This week's mailbag features your questions on Kyle Kuzma's 3-point shooting, various season projections and more.
@kpelton what's the sample size where we start to take Kuzmas 3 pt shooting serious?
— Kevin (@just4sportsacct) October 3, 2017
In three years at the University of Utah, Los Angeles Lakers rookie Kyle Kuzma shot 32.1 percent from the college 3-point line. For whatever reason -- he has attributed it to greater focus because he knows he has to jump to shoot -- the longer NBA 3 has agreed with Kuzma. He shot 4-of-5 on 3s in the one scrimmage he played at the NBA draft combine, 24-of-50 (48 percent) at the NBA summer league in Las Vegas and now 8-of-17 (47.1 percent) during his first three preseason games.
My hot take is Kuzma likely won't surpass Steve Kerr as the most accurate 3-point shooter in NBA history, which I'm pretty sure is not what you mean by take it seriously. The question is probably more like, at what point can we expect Kuzma to be an above-average 3-point shooter? From a Bayesian perspective, that's a product of two things: How many NBA attempts Kuzma has and how well he has shot on them.
There's one key underlying question here: Does each individual NBA 3-point attempt tell us more about a player's NBA 3-point shooting than an NCAA 3-point attempt? And the answer here seems to be yes. Using summer league stats for rookies who attempted at least 20 3s in the summer and 50 the following NBA season, I found the best prediction of their rookie 3-point percentage weighted summer league attempts 70 percent more than college attempts (using their projected NBA 3-point percentage, not their actual college percentage).
So, taking preseason 3-point attempts as equally predictive as summer league ones, Kuzma's projection would now be weighted by 169 college attempts to the 29.2 percent he was projected to shoot based on his NCAA accuracy and 117 NBA attempts to the 47.8 percent he has shot in the NBA. That yields a projection of 36.7 percent, a little better than last year's league average of 35.8 percent.
I'm fascinated to see how Kuzma continues to develop as a shooter. I can't recall a case of a player showing this kind of improvement from the NBA 3-point line so early. The change to his projection incorporating summer league stats dwarfs the next-highest improvement, for Anthony Morrow when he went 11-of-20 from 3 in summer league play in 2008. (Morrow went on to exceed even his revised projection by shooting 46.7 percent on 3s as a rookie.)
Any predictive value in preseason performance?
— ROBCO (@Dav93richards) October 6, 2017
Absolutely. As with the previous example of 3-point percentage, the key question to ask here is how different is a team's performance during preseason from what we previously expected from them. I've found some predictive value to exceeding or underperforming preseason over/under lines, on the order of about three wins per season above or below those lines at the extremes.
The exception to that is for teams with an over/under line of more than 50 wins, so the Cleveland Cavaliers, Golden State Warriors and San Antonio Spurs all starting 0-2 shouldn't be reason for concern.
For individuals, the story is similar. The most recent time I studied the issue, I found preseason performance had about a fourth the predictive power of my SCHOENE player projections, which isn't nothing but also isn't everything.
"When looking at best case and worst case scenarios for a team, based off a forecast like ESPN's real plus-minus (RPM), how big a range is typical? What would be the band of outcomes (call it 10th to 90th percentile) we might expect?"
As the saying of disputed origin goes, prediction is hard, especially about the future. So the band is pretty large. Over the three seasons we've published RPM projections, the average team error is about 5.4 wins -- slightly better than over/under lines over that span (6.1 wins, granting that this doesn't account for bookmakers changing the odds on the over and under in response to action rather than moving the line). Here's how they compare graphically:
The over/under lines have done slightly better at predicting records almost perfectly (in this case half a game; I just used the integer of difference), but RPM has been slightly better at getting within three games (52.2 percent of teams as compared to 46.7 percent) and avoiding huge errors (that 23-game difference belongs to the 2014-15 New York Knicks, who won 17 games with an over/under of 40.5).
But either way, even above and beyond the challenge of rating players and how they'll fit together, injuries and midseason transactions ensure that there's a wide degree of variability in how teams actually perform. So there's reason to hope even if your team has a weak projection.
"For a variety of reasons, Dallas and Memphis have a history of exceeding statistical projections. What do you see of the likelihood of that happening this year?"
I don't know that I'd put the Dallas Mavericks in that category. The Mavericks have actually fallen short of their preseason RPM projections two of the past three years, though that can be attributed to the Rajon Rondo trade in 2014-15 and the decision to punt late in the season last year. I'd say there's a decent chance of something similar happening this year, where Dallas is more competitive than the RPM projection (34.8 wins before accounting for schedule) indicates but ends up near that mark because of late-season losses.
The Memphis Grizzlies also had a stronger track record of exceeding projections early in the Grit n' Grind era. (Memphis beat its retroactive RPM projection by at least six wins each season from 2009-10 to 2012-13.) To the extent the Grizzlies have exceeded expectations lately, it has often been by outperforming their point differential.
I'd give Memphis a pretty good shot at beating this year's RPM projection (also 34.8 wins before accounting for schedule) just because it's so low. That would be an eight-win decline from last season and seven wins worse than the 42 wins the Grizzlies' 2016-17 point differential typically would have yielded. I'd say my expected average outcome for Memphis is closer to 40 wins than 35.
Some say say that Wilt never fouling out was bad because it meant he didn't play aggressive defense. What's your opinion? #peltonmailbag
— George Zip (@JSlakov) October 4, 2017
In his great "The Book of Basketball", Bill Simmons quotes John Havlicek's autobiography, "Hondo", on this topic: "Wilt's greatest idiosyncrasy was not fouling out. He had never fouled out of a high school, college or professional game and that was the one record he was determined to protect. When he got that fourth foul, his game would change. I don't know how many potential victories he may have cheated his team out of by not really playing after he got into foul trouble."
While it's impossible to analyze Chamberlain's own play, there is evidence that, in general, players perform worse when in foul trouble. A 2011 study found that teams underperform when they leave starters in foul trouble in the game, which makes sense logically -- a player concerned about fouling won't be as effective defensively as one who can play without fear of fouling.
At the same time, I think attributing Chamberlain's desire to avoid fouling out strictly to vanity is a mistake. Pretty obviously, there was an immense drop-off from Chamberlain to his backup, which is why he averaged 45.8 minutes per game for his NBA career. (This probably doesn't get enough notice.) So what was worse for Chamberlain's teams: him playing less aggressively to avoid fouls, or him fouling out?
Granting that I wasn't there and can't say just how limited Chamberlain was in foul trouble, I think Havlicek and other critics are probably confusing correlation with causation here. The problem wasn't how Chamberlain played in foul trouble; it was him getting in foul trouble in the first place.