Gary Neal

Randy Wittman vs. Analytics

wittman grimace

Yesterday, Wizards head coach Randy Wittman appeared Sportstalk 980’s “Sports Fix” with Kevin Sheehan and Thom Loverro (as transcribed by the Washington Post’s Dan Steinberg) and provided a treasure trove of comments worth some analysis.

Let’s unpack a bit, shall we?

Writes Steinberg:

The Wizards have their own special history with the [analytics]. There were critics throughout much of the 2014-2015 season who wanted Randy Wittman’s Wizards to play a smaller, faster, more three-point-friendly game, and who sometimes used numbers to make their case. Owner Ted Leonsis, at least in one blog post, seemed sympathetic with their cause.

ESPN the Magazine gave the Wizards a mediocre “analytics” rating, writing that “Washington lags in terms of applying the lessons of analytics to its shot chart even in the midst of the team’s best season since 1978-79.” The Wizards went smaller in the playoffs and found some success. After the season ended, the Wizards held an “analytics” scrimmage, and Leonsis defended the franchise’s use of analytics. And by the start of this season, they were debuting a faster “pace and space” offense that seemed more aligned with modern NBA thinking.

A very fair and cogent summary of the Wizards’ recent history with statistical analysis. The thing that makes me twitch is the notion that “analytics” said the Wizards should play faster. The numbers I track indicated that a) playing fast hadn’t helped the team in recent years, b) that if anything the Wizards were slightly better in slower-paced games, and c) that playing fast or slow or in-between is a bad goal because it doesn’t mean anything. There’s no prize for having a lot of possessions.

During 2014-15 (and the several preceding seasons where stat guys made similar points), the lesson from the numbers was that the Wizards could benefit by exchanging two-point jumpers for threes, at-rim attempts and free throws — to the extent possible. The team’s “go fast” approach was a leap of faith unsupported by the numbers.

From Steinberg:

At least one local critic — ESPN 980 host Kevin Sheehan — has pointed the finger at the “analystics”-inspired change.

“I wouldn’t be surprised if we find out down the road that this small lineup pace-and-space style of play was forced on Randy Wittman and his staff,” Sheehan said recently, “forced on them by some advanced-analytics stats geek who convinced the technological visionary who owns the team that this team was stuck in yesteryear, that this team was stuck in old ways of playing basketball that weren’t going to work anymore:

Hey Ted, Ted, I’ve got this logarithm that I wrote, it’s really cool stuff, I got the idea from an app that I created for sci-fi movies and it’s really gonna work in the NBA, bad twos, you can’t take those anymore, you’ve got to take threes, you’ve got to space and pace, you’ve got to go small, you’ve got to play a stretch 4, this is the way of the future, Ted, this is the way you’ve got to do it. And Ted said to Randy ‘Hey Randy, what do you think about this pace and space and stretch 4s and shoot more threes and it worked in the playoffs and we almost made it to the Eastern Conference finals.

“Jesus!” Sheehan concluded. “This whole thing all season long is just a blown opportunity. A major blown opportunity. I would love to know what Ted is thinking right now.”

If Washington has a stat goober who told them the numbers said they should play fast, he should be fired. Second, the space part of the new offense has probably helped a bit. Last season, the team ranked 22nd in offensive efficiency, 1.9 points per 100 possessions below league average. So far this season, they rank 18th — about 1.0 points per 100 possessions below average.

This is a fairly small effect, which at least one stat goober expected before the season. The Wizards weren’t the only team to apply the lessons of statistical analysis, defenses have been adapting, and games generally come down to overall talent and execution. And the team has middle-of-the-road talent.

Washington’s real problem, of course, has been defense, which has nothing to do with pace or space. Last year, they ranked fifth overall defensively, 2.6 points per 100 possessions better than average. This year: 17th, about a half point per 100 possessions below average. Over the past two seasons (individually or combined), there was no relationship between pace and defensive efficiency.

Appearing on Sheehan and Loverro’s show, Wittman had this to say (courtesy Steinberg):

“I’ve got to coach the team. Analytics haven’t won a ballgame. You’ve got to take what you have and put guys in position that they can best succeed at. And there are some things with numbers that help that, but if you see some of the number sheets that we have, it would drive you crazy. But you know what, that’s the world we live in. You can fight that, but that does you no good. Listen, I’ve been in the business 32 years now. We had analytics back in the ’80s, alright? We had numbers. Plus-minus, and guys playing with certain guys, and that’s never changed. It’s just now, for whatever reason … Hey, it’s good for some people. Because guys have gotten a lot of jobs because that of word.”

A few thoughts. “Analytics” is the study of what wins and loses basketball games. “Analytics” are drawn from the actual games. They’re not made up. When done well, they reveal what’s really happening on the floor, pinpoint what’s important, help coaches and players identify advantages and disadvantages that can be discerned in the numbers, but might escape the naked eye. Analytics are a tool to help coaches and players perform their jobs better, and (hopefully) win more games.

Wittman’s comments suggest the Wizards have some serious internal problems, though — and NOT because he’s resistant to “analytics.” The telling statement is “…if you see some of the number sheets that we have, it would drive you crazy…”

A head coach should not be getting buried with sheets of numbers. He’s a basketball coach, not a statistical analyst. Like many busy people, when presented with an overload of information, he’s going to ignore most of it, seize on a few things he thinks he understands, and then go with what his experience tells him is the right strategy.

The proper role for a statistical analyst is to crunch the numbers, perform the analysis, and then communicate the findings in a way that coaches and decision-makers can understand. If Wittman is being driven crazy by the data, then the analysis department is failing. It sounds like the Wizards may be missing the crucial ability to communicate the findings of their analysts.

Also worth considering is the kind of numbers and information being analyzed and presented. It has become fashionable in recent years to break players into their component skills and seek to construct a roster as if completing a puzzle.

Based on comments made by GM Ernie Grunfeld, the Wizards are big into this kind of analysis. Symptoms include statements such as: the team needs to add “shooting” or “defense” or “rebounding” or “ball handling” or “length” or…you get the idea. As if “shooting” can be “added” to a lineup.

This approach has been borrowed from other sports like football or baseball, where specialists can be extremely valuable. This is much less true in a flowing game like basketball. “Adding” a shooter to the lineup means “subtracting” another player from the floor. Whatever specialty a player is put on the floor to perform, the team gets his whole game — offense and defense. So while it’s worth analyzing what guys are good at doing, it MUST be coupled with analysis of his overall impact on the game.

From Steinberg:

“And not to try to get you into trouble, but it’s been sort of a season-long question for Wizards fans, and I’m a big one,” Sheehan said. “And that is how on board were you with sort of this space-and-pace and pace-and-space and going small?”

“Well, I didn’t have big decisions to make,” Wittman said, “because after the roster was put together with the guys that left and the additions that we had, I had nobody that could back up Marcin [Gortat] at the 5 spot. Kevin Seraphin left and I had nobody there. I thought what was best for our team was to take Nene out of the starting lineup and play him more at 5 than at 4. And that was more just because of the makeup, and we had success with it.”

Good question from Sheehan. Wittman’s answer is…interesting. I agree with his point that the roster construction left him with few lineup choices. I’m baffled by his comment about Seraphin because the big fella was terrible with the Wizards and has been even worse with the Knicks.

From Steinberg:

Wittman said he’s sympathetic with armchair coaches, because he does the same thing when he’s watching baseball or football. But he noted with some amusement that last year critics said his team was playing too big, and this year other critics say his team is playing too small. He said he would run out of minutes if he started a big lineup but then also used Nene as his second-string center, but added that a bigger lineup could be used in a shorter playoff series. And he said this year’s changes have both helped Washington’s offense and hurt its defense.

“There’s no question about it, [it] hurt our rebounding a little bit as well,” he said. “And that’s an important factor for us because we want to run. If you don’t rebound the ball, you can’t run.”

The first part of this struck me as a strawman. Statistical analysis suggested the team would be better off taking fewer two-point jump shots, and that the team could probably benefit by adding a stretch-four. That doesn’t mean “playing smaller” — at least not to me.

I disagree with Wittman that the changes are what hurt the defense. And the rebounding really hasn’t suffered much at all. Last season, the Wizards were third in defensive rebounding at 77.3%. This year, they’ve fallen to tenth, but their defensive rebounding percentage is still a robust 77.0%. More teams than ever are opting to emphasize getting back on defense rather than going for offensive rebounds.

If it’s not rebounding, what’s causing the decline in Washington’s defense? Answer: an inability to make opponents miss. Like last year, the team still does a good job keeping opponents out of the paint (fifth best at preventing opponent at-rim attempts; down from third best last year). However, opponent efficiency on at-rim and three-point attempts has improved. That could be about playing smaller lineups (taller players tend to force lower opponent shooting percentages), but it could be something else such as less effective close-outs on three-point attempts, and/or random variation.

Bottom line: bad analytics didn’t sabotage the Wizards — at least not at the coaching level. What’s hampered them this year is the reality that they have very average talent across the board.

Player Production Average

The ratings below are from a metric I developed called Player Production Average (PPA). In PPA, players are credited for things they do that help a team win, and debited for things that don’t, each in proportion to what causes teams to win and lose. PPA is pace neutral, accounts for defense, and includes an adjustment based on the level of competition faced when a player is on the floor. In PPA, average is 100, higher is better, and replacement level is 45.

League-wide PPA scores through games played 3/1/16 are here.

PLAYER GMS MPG 11/10 11/22 12/3 12/13 12/21 12/30 1/6 1/13 1/27 2/11 3/1
Marcin Gortat 53 31.0 91 112 128 133 132 138 147 145 148 151 172
John Wall 59 35.9 153 129 136 168 157 157 149 144 142 146 153
Otto Porter 52 30.5 144 158 104 116 107 115 122 127 130 130 134
Jared Dudley 58 28.4 36 92 90 85 98 103 100 105 99 104 106
Alan Anderson 3 16.0  –  –  –  –  –  –  –  –  –  – 97
Bradley Beal 38 31.0 128 108 96 87 87 86 85 86 98 108 94
Ramon Sessions 59 20.4 131 119 84 90 87 89 88 91 90 89 88
Nene Hilario 35 18.7 58 90 80 74 79 78 79 88 92 84 86
Gary Neal 40 20.2 23 49 64 75 78 74 75 78 71 70 69
Jarell Eddie 18 5.0  –  –  –  –  – 153 119 113 110 86 68
Garrett Temple 57 25.3 38 106 57 54 70 63 68 79 79 69 59
Markieff Morris 7 24.1  –  –  –  –  –  –  –  –  –  – 41
Drew Gooden 27 10.8 99 51 57 56 56 56 38 47 34 31 26
Kelly Oubre 48 11.2 -103 -4 -40 -44 9 37 43 39 36 29 22
J.J. Hickson 2 5.5  –  –  –  –  –  –  –  –  –  – -14
Kris Humphries 28 16.6 90 121 95 80 78 76 79 79 78 76  –
Ryan Hollins 5 9.6  –  – -40 60 59  –  –  –  –  –  –
DeJuan Blair 29 7.5 -345 -129 -112 -45 -34 -38 -38 -28 -6 -15  –

On a per minute basis, Gortat remains the Wizards’ top producer. Wall leads in total production because he plays more minutes.

Markieff Morris has had a rough start to his career in Washington, but should improve over time.

Reality vs. the Wizards

The Wizards are not going to make the playoffs this year. They’re not quite as good as they’ve been the past couple years, which is to say they’re a bit worse than mediocre.

Wizards fans — and perhaps the Wizards themselves — are hanging their dreams of postseason participation on some fragile hooks:

  1. the schedule gets easier, and
  2. key Wizards players have returned from injury.

Neither factor holds as much promise for the Wizards as adherents might imagine. The schedule does get easier, but the Wizards are five games below .500 because they’ve played below average basketball through their first 51 games, not because the schedule was difficult.

Before the All-Star break, Washington’s opponents were 0.67 points per game better than average. That’s the approximate quality of a 43-win team over an 82-game schedule. Against this very slightly better than average level of competition, the Wizards got outscored by 2.7 points per game. Which means Washington performed about two points per game worse than average.

The schedule does get easier. Opponents the rest of the way have been a point per game worse than average the rest of the way — about the quality of a 38-win team over an 82-game schedule. In other words, the Wizards are going from playing the approximate quality of an eighth or ninth place team to playing opponents the approximate quality of a ninth or tenth place team.

There are a couple problems, though. First, the Wizards have been worse by a point per game than the aggregate of their remaining opponents. And second, 17 of their remaining games are on the road, and the post All-Star break sprint starts with three games in three nights.

Run the numbers on their remaining schedule, and at this point they’d be the favorite in 11 of their remaining 31 games. If you believe (as I do) the real quality of the team is about average, they’ll probably win between 14 and 17 games the rest of the way – meaning they’d finish with 37 to 40 wins, and would miss the playoffs.

So what about the injuries? The Wizards have indeed lost the NBA’s highest number of player games to injury so far this season. Unfortunately, the players who missed games weren’t the team’s major producers of wins, and their replacements did a reasonable job filling in.

Prior to the All-Star break, I used PPA to quantify how much injuries cost the team. That involved reapportioning minutes (taking playing time from Garrett Temple, Gary Neal and others, and assigning it to players who missed time like Bradley Beal, Nenê, and others). Based on the performance levels of the players involved, I estimate the team would have won two additional games with perfect health.

Two wins is a significant number, but it wouldn’t change the overall trajectory of the season. Washington would be 25-26 – closer to eighth in the East, but still in ninth or tenth place. Even with an easier schedule the rest of the way, the numbers in this “ideal health” scenario would project the Wizards to finish with 41 to 42 wins, and a likely ninth place finish.

Washington still has a theoretical chance to finish eighth. Chicago and Charlotte have suffered injuries — Jimmy Butler’s being far more damaging to the Bulls than Michael Kidd-Gilchrist’s to the Hornets. But, the eighth seed in the East is going to need a minimum of 42 wins, and to do even that well, the Wizards would need to transmogrify from playing like a 37-win team to playing like a 50-win team.

Here’s a simple table showing the task the Wizards have created for themselves. It’s not impossible for them to make the playoffs, but it is highly improbable. The W82 column shows the quality level the Wizards would need to reach the target wins. So, to reach 44 wins this season, they’d need to finish 21-10, which is the approximate level of a 56-win team over an 82-game schedule.

TARGET   NEED   W82
W L W L
44 38 21 10 56
43 39 20 11 53
42 40 19 12 50

Player Production Average

The ratings below are from a metric I developed called Player Production Average (PPA). In PPA, players are credited for things they do that help a team win, and debited for things that don’t, each in proportion to what causes teams to win and lose. PPA is pace neutral, accounts for defense, and includes an adjustment based on the level of competition faced when a player is on the floor. In PPA, average is 100, higher is better, and replacement level is 45.

League-wide PPA scores through games played 2/11/16 are here.

PLAYER GMS MPG 11/10 11/22 12/3 12/13 12/21 12/30 1/6 1/13 1/27 2/11
Marcin Gortat 45 31.2 91 112 128 133 132 138 147 145 148 151
John Wall 51 36.1 153 129 136 168 157 157 149 144 142 146
Otto Porter 44 31.0 144 158 104 116 107 115 122 127 130 130
Bradley Beal 30 31.8 128 108 96 87 87 86 85 86 98 108
Jared Dudley 50 28.9 36 92 90 85 98 103 100 105 99 104
Ramon Sessions 51 20.4 131 119 84 90 87 89 88 91 90 89
Jarell Eddie 14 5.6  –  –  –  –  – 153 119 113 110 86
Nene Hilario 28 18.8 58 90 80 74 79 78 79 88 92 84
Kris Humphries 28 16.6 90 121 95 80 78 76 79 79 78 76
Gary Neal 40 20.2 23 49 64 75 78 74 75 78 71 70
Garrett Temple 49 25.6 38 106 57 54 70 63 68 79 79 69
Drew Gooden 22 12.4 99 51 57 56 56 56 38 47 34 31
Kelly Oubre 43 11.9 -103 -4 -40 -44 9 37 43 39 36 29
DeJuan Blair 29 7.5 -345 -129 -112 -45 -34 -38 -38 -28 -6 -15
Ryan Hollins 5 9.6  –  – -40 60 59  –  –  –  –  –

Beal has returned from injury playing well. Wall, Gortat and Porter continue to lead the team in production. The numbers indicate the Wizards actually have some decent professional players. What they lack is an elite producer. More on that coming tomorrow.

Wizards Update: Porter Leading The Way

otto porter

I have to say I’m pretty torn on the Wizards so far. On one hand, they’re 6-4 despite injuries to Bradley Beal, Alan Anderson and Martell Webster, as well as growing pains as they figure out how their personnel can mesh with their new Pace & Space offense.

Sticking with the positives: Otto Porter is having a terrific season (leads the team in PPA — see below), and they’re getting some decent play Ramon Sessions, Nenê and Jared Dudley off the bench.

On the other hand…John Wall — despite his publicly stated desire to be an MVP candidate this season — is performing like a pretty average starter, the team seems to crumble whenever Kris Humphries is on the floor (despite decent production from Humphries), and Marcin Gortat is lost on offense.

Since I believe Wall and Gortat will perform more like they did last season, my biggest area of concern is at power forward. Humphries continues to be fairly productive, and has even added a three-point shot. So far this season, he’s at .412. An analysis I did in July suggested .365 could be anticipated.

The problem: the Wizards collapse when he’s out there. So far, Washington is 5.2 points per 100 possessions worse offensively and 5.9 points per 100 possessions worse defensively (net -11.1 points per 100 possessions) when Humphries is out there. While the results are a bit extreme because of the small sample size, keep in mind the Wizards were worse on both ends of the floor with Humphries last season, and Boston was worse on offense with him out there.

My theory: Humphries plays like my iPhone 4 trying to load an app. When it’s time for business (like catching a pass), there’s a pause, then a blank screen, then a spinning wheel, and then maybe some action. It could be that Humphries is just better against reserves than starters.

Unfortunately, the Wizards don’t really have a viable option to start at power forward. Nenê can’t be relied on in that role, and moving him into the starting lineup leaves the team without a reserve center. I’m dubious about Jared Dudley as a viable option — he’s been a below-average rebounder for a SMALL forward throughout his career. (It’s worth mentioning that in the tiny sample size recorded thus far, Dudley is rebounding at his best rate since his rookie year.)

If the team can’t figure out how to play with Humphries as a starter, they may be forced to start Dudley and take some lumps on the board.

Player Production Average

The ratings below are a metric I developed called Player Production Average (PPA). In PPA, players are credited for things they do that help a team win, and debited for things that don’t, each in proportion to what causes teams to win and lose. PPA is pace neutral, accounts for defense, and includes an adjustment based on the level of competition faced when a player is on the floor. In PPA, average is 100, higher is better, and replacement level is 45.

League-wide PPA scores through games played 11/22/15 are here.

 

PLAYER GAMES MPG 11/10 PPA
Otto Porter 10 32.9 144 158
John Wall 10 32.7 153 129
Kris Humphries 10 20.8 90 121
Ramon Sessions 10 18.0 131 119
Marcin Gortat 10 27.5 91 112
Bradley Beal 7 33.9 128 108
Garrett Temple 8 15.4 38 106
Jared Dudley 9 21.7 36 92
Nene Hilario 9 18.6 58 90
Drew Gooden 6 12.8 99 51
Gary Neal 10 18.7 23 49
Kelly Oubre 6 10.0 -103 -4
DeJuan Blair 4 8.5 -345 -129

 

Wizards Renovated Offense Needs Time

beal rips

The Washington Wizards have embarked on a major renovation of their offense, and while the results are decidedly mixed through their first six games, the right thing is to stay the course and see if players can grow into new roles.

Last season, the Wizards were 18th in pace and 28th in three-point attempt rate (the percentage of field goal attempts from three-point range). So far in 2015-16, they’re second in pace (a whopping 8.4 possessions per 48 minutes faster) and 12th in three-point attempt rate.

The positives so far: decent shooting and an uptick in trips to the free throw line. The negatives: the league’s worst turnover rate, a drop-off in rebounding effectiveness, and an elevated opponent shooting percentage.

While patience is warranted as the team figures out how to operate in the new system, the emphasis on playing at a fast pace continues to be a concern for me. Why? Because pace of play has nothing to do with what causes teams to win games. Go through the record of games and seasons, and you’ll find good (and bad) teams that played fast, slow and in-between. What makes sense is for a team to play at a pace where it’s comfortable, where the players are under control, and where it can get good shots and maximize efficiency. If that’s fast, then play fast. If it’s slower, then slow down.

At the risk of oversimplifying, basketball is a game where teams take turns with the ball until the clock runs out. The number of possessions are about the same for each team over the course of a game — sometimes there’s a possession or two variance based on end-of-period exchanges. Playing fast doesn’t get a team extra possessions because no matter how quick they shoot the ball, the other team gets it back. What matters is efficiency — generating more points on your possessions than the other guy does on his.

The emphasis on playing fast — at least as it’s been implemented by the Wizards so far — seems misguided to me. Pushing the ball up the floor every possession is a good strategy because it can stress the defense, catch defenders out of position, and give Washington more time to run its half court offense. But if rushing leads to more turnovers, and weakened defensive rebounding, the strategy is missing the point.

What really concerns me about the “play fast” mantra is the question of whether the Wizards truly have a grasp on what’s important for a team to win. There’s a palpable feeling that they’re mimicking Golden State without a true understanding of why that style of play works for the Warriors, and how it fits Washington’s personnel. What wins games in the NBA? Doing these things better than the opponent (in order of importance): shooting from the floor, controlling turnovers, getting to the free throw line, and rebounding. Pace ain’t on the list.

Player Production Average

The ratings below are a metric I developed called Player Production Average (PPA). In PPA, players are credited for things they do that help a team win, and debited for things that don’t, each in proportion to what causes teams to win and lose. PPA is pace neutral, accounts for defense, and includes an adjustment based on the level of competition faced when a player is on the floor. In PPA, average is 100, higher is better, and replacement level is 45.

PLAYER GAMES MPG PPA
John Wall 6 34.8 153
Otto Porter 6 34.0 144
Ramon Sessions 6 17.2 131
Bradley Beal 6 34.7 128
Drew Gooden 4 15.0 99
Marcin Gortat 6 28.0 91
Kris Humphries 6 20.5 90
Nene Hilario 6 18.0 58
Garrett Temple 4 3.5 38
Jared Dudley 5 21.6 36
Gary Neal 6 17.7 23
Kelly Oubre 3 5.3 -103
DeJuan Blair 2 6.0 -345

In many ways, I find these numbers (drawn from a small sample size) encouraging. The team’s most productive players so far are its youthful trio of high draft picks: John Wall, Otto Porter, and Bradley Beal.

It’s hard to believe veterans like Marcin Gortat and Jared Dudley will continue playing this poorly. Look for Nenê’s production to improve some as well.

On the other hand, don’t expect big jumps in productivity from DeJuan Blair (who apparently has forgotten how to play basketball), Gary Neal and Garrett Temple.

Also, much as I hate to write it, I think Porter’s PPA is ripe for a drop-off. It’s not likely he’ll continue to shoot 70% from two-point range or have a 60-plus percent efg.

Projection: Wizards to Win 42

wall 02My 2015-16 Wizards preview is up at Vice Sports. I won’t steal my own thunder much, except to say that my projection approach says the Wizards will win 41-42 games this season. Not as encouraging as I’d hoped, and I had some mild surprises in the numbers — younger players not being predicted to improve as much as I’d have intuitively expected.

At any rate, please click over and give it a read at Vice.