This week’s mailbag features your questions on perimeter defense metrics and more.
“Through 36 games, the Washington Wizards have a record of 11-6 against teams that are .500 and above and 9-10 against teams that are below .500. It is somewhat easy to see how a team could be better against good teams over a small sample size, whether it be because key players happened to be injured against bad teams or out of simple randomness. But there’s often a narrative that emerges around playing down to one’s opponent. Is there any empirical reason to believe that that is a real phenomenon over the course of a full season?”
— Thomas Silverstein
So, yes, I think there is certainly such a phenomenon as playing to the opponent. One way I’ve tried to quantify this in the past is to look at the difference between the standard deviation of a team’s point differential (how consistent it is from game to game) and the standard deviation of point differential adjusted for game location and opponent quality.
Naturally, we’d expect that teams will become more consistent when we account for these factors, since they make up part of the variation in game-to-game point differential. And indeed that’s true on average: The typical team sees its standard deviation go down about 0.8 points per game when the adjusted point differential is compared to the raw version. But there are substantial differences from team to team.
By this measure, the Utah Jazz, San Antonio Spurs and New York Knicks show the least tendency to play up or down to their level of competition, so adjusting for it makes them appear much more consistent. By contrast, four teams actually become less consistent when we adjust for context, which suggests they are playing to the level of their competition.
The Cleveland Cavaliers are perhaps the ultimate example of “good” (though probably still frustrating) playing to competition. They’re 17-5 against below-.500 teams but haven’t beaten a team with a sub-.350 winning percentage by double figures all season. Cleveland is truly doing just enough to win. It’s probably no coincidence the Cavaliers and Golden State Warriors rank near the top of this list. I found something similar with the eventual champion Los Angeles Lakers when I looked at this measure during the 2008-09 season.
Washington’s performance against weak opposition is a little more complex. The Wizards don’t score as more consistent without considering context because they have tended to blow out bad teams at times. However, their poor games against weak opposition have frequently resulted in bad losses.
Typically, fans tend to think of the better performance against good teams as the “true” ability of the team, where the bad performance against weaker opponents is a fluke. As you hint, I think the truth lies somewhere in between. The Wizards are much better than their play against sub-.500 teams but not likely as good as they’ve looked against above-.500 teams thus far.
One source of this randomness is the most common one in the modern NBA: 3-point shooting. The Wizards have made 38.9 percent of their 3-point attempts against winning teams and 34.3 percent against losing teams, a trend that surely won’t continue. On the other end, below-.500 teams have made 33.8 percent of their 3-point attempts against the Wizards, while above-.500 teams have shot just 33.6 percent as compared to a weighted season average of 36.9 percent.
As those marks even out, Washington will probably win more games against weaker foes but be unable to keep up the wins over top competition.
@kpelton Klay Thompson is known as a great defender, but his defensive Real Plus-Minus always seems to say otherwise. Why is that? Is it something like multicolinearity from the Warriors’ common lineups screwing up the regression? #peltonmailbag
— Kevin O’Neill (@WhatWinksThinks) December 22, 2017
No, that’s not the issue.
Readers shouldn’t be scared off by the word multicollinearity, which in this context means that adjusted plus-minus has a difficult time determining which player to credit for a team’s success or failure when they tend to play together frequently. I don’t think that’s a big issue for the Warriors because of games their stars miss due to injury and the fact that Steve Kerr tends to mix his starters and reserves. Multicollinearity is a bigger issue when coaches tend to exclusively play their starters and reserves together, or there are specific players who play only when the other is off the court.
Even if multicollinearity were an issue for the Warriors, there has been plenty of credit to go around for their defense. So if you look at RAPM, the cousin of RPM from co-creator Jerry Engelmann that includes only lineup data, Thompson typically rates as elite defensively for a guard. Engelmann reports that he currently ranks 20th defensively in single-season RAPM and 82nd in the more accurate multiyear RAPM.
The issue, instead, is the box score-stat component of RPM that helps stabilize it. Thompson’s steal rate in particular is exceptionally poor, while his defensive rebounding has also been worse than the average shooting guard’s. Players with box score defensive stats like Thompson’s are typically poor defenders, so his defensive box plus-minus rating is far worse than league average. That rating is similar to the one RPM uses as a starting point to rate players, which means it’s beginning with the assumption that the Golden State defense is succeeding in spite of Thompson rather than because of him.
The lack of steals suggests that Thompson’s defense is probably somewhat overrated, since these contributions tend to be undervalued in favor of one-on-one defense by most observers. Still, this isn’t a case like Avery Bradley, in which elite individual defense doesn’t seem to be translating at the team level. Thompson is clearly a very good defender, just in a way that’s difficult to measure for a system designed to provide the best estimate for all players and not unusual cases like him.
#peltonmailbag Is there a metric that values perimeter defense? If so, who are the league’s elite stoppers based on that?
— Nikolaj Pavlovic (@nikolajdst) December 28, 2017
Speaking of perimeter defense, of course RPM is a metric that values perimeter defense, though I think you’re looking for something that more precisely homes in on individual defense on the perimeter. That being the case, I don’t think there really is a good one, partially because such one-on-one situations rarely occur in reality.
One favored practice early in the advanced stats movement was to look at the performance of a player’s counterpart. This is limited for several reasons: Players don’t always guard their position, switches mean other players can be the primary defender and, of course, the quality of the other four defenders and the team concept also affects the success or failure of any given offensive player.
More recently, analysts have used tracking data to look at how well players shoot when a given defender is nearest to them. (These numbers are available on NBA.com/Stats.) Unfortunately, as Krishna Narsu showed last year on Nylon Calculus, there is little year-to-year correlation in how opponents shoot against a defender outside 6 feet. So while opponent shooting is a useful measure for rim protectors, it doesn’t appear to help us much with perimeter defense.
Ultimately, I think the best way to quantify the best perimeter defenders is by their impact on their team’s defense when they’re on and off the court, an effect that becomes consistent really only over several seasons.
“Which player’s shots have been blocked the most in the NBA? I feel Brandon Ingram‘s drives result in a lot of blocked shots. Is there some data for this?”
— Devaansh Goyal
You can find data on a player’s own shots that are blocked on NBA.com/Stats (good for a leaderboard) and their Basketball-Reference.com player pages (good for year-over-year data). As the first link shows, Ingram has had the seventh-most shots blocked this season, so you’re definitely seeing something accurate.
Perhaps more interesting is how often players have their shot blocked as a percentage of shot attempts.
With a few exceptions (notably rookie Dennis Smith Jr., who has struggled to negotiate shot blockers at the rim), it’s interesting that these players are almost all big men, which makes sense from the standpoint that they’re unlikely to shoot jumpers that are rarely blocked. Ingram ranks 17th by this measure and doesn’t even lead the Lakers.
How about the other end of the spectrum?
Conversely, this list features a lot of 3-point specialists. Dirk Nowitzki both shoots a lot from the perimeter and has a famously unblockable high release, so his presence here makes sense. I was more surprised by Chris Paul, who doesn’t get as much credit for being difficult to block.