold their value longer, since their size never ages. In general, I'd agree that player type is at least as important as position. Versatile players, for example, tend to age better at any position. The factors I mentioned for point guards (height for position and shooting ability) are also generally important.

R: How do you anticipate Paul aging as he gets to his late 20's and early 30's?

KP: Well, the wild card with Paul is how his knee will hold up. Assuming average health, he probably has another three or four years left after this at a superstar level, followed by a gentle decline. John Stockton and Steve Nash have both stayed strong into their late 30s. That's the ceiling.

R: I'm of the opinion that building around a guard vs. building around a big is an overrated debate- a championship team needs quality in a lot of different places. Is there actually a good answer to this?

KP: I agree. I don't think there is any one formula and the discussion about this tends to ignore counterexamples. If you were going to compare an idealized version of these Hornets to anyone, I suppose it would be the Bad Boys Pistons. Not the same kind of toughness in the frontcourt, of course, and Thomas had taken a step back by that point, but it would have been hard to argue against building around a point guard when Thomas and Magic Johnson were squaring off in the Finals.

R: One thing we often see in the early season is people overreacting to small sample size induced stats/trends/etc. Is there a time period or general time of year at which we can note that teams are "for real" (or this mostly a function of how much of a surprise a given trend is)?

KP: I looked at something like this question last Thursday and found at the team level that at this point preseason expectations and performance to date should be considered about equally. Like you said, the level of surprise factors in. The Hornets have played well enough so far that we can conclude confidently that they are an above-average team, but it is difficult to get much stronger than that. Here's a more positive note, though. I have an annual tradition of looking for non-playoff teams that start 7-3 or better. Even though it's early to judge them, these teams are very likely to make the playoffs and often accomplish much more. Already, New Orleans has qualified with three games to spare. So that's an excellent sign.

************

Halftime! Let's take a day to digest some of that... more Hornets related topics scheduled for Friday. Much thanks to Kevin Pelton for taking the time to talk to us.

Tweet Comment 4 comments | 0 recs |

Do you like this story?

Comments

Despite the loss to the Clippers

and how lousy, dejected or whatever we felt; we should still remember how well we’ve done overall. A season no matter how good/bad always has peaks and valleys. As Rohan stated a few articles ago, the last few games have shown an semi-alarming downward trend in performance, but it doesn’t mean we were just a hot start team.

I have an annual tradition of looking for non-playoff teams that start 7-3 or better.

I’ve usually felt similarly about the first ten or so games. Flukes are rare and we must not forget the quality of competition we were beating on a nightly basis before the spiral down. I’m hoping that the team simply got a little complacent with the fast start and a few players curbed their effort just a bit because they started at a 110% level. However, if it’s one thing we’ve learned, this team has to play at such a high level because we really aren’t among the most talented teams in the league. What will propel us to more victories and losses is effort and execution. Let’s hope the team realizes this and starts trending towards another peak in our next game.

by RedHopeful on Nov 24, 2010 9:42 AM CST reply actions

Fantastic Read

I’m not big on the statistical movement here lately in basketball but that’s because I’m not a big Hollinger guy as well. Because I want to have a career involved in the sport, it’s imperative that I’d find some way to digest the analytical portion of the game and use it to maximize the talent on my team. I may even check out his book now because of this interview and, overall, it’s an interesting and impressive system on the outside looking in.

Good reporting, Rohan!

"You play to win the game."

by MrWayneKeller on Nov 24, 2010 12:46 PM CST reply actions

Hollinger is probably not the best representative for the advanced stats movement...

But that seems to be what he’s become. I definitely respect him though, and he’s certainly done great work. He rubs a lot folks the wrong way though and I can understand why.

I would generally rather read Kevin Pelton, Dean Oliver, and Neil Paine (just as a few examples) than Hollinger.

by Caleb462 on Nov 24, 2010 5:25 PM CST up reply actions