Neil Paine was nice enough to speak with members of At The Hive recently for a quick Q&A session. Among the topics discussed were selfish rebounding, a possible explanation for the slow start for the offense and why there is still hope for the Hornets.
For those who do not know, Paine is a writer for fivethirtyeight.com, a "data-supported sports, politics and culture site" according to ESPN. Among his most recent works for the site: how exactly the University of Kentucky would fare against the Philadelphia 76ers and a goodbye to Jason Collins, "LGBT And Advanced Stats Pioneer". Paine was a part of the team that previewed the 2014-2015 NBA season and predicted that the Hornets would finish eighth in the East.
At The Hive: We really enjoyed the data from (538's) section in the NBA season preview in regards to selfish rebounding. Considering some of the comments made last year about the Indiana Pacers and the Charlotte Hornets' acquisition of Lance Stephenson, I'd really like to see that data for a) current Hornets and b) Stephenson when he was part of the Pacers.
Neil Paine: The Hornets didn't actually have many selfish rebounding bigs last year... Defensive rebounding is where this would show up most, and here are the numbers for the 2014 Hornets/Bobcats:
DR is defensive rebounding percentage, def is GotBuckets' plus/minus defensive rebounding percentile, defE is expected percentile from defensive rebounding percentage, and defDiff is the difference between actual percentile and expected (negative equates to more selfish). Stephenson's defDiff was -9.5 last year, so he was slightly overrated by raw defensive rebounding percentage.
ATH: With all of the advanced statistics out there today, as a writer, how do you one digest it all, and how do you balance it with the "eye test", and what stats today do you think are the most useful in evaluating individual performances of players (considering both sides of the ball)?
NP: I think the best statistics are the ones that are denominated in logical units (like points or wins, instead of an arbitrary scale like PER) and do a good job of predicting how a team will perform out of sample. For instance, if I compile a team of guys who are good at Metric X, and Metric X is a good statistic, then my team should be good as well. If they aren't, and that pattern persists across many different teams/years, Metric X is probably not a good metric. By that standard, a statistic like Real Plus/Minus is currently the best publicly available metric, in no small part because it does take into account defense. Among boxscore stats, a number like BPM (which Basketball-Reference recently added) also does better at the out-of-sample prediction test than PER, Win Shares, or (especially) Wins Produced.
As far as the eye test goes, it's really important to combine scouting with the numbers to get a complete picture about a player's game. You have to be able to know what type of player a guy is, and recognize that there's still a lot of context fueling the numbers he produces. You especially need scouting to predict whether a player will be a good fit or not.
ATH: Could you think of a more apt player to lead the league in True Usage percentage than Jannero Pargo?
NP: True Usage puts a lot of emphasis on passing in addition to shooting, and it's rare to see a player like Pargo last year post a traditional Usage over 28 percent and an assist percentage of over 40 percent. But certainly Pargo is the worst player on that list, and he never really came close to combining those stats at the same level in the past.
ATH: In your opinion, who is the most valuable member of the Hornets' Big 3: Kemba Walker, Al Jefferson or Lance Stephenson?
NP: Right now, Walker has played the best in 2014-15, but I still think Stephenson is probably Kemba's equal, at least in terms of potential long-term value.
ATH: You talked a lot about the Plexiglas Principle in your season preview for the team. Can you expand upon that further and talk about ways the team can avoid the principle, if there are any?
NP: Plexiglas just speaks to the tendency of a team that showed rapid improvement to regress a bit back to their previous level the following year. The implication is that part of the perceived improvement was luck -- players performing above their true talent levels. But it's certainly not a hard and fast rule; the Bobcats themselves bucked it last year, improving yet again after making huge strides following their horrendous 2012 season.
Coaching seems to be one way to try to combat it; in my Spurs preview, I found a few coaches who had the persistent ability to resist regression to the mean. And another way to avoid being hurt by the principle is to simply embrace it and not overestimate the talent level of the team on the basis of an improved season or two.
ATH: Why do you think Charlotte's offense has been off to such a slow start? What would you do to fix it, if anything?
NP: The big problem is shooting -- from the floor and the line. They aren't taking many threes, which is increasingly a staple of efficient offensive teams, and they also aren't effectively working the ball inside enough to draw a lot of fouls (they're also shooting poorly from the line, but that's another issue). Charlotte is one of the leading mid-range shooting teams by volume, which is a bad recipe for offensive efficiency, and it's a large part of why they rank 28th in 2P% and 27th in eFG%. They're also not getting enough opportunities in transition, per Synergy numbers. It's a distribution of shots that is strongly weighted toward the least-efficient attempts in the game. As far as fixing it, I think the coaches can encourage a more optimal shot distribution, but you also know Stephenson, Walker, Jefferson and Marvin Williams won't keep having a collective offensive rating lower than 99.
ATH: Which players do you expect to see the biggest progression/regression from this season?
NP: Our RPM projections thought Kidd-Gilchrist would continue to improve (which has been true in the early going), and that Gary Neal would have a poor season (he's been much better than expected so far).
ATH: Can you give one Hornets specific piece of analysis for the At The Hive faithful?
NP: Okay. The Hornets are currently -65 in point differential through 12 games. How have teams with similar starts fared by season's end? I gathered all teams since 1978 who were between -85 and -45 through 12 games. The mean of those teams was -62.2 through 12 games, so they're pretty representative of how the Hornets have started. Then I looked at the eventual pythagorean winning percentage of those teams. The mean end-of-season pyth% was .381, or 31 wins per 82 games. The eventual pyth% values were also approximately normally distributed around that mean with a standard deviation of about 7.9 wins/82, so the upper bound of 95% confidence interval for teams in that Charlotte-like group was about 46.7 wins.
That means if the Hornets are, for whatever reason, atypical of that general group of poor-starting teams through 12 games, they could conceivably play like a 45+ win team from here on out, even in spite of the poor start. (Granted, that would definitely make them an historical outlier.)
Make sure you give Paine a follow on Twitter.