A Team Greater Than The Sum Of Its Parts

[Nima]

I interpreted the findings of my research analyzing the distribution of various statistical categories to draw some conclusions about how to best build a team that will be greater than the sum of its parts.

Here is the post:

http://basketballanalytics.blogspot.com ... build.html

Older posts cover the distribution analysis in detail.

I'd love to hear your thoughts.

Thanks,

Nima

[Crow]

Nice post. I agree with it and think that you have written something quite readable.

Are you willing to share any detailed research data and analysis supporting the recommendation of the one lead initiator?

Is sharing offensive rebounding duty important across 3,4 or 5 guys on the court or is it mainly just important to have two 12+% OR% guys on the court at the same time as much as possible? Do you get greater than expected returns when this condition is met? I haven't looked at everybody or the question systematically yet, but of the top 5 OR% teams they generally have 3 strong offensive rebounders.

[Mike G]

Miami and Dallas won titles the last 2 years largely because some role players hit more than their usual number of three-ptrs.

Does this then become an effective strategy for all teams? -- Have your role players shoot a lot of 3's and make more than they normally do.

[Nima]

Nice post. I agree with it and think that you have written something quite readable.

Are you willing to share any detailed research data and analysis supporting the recommendation of the one lead initiator?

Is sharing offensive rebounding duty important across 3,4 or 5 guys on the court or is it mainly just important to have two 12+% OR% guys on the court at the same time as much as possible? Do you get greater than expected returns when this condition is met? I have looked at everybody or the question systematically yet but of the top 5 OR% teams they generally have 3 strong offensive rebounders.

Thank you. This post was based entirely on my interpretation of findings of the research in the older posts (the analysis is there), in an attempt to create something more readable and less a repetition of the research findings themselves. Essentially I tried to look at how various statistics were distributed in a lineup and how it influenced offensive efficiency.

The data I used was lineup data that Dean Oliver provided after I presented my initial research (based on inferior data - resulting in what turned out to be incorrect findings initially). The lineup data provides what each of the 5 players puts up in each category in that lineup. I then reduced each of those numbers to a percentage of the lineup's total in that category and then took the standard deviation of those percentages (a sum-independent standard deviation) in each category among those 5 players. The offensive categories being FGA, 3PFGA, 2PFGA, Assists, TO, Offensive Rebounds, and FTA. I also looked at the defensive side of the ball simply using defensive rebounds, blocks, steals and personal fouls, but the results of that analysis were pretty simple in that we want everyone playing defense (getting boards, block and steals; we want a low standard deviation/even distribution), and we don't want them fouling while they do it (we want a high standard deviation/wide distribution of personal fouls).

The results were simply that offensive efficiency increased as the standard deviation of:

2P FGA increased,
3P FGA decreased,
Offensive Rebounds decreased,
Assists decreased,
Turnovers increased,
FTA decreased.

I also did a simple correlation looking at the relationship between the ratio of 2PFGA/FGA and offensive efficiency, as well as the relationship between 3PFGA/FGA and offensive efficiency, among other similar correlations, to find that we want more 3P FGA and fewer 2P FGA.

The rest was simply my interpretation of what these results meant in basketball terms; what conclusions could we draw from the findings themselves.

Another important note. My initial analysis involved FGA as a whole, not 2P FGA. After the initial analysis that showed we want an even distribution (lower StDev) of 3P FGA and a wider distribution (higher StDev) of FGA as a whole, one could only conclude that it was 2P FGA that was the driving force behind wanting a wider distribution of FGA since a more even distribution is preferred with 3P FGA (and FGA are entirely comprised of 3P FGA and 2P FGA). I had the 2P FGA data to study this and found that to be true. Then I took it a step further and simply thought, what was most likely the driving force behind the result of wanting a wider distribution of 2P FGA. If we divided 2P FGA into two groups, those at the rim/restricted area, and those outside of the restricted area/mid-range (the paint is still kind of a gray area, but I would say shots in the paint that aren't at the rim are closer to mid-range shots than shots at the rim), which of those groups is likely the driving force behind wanting a wide distribution of 2P FGA. The distribution of one of these groups falls on the side of the 2P FGA result wanting a more even distribution, while the other falls on the side of 2P FGA wanting an even wider distribution. I don't have the data to study this, but I think it only makes sense, given the greater difficulty of mid-range shots than layups/dunks, that it is the distribution of mid-range shots that is the driving force behind prefering a wider distribution of 2P FGA as a whole.

Regarding offensive rebounds, the finding from the research was simply that offensive efficiency increased as the standard deviation of the offensive rebounds in a lineup decreased. I haven't taken the analysis beyond that. However, I am not surprised at all that the top 5 OR% teams have 3 strong offensive rebounders. That is consistent with my findings, as having 3 or more strong offensive rebounders would result in a more even distribution of offensive rebounds, by lineup.

[DSMok1]

Intriguing work, and worthy of a new thread. I've just split these posts off into a new thread.

[Nima]

Miami and Dallas won titles the last 2 years largely because some role players hit more than their usual number of three-ptrs.

Does this then become an effective strategy for all teams? -- Have your role players shoot a lot of 3's and make more than they normally do.

Well, I think teams need to have reliable three-point shooters for such a strategy to be most effective, but in the NBA, teams have multiple players who are reliable open shooters, though some teams have a greater wealth in this area than others. They are professionals after all. (Miami really re-upped their 3-point shooting this offseason, and with James being such a significant threat to get to the rim, unless they take a large step back defensively, I think they will be even better this year.) As the study was based on the distribution of attempts, not made shots, I feel it is an effective strategy for all teams to shoot more threes (and make more as a result), specificially when those three-point attempts are coming at the expense of mid-range shots, rather than easy shots at the rim or trips to the line. That said, the better facilitator/threat to score teams have, the greater the quality of these three-point attempts will be, and they will also be more readily available as well (e.g. James' dominance to the rim this year and Nowitzki's dominance in the high post last year, created a wealth of quality three-point attempts for teammates).

One of the main points is that three-point attempts should be taken more liberally, while mid-range shots should be taken more conservatively. While this may result in a slightly lower three-point percentage and a higher percentage from mid-range, I am confident that the end result would be increased offensive efficiency. In the NBA, the Spurs may fit this mold more than anyone, as they take a great number of threes, and use multiple passes after creating a threat to get them, resulting in, what seems to be, a relatively even distribution of assists. Also, their mid-range jump shots come almost entirely from Parker and Duncan (who, along with Manu, almost completely take on the role of initiating the offense) Not surprisingly, though Duncan is one of the greatest of all time and Parker and Ginobili are great players in their own right, they are an older team and the prevailing opinion was that they were over-achieving last year. I think this comes from their style of play and heavy reliance on the three-point shot. Other teams should take note.

A more recent example is Team USA. They relied very heavily on the three-point shot. Everything else was at the rim. Mid-range twos were taken, preferably, only when necessary. Durant became the greatest role player in the history of basketball, as he focused on finishing what James and the PGs created, rather than creating for himself and others. I was also glad to see Carmelo Anthony eschew his beloved mid-range two for the three. Granted, the international line is a little closer, but I feel it would be wise for him to carry over this shot selection to the NBA season.

Another note on Anthony; I feel if he focused on creating a threat by getting to the rim, and replaced his mid-range jump shots with passes to create easy opportunities from 3 and at the rim for teammates, the Knicks would be far better, and Anthony, as a team player, would be far better. He can be such a great threat offensively, especially when taking it strong to the rim (not as good as, but not unlike, James), that there is a great opportunity for him to be a great facilitator if he only passed the ball instead of taking mid-range jump shots. And given that he is such a great shooter, after passing, he could help spread the floor, and perhaps the ball would come back to him for an open three. Given his shooting ability, he can also help spread the floor anytime someone else facilitates, by catching and shooting the three, rather than catching, stopping the ball, and essentially resetting the possession, as he often does. Given that the Knicks don't have the talent that Team USA had, I fear that his better team play had a lot to do with taking a step back into being a role player due to the talent Team USA had, and that it will take some stern coaching for him to truly realize his potential as a team player when returning to the Knicks.

[Nima]

Intriguing work, and worthy of a new thread. I've just split these posts off into a new thread.

Thank you. Very kind of you to say.

[Mike G]

... three-point attempts should be taken more liberally, while mid-range shots should be taken more conservatively.

If you've been around these parts for very long, you'd know that LOTS of people have come to this conclusion. It would only be a slight exaggeration to say Everyone has held this theory.

Defenses (a) protect the paint and (b) chase shooters off the 3-pt line. The mid-range shot is what's left. If you don't have any mid-range shooters, your offense is easy to stifle.

Miami, with all its fearsome rim-attackers, was just 2 pts better than avg on offense this year.

Opponent (defensive) 3pt% ranged from .308 (Bos) to .383 (Den). With opponents being essentially the same for all teams, some obviously stress denying the three more than others.
But I bet all teams recognize the importance: 3pt accuracy peaked in 2009 (.367), but attempts continued to rise to this season (18.4 per game), at a reduced success (.349) .

One could argue for getting more FTA, since that is an even greater return per attempt. But there's only so much shot selectivity possible, and in general one takes what the defense allows.

The other limitation on having guys at the 3-pt line is that they are not in a position to make a pass to a player near the rim. They're almost "out of the offense", except in the limited sense that they might get a pass when they're open.

Most players' shooting% drops off steeply at a certain distance. Carmelo is probably one such player, whose critical dropoff is right around 21-22'.

[Crow]

So you concentrate the tough shots and tough passes. But lots of folks should take 3s and make easier passes to 3 pt shooters or cutters or post players. And everyone crashes the lane and the offensive glass.

Sounds closer overall to the Clippers than anyone else I can think of immediately. I guess it is fairly close to the Thunder too but the Clippers are better as a team on some of these stats. Concentration comparison could be aided by that type of data.

[Nima]

... three-point attempts should be taken more liberally, while mid-range shots should be taken more conservatively.

If you've been around these parts for very long, you'd know that LOTS of people have come to this conclusion. It would only be a slight exaggeration to say Everyone has held this theory.

Defenses (a) protect the paint and (b) chase shooters off the 3-pt line. The mid-range shot is what's left. If you don't have any mid-range shooters, your offense is easy to stifle.

Miami, with all its fearsome rim-attackers, was just 2 pts better than avg on offense this year.

Opponent (defensive) 3pt% ranged from .308 (Bos) to .383 (Den). With opponents being essentially the same for all teams, some obviously stress denying the three more than others.
But I bet all teams recognize the importance: 3pt accuracy peaked in 2009 (.367), but attempts continued to rise to this season (18.4 per game), at a reduced success (.349) .

One could argue for getting more FTA, since that is an even greater return per attempt. But there's only so much shot selectivity possible, and in general one takes what the defense allows.

The other limitation on having guys at the 3-pt line is that they are not in a position to make a pass to a player near the rim. They're almost "out of the offense", except in the limited sense that they might get a pass when they're open.

Most players' shooting% drops off steeply at a certain distance. Carmelo is probably one such player, whose critical dropoff is right around 21-22'.

Thanks for the heads up. I have perused these forums for a couple years, but not too often. Though, I would be surprised if many hadn't already come to that conclusion regarding threes. That said, by more liberally (for 3s) and more conservative (for mid-range 2s), I don't just mean take more threes and take fewer mid-range twos. I'm referring more to the decision to shoot, so it's more about player shot selection. I'm essentially saying a mid-range jumper should be wide open and taken by the few who are particularly efficient from that distance (as I noted in the post, like how Boston uses Garnett to take the significant majority of their mid-range jump shots), whereas three-point shot attempts don't need to be of the same quality based on who is taking them and how open they are.

Though Miami was just two points better than average on the season, how were they during the finals? From observation alone, it seemed like they turned things around once their backs were against the wall against Boston. They were particularly disciplined during the finals and I believe their shot selection improved.

As for guys at the three point line not being in position to make a pass to a player near the rim, you may be right, but if the shot is there, they should take it, if it is not, they should attack the rim, and along the way, if they find that they can't get it all the way to the rim, they will be in a better position to make a pass to someone who might be at the rim, and if not, they have likely attracted enough attention by beating their man that someone else will be open from three. So, with quick ball movement and attacking without hesitation, they can put themselves in position to make that pass while creating space for teammates to shoot from 3 or do the same. So I disagree that they would be "out of the offense." If the players aren't at the 3 pt line, though they might already be in a position to make that pass, they don't have the option to shoot the 3, which is one less option for the defense to worry about, at which point they simply have to be concerned with them attacking the basket, passing, or taking a jump shot that is only worth 2 points, instead of 3. When the defense has to worry about a shooter taking an open shot from 3, it makes the option to attack the basket an easier one.

Regarding Melo, that may be true, but it's likely because he takes and works on the mid-range shot much more. Clearly, he has the foundation to be a great shooter from any distance, and more time and attempts from greater distance would change that. And if they aren't capable of shooting the three (but I would argue anyone who can hit a jump shot consistently from just inside the three point line has the foundation to shoot efficiently a step or two further back), then they should focus on attacking the rim to get easier shots, to the line, or to open up opportunities for teammates beyond the arc.

Take a close look at the Dribble Drive Motion Offense created by Vance Walberg. I think it is likely superior to what a lot of NBA teams run, and certainly superior to isolation.

[Crow]

"Everybody knows that it makes sense to take more 3s" but only 8 of the playoff teams were in the top 16 on 3 pt attempts and 3PTA / total FGA in the regular season. Only 2 of the conference finalists were in this group and they were 7th and 10th on regular season 3PTA / total FGA.

In the playoffs though, 3 of the conference finalists were in the top 6 for 3PTA / total FGA. They wised up when it really counted. But not the Celtics (with analytic staff knowledgeable in and presumably running fancy programming languages on stuff, maybe something else); they were a below average 10th of 16 in the playoffs and a way below average 24th of 30 in the regular season.

[Nima]

So you concentrate the tough shots and tough passes. But lots of folks should take 3s and make easier passes to 3 pt shooters or cutters or post players. And everyone crashes the lane and the offensive glass.

Sounds closer overall to the Clippers than anyone else I can think of immediately. I guess it is fairly close to the Thunder too but the Clippers are better as a team on some of these stats. Concentration comparison could be aided by that type of data.

Well, the Thunder take more mid-range twos than I would like, but they certainly attack the rim well and have many three-point shooters.

But yes, I agree, the right data would make it easy to do such a comparison.

Anyone have lineup data from last year that breaks down each statistical category by player (P1 to P5, rather than for the lineup as a whole)? Ideally, it also includes the possessions played by each lineup to more accurately calculate each lineup's offensive efficiency. Once it is in that P1-P5 format, it is easy to break each down to a percentage and take the StDevs to compare the distributions of various statistics.

Even more ideal would be data that would break down the two point attempts into various types, though most simply they would be broken down to shots at the rim (layups/dunks, shots in paint very close to rim, and shots away from the rim).

[Crow]

I think the NBA is promising public access to lineup data that breaks down each statistical contributions by player in the future.

I don't think anyone currently offers it. I would imagine Evan with his coding skills probably could though.

ESPN with Dean's assistance could too perhaps.

[Nima]

"Everybody knows that it makes sense to take more 3s" but only 8 of the playoff teams were in the top 16 on 3 pt attempts and 3PTA / total FGA.

1) Well, we're just talking about offense here, and defense factors in as well.

What if we looked at the 3PTA/FGA of the most efficient teams offensively rather than playoff teams?

2) I'm also taking the conclusions from the research to make statements about half-court offense here, so just as defense doesn't really factor into these conclusions, transition offense also doesn't factor in.

It would be even better if we looked at 3PTA/FGA in the half-court offense of the teams that are most efficient in the half-court offense.

3) Though this is likely less of a factor, the teams that are actually winning, particularly winning big, are probably shooting fewer threes at the end of games.

When I looked at the correlation between 3PTA/FGA and Offensive efficiency, I was using lineup data, and I looked at lineups that had played over 100 possessions together (maybe 50, don't recall). That data is less likely to be influenced, in my opinion, by the things that happen at the end of games that aren't consistent with what happens the rest of the game. The lineups teams have out there when they are up big or down big at the end of a game are unlikely to be included in that data set.

4) I should clarify that it makes more sense to take more 3s relative to other jump shots. Certainly, if a team can get to the line or get a layup/dunk every possession, they should.

[Nima]

I think the NBA is promising public access to lineup data that breaks down each statistical contributions by player in the future.

I don't think anyone currently offers it. I would imagine Evan with his coding skills probably could though.

ESPN with Dean's assistance could too perhaps.

I know Evan can with his coding skills, as he was kind enough to provide me some data for the 2011 playoffs on Miami and Dallas. Problem was, the small number of possessions by lineup for those teams during the playoffs made it hard to get anything out of the data.

Since Dean provided me that original data set, I imagine he could too. I will reach out to him.