This week's mailbag features your questions on yearly player development, the NBA schedule and more.
"I know that younger vs. older is better when projecting players on their first contract. But how does accrued playing time play into projections? For example, if you ... compare a player who came in two years ago as a 19-year-old vs. a 21-year-old rookie. Let's say a player like Brandon Ingram, who has already played 3,500 minutes in the league at age 20. Do you still project Ingram as a 'young player' with more space to improve than a 21-year-old rookie? Does it matter that the Ingram-type already has a body of work, and does that body of work adjust projections?"
-- Alejandro Yegros
My player projections rely exclusively on age without consideration for experience, because I've found no improvement in terms of predicting development when experience is factored in. If you run a regression on players with at least 500 minutes played both seasons, adding experience doesn't improve the correlation with change in their player win percentage (the per-minute version of my wins above replacement player metric, akin to PER) at all as compared to just considering age.
However, looking at it that way does seem to have obscured one key caveat. I took a slightly different view in response to your question, sorting players by both their age and experience. Here's how that looks for players ages 20 through 24:
There does seem to be a consistent effect in which players improve more between their first and second seasons than players of the same age with more experience. On average, rookies tend to improve twice as much as second-year players of the same age.
Beyond that, however, the effect is lost. And when we're talking about predicting long-term development for a 20-year-old player, that single season of development is relatively unimportant. So I see little reason to discount the expected improvement a player like Ingram would make relatively to a less experienced player of the same age will make.
"Fact or fiction: A year from now, Zach LaVine will be deemed a better player than Wiggins. The answer is I don't know, and that's not a good sign for Andrew Wiggins at this point. Ask Pelton that one."
-- Brian Windhorst
Though he didn't use the hashtag, my ESPN colleague had a question for me in this week's 5-on-5 on the Minnesota Timberwolves and Oklahoma City Thunder in response to a question about whether Wiggins has lived up to his forthcoming max extension.
LaVine, who made his season debut and Chicago Bulls debut Saturday night, has the chance to become a go-to scorer after trading Wiggins and Karl-Anthony Towns as teammates for Chicago's less talented starting five.
Offensively, that could work quite well. In Minnesota, LaVine showed the ability to create his own shot when those teammates were off the court, averaging 21.3 points per 36 minutes in those situations with adequate .520 true shooting, according to NBA.com/Stats. LaVine's strong 3-point shooting allows him to remain efficient even if he's forced to take more difficult shots off the dribble, in addition to the 3s.
The issues are twofold. First, there's the matter of defense. LaVine's minus-2.4 defensive rating per 100 possessions by ESPN's real plus-minus dropped his rating nearly to replacement level last season. That's unlikely to improve after coming back from a torn ACL. And as I noted on Twitter earlier this week, the history of players returning from ACL injuries suggests LaVine will probably be a less accurate shooter than usual for the remainder of this season.
So, I think his value is likely to go down rather than up after his comeback. If he's deemed a better player than Wiggins when we do next year's #NBArank, I think it will say a lot more about Wiggins and the high expectations for him than LaVine.
"Is there a metric to independently measure the skill of coaches vs. the skill of their rosters? I.e., Phil Jackson won all the games but had all the talent in the world. How can we estimate the value added of Kenny Atkinson, whom everyone seems to agree is wringing more wins out of his talent, or Pop (duh) or Rick Carlisle, who is a great coach and somehow won 50 games with an old Dirk Nowitzki? Could we use wins added over the Vegas-expected preseason win total? This seems a bit simplistic. Perhaps the relationship between actual net rating and expected net rating?"
-- Troy Reinhalter
The short answer is no. To touch on the two ideas you offer, the problem with comparing team performance to preseason lines is that they can price in coaching. If all bettors knew about was the talent on the San Antonio Spurs' roster, this would be an ideal metric for evaluating coaches; but of course, they're also aware of Gregg Popovich's track record and adjust accordingly.
Actual performance vs. a projection based on the talent on the roster gets us closer to what we actually want to measure, but it is still challenged by the limitations of our projections. If we're not very good at capturing the defensive value of individual players, for example, a coach might get credit or blame for our misestimates. And if the roster is relatively stable, as they typically are, coaching is baked into the projections to some extent or another.
In the first edition of his version of the Pro Basketball Prospectus series, my Insider predecessor John Hollinger used team record over the previous two seasons plus a regression factor to set a baseline and evaluated coaches based on how well their team's record matched up to that. That method works pretty well, pegging Jackson as the best coach in the league at the time, and the regression factor allows it to capture what Popovich has done in terms of consistently winning at a high rate. But it really rates a team's management as a whole, since we don't account for personnel changes, and therefore harshly penalizes coaches who take over rebuilding teams.
The most creative method I've seen to evaluate coaches was from Jerry Engelmann, the co-creator of ESPN's real plus-minus (RPM), who has used coaches as essentially a sixth player on the court at all times in the same regression used to create player ratings. In theory, this should account for everything that throws off other metrics. Unfortunately, because there's no way to use box score stats for coaches, as RPM does for players, coach ratings are as noisy as traditional adjusted plus-minus for players. I've found it more useful for showing general trends -- for example, that coaches have more influence on defense than offense -- than evaluating specific coaches.
Even if someday we are able to determine exactly how much better or worse a team is playing than it "should" based on talent, we'd be unable to determine how to split that credit between the head coach, the rest of the coaching staff and even player leadership. So I don't see a reliable coaching metric anywhere on the horizon.
"What would you think about the NBA adopting the NHL's new scheduling philosophy incorporating a midseason 'bye week' into each team's schedule?"
-- Jonathan Dennis
I must confess that not being an avid hockey fan, this is the first I've heard of the NHL's bye weeks, which were first adopted before last season. The NHL found teams struggled coming off their "bye" when playing opponents that had been playing regularly, and therefore decided this season to condense all byes into a two-week span that we're currently in the midst of so that teams would be on a level playing field.
This is all somewhat consistent with research into NBA performance by days of rest, which shows a huge difference between back-to-back games and those played with one day off in between, but diminishing and eventually negative returns to additional days of rest. If we were strictly looking to maximize the product on the court, the best strategy would apparently be to offer a day of rest between each game but no more than two -- more or less what the NBA has tried to do in terms of limiting back-to-backs with its own scheduling changes.
Of course, physical performance isn't the only thing that matters here, and the mental break the NHL's byes provide might convey longer-term benefits that aren't captured by simply looking at performance in the following game. Still, for now I'd say I prefer the NBA's approach to managing rest to the NHL's.