Rockets Shot Selection: Amateur Study of the '3 & Paint pts' Strategy

Rockets Shot Selection: Amateur Study of the '3 & Paint pts' Strategy & Performance

With all the talk about the Rockets’ offensive philosophy being heavily influenced by analytics, I tried to see if the Rockets’ wins this year showed the 3’s and Layups strategy to be successful. This is obviously an amateur’s attempt, but I hope you still find it interesting.

The data is based on the shot charts for each game from nba.com/stats which divides the half court into the following sections (Figure 1). I took four main shot types: 3’s, long 2’s (L2), short 2’s (S2), and the Paint. The areas are accounted as below.

3’s: 3LC, 3LA, 3C, 3RA and 3RC
Long 2’s: L2LC, L2LA, L2C, L2RA, L2Rc
Short 2’s: S2L, S2C, S2R

Spoiler

Figure 1

Assumption 1: The efficient shots were 3’s + Paint only which means my definition says Long 2’s and Short 2’s are inefficient. This is obviously up for debate whether or not to include sections of the short 2’s area, but I had to start somewhere.

This (Figure 2) is the collection of data for the Rockets season so far, excluding tonight’s Memphis game (1/24/2014). This is a cumulative area plot showing the percent of each shot style taken in each game. Left axis shows the % of each shot taken in a particular game, right axis shows the Total Shot attempts for the particular game, x-axis has the game number in chronological order.

The yellow area shows the % of total shot attempts in the paint, blue is % of total shot attempts from 3, grey is the % of total shot attempts as Short 2’s, and orange is Long 2’s. The points are the total shot attempts for the particular game, green indicating a Win and red indicating a Loss.

Spoiler

Figure 2

Overall, we can say that the fraction of Paint shots varies from 45 – 55%, and % of shots from 3 ranging from 23 – 43%. We haven’t taken many 2’s, long or short, ranging from 5 – 20% (rare). We can’t see a clear relation between wins and types of shot either. The # of shot attempts per game varies wildly from 65 – 95, but typically they take 73 – 83 shots per game. Other teams’ average ranges from 76 – 88 with a 83 mean, and the Rockets average 79 with huge std dev.

Arranging the games by increasing Opponent D Rtg, I get the following chart (Figure 3). The x-axis is now Opponent D Rtg and the Opponent is listed at the top.

My hypothesis was that I can use DRtg and get a correlation with the shot selection for a particular game. Net Rtg correlates well with winning, so I assumed DRtg is the best metric to evaluate defense. Weirdly, DRtg did not correlate well with any of the shot attempts by itself (data not shown). One immediate problem is the use of present DRtg as of 1/24/2014 so it doesn’t reflect the opponent team’s defensive ability at the time the game was played.

Spoiler

Figure 3

Once again, the Rockets have won games against teams with good D Rtgs, and they have lost games to teams with bad D Rtgs. Contributing factors to this: schedule, players are out-of-form, injuries, and too many more to list. What we do see a correlation for though, is the number of shot attempts and the Def Rtg for the stronger defensive teams that the Rockets played. Low Def Rtgs indicate a strong defensive team and the Rockets shot the ball more times when playing them. Between the two white lines, as the defensive rating increased from 94 to 104 (weakening defense) the total number of shots decreased from ~85 to ~70, almost the lowest mark. This makes sense as we expect good defensive teams to make the Rockets miss more often, the reduction of points for which the Rockets compensate by taking more shots.

The aim of course was to see how the shot selection changes. Against the best defensive teams that we’ve seen so far, 3’s and Paint shots contributed the least to the total shot attempts taken (75 – 80%) [looking at the LHS of the chart and the fraction of grey and orange areas]. The fraction of 3’s and Paint shots then increased gradually to ~87% at the defensive rating of 102.6 (MIN & PHX). Against the good defensive teams, the Rockets also shot the highest number of Short 2’s. This is interesting! Even when the Rockets struggled, they were conscientious enough to not shoot the Long 2’s!

Above the 102.6 Def Rtg, the fluctuation begins and there is no clear relation to be seen from this chart. The Rockets lost some high scoring games, and won some low scoring games.

I actually think the function of Ttl shot attempts = f(Opp Def Rtg) looks cubic, where the shot attempts increases again for the middle to lower middle ranks of Def Rtgs, then falls again against the really bad defensive teams (presumably because the Rockets would just make a high % of shots against them).
---------------------------------------------------------------
I then played around with some bubble charts to see when the Rockets best followed the 3’s and layups plan. The chart below (Figure 4) is again just the first 43 games played this season showing the % of shots taken that were 3’s and Paint shots against the Opponent Def Rtg . The size of the bubble indicates the shooting percentage (Shooting%) from 3 and the paint simply calculated as the (FGM from 3 and the paint)/(FGA from 3 and the paint)%. The bubbles are scaled to emphasize the spectrum. Red bubbles are games that the Rockets lost, and blue bubbles are games we won.

Spoiler

Figure 4

The Rockets are clearly inconsistent. Out of these 43 games played, they shot less than 85% of shots from 3-land and the paint in 17 games (40% of games). Of these 17 games, they had a decent Shooting% (bubble size) in 8 games, losing two of them (OKC & SAC).

In the 26 games where they shot 85% and higher, they had a decent Shooting% in 9 games (BRO, DAL, DAL, BOS, WAS, MIN, LAC, DET, POR), losing only 3 of them (LAC, DAL, DAL). So when the Rockets stick to the plan and make their shots, their offense alone is enough to win (referencing the shootouts with DAL and LAC). Of the remaining 17 games with poorer shooting%, the Rockets still won 12 games! These games were against opponents with only average Def Rtgs, but it’s a good indicator that even if the shots are not falling at a high clip, following the 3’s and paint points method is a solid strategy to winning regular season games!

The anomalies in this chart are clearly IND and OKC. Against IND the Rockets couldn’t stick to the overall strategy and couldn’t make those shots. Against OKC, they lost one game by abiding by the strategy and missing shots (top right) and lost another by taking more Long & Short 2 pointers, even though the 3’s and paint shot attempts were made at a decent rate.

---------------------------------------------------------------
In the last section, I looked at other teams’ shot selection.

The bubble chart below (Figure 5) has been posted by Grantland before, but I still made it to assess that my choice of areas on the court to count in the 3’s or paint scoring was appropriate. And I think it is! The Rockets, MIA and PHI are the three main teams to use the 3’s and layups strategy. We’ve heard about the analytics use in Miami and with Sam Hinkie as the PHI GM, it’s not surprising to see that PHI are following a similar strategy. The chart also shows that more Western Conference teams shoot a large number of 3’s and paint shots, than Eastern Conference teams. The Spurs are not featuring the 3 and paint pts as highly as the Rockets or Miami but they are making these shots at a high rate. PHI unfortunately have a low shooting%, probably because their roster is so young.

Spoiler

Figure 5

To get an idea of how the shot selection correlates with winning, I plotted the data against the Net Rtg (Figure 6), which is the difference between a team’s Off Rating and Def Rating, said to be an excellent metric to indicate winning.

Spoiler

Figure 6

A trendline on the western conference teams gives a positive correlation between the %Shots taken from 3/Paint and the Net Rating, albeit with a crappy R2. Teams like DAL, UTA, NOP and SAC are making these shots at a high rate, so it may be feasible to say that if these teams shot more of these shots, their Offensive Rating could increase (previous plot) and hence their Net Rating.

Just to check, I plotted the same thing against the Winning% of NBA teams (Figure 7), and it looks almost identical to the chart above against the Net Rating.

Spoiler

Figure 7


That was probably a tedious read but I hope it was somewhat insightful, and hope to read some suggestions and discussion.

PS: Can someone post a tl;dr summary? :grin:

 

Thanks for the comment.

I think that's the point of the strategy: that over the course of 82 games when the Rockets don't tailor their approach to an opponent as much as they would in a playoff series, the 3's and Layups philosophy increases the probability of a win (clarification: there's no rule against shooting the L2's and S2's, just that there is an order of preference in the shot type). Accounting for the fact that the Rockets will be cold some nights, over the season sticking to this strategy seems to yield more wins than shooting more midrange J's, as indicated on figure 6: the more winning teams are positioned towards the right and top, even if the regression line's R^2 value is low.

Figure 4 showed that for the Rockets this season, in games where they shot 85% or more of their shots as 3pters or paint shots, they ended up winning 18/26 games. Pretty decent rate! And they didn't shoot particularly well in those games. In fact they lost 3 games despite hot-shooting from these areas, so it was the defense that was failing them.

To your point about OKC and IND having the big men to shut down the paint - it's valid, and my attempt is to quantify what we saw in those games into a model or chart. I think I need to add in stats of players from each team, perhaps accounting for players with the best DRB% or DRtg or even DWS on their team. That's part of the problem: that it's tough to choose any single advanced metric to characterize the defense, or any other aspect of the game.

 

Update to include the two Memphis disasters.

Figure 1 is the same as that discussed in the OP. In both Memphis games, the Rockets couldn't get the shots they wanted and weren't making much else. The range for the bubble size i.e. Shooting% is from 34% (MEM) to 61% (BOS).

Spoiler

Figure 1

Figure 2 is similar to figure 1 except that it shows the % of shots taken as long and short 2's. The bubble size shows the rate at which they made those shots. The range for the bubble size i.e.Shooting% is from 0% (BOS) to 61.5% (POR).

Typically, in games that we won: if we shot well from 3 and the paint, we didn’t shoot as well on long and short 2’s and vice versa (if we shot badly from 3 and the paint, we shot decently on long and short 2’s).

Spoiler

Figure 2

Figure 3 is the 3pters and paint shots data plotted against the point differentials of our games so far. It shows the record disparity between Eastern and Western Conference opponents. Negative point differentials shows the amount we lost the game by. Point differential is not the best way to indicate performance in a game, but I used it for the sake of getting a different lateral spread to the data points than using DRtg.

The way to distinguish between the shootout games and the low scoring dogfight games (both close to 0 pt differential line) is to look at the % of 3s&Paint and the shooting%.

Spoiler

Figure 3

Figure 4 is the same thing but with Long and Short 2’s. Combined with Figure 3 it basically shows that when the Rockets win handily, they shot well from 3s&Paint and/or from Long&Short2s. When the Rockets lost in blowouts, they shot badly from 3s&Paint and/or from Long&Short2s, with a generally lower fraction of shots from 3&Paint.

In close games, the Rockets shot below average either from 3s&Paint or from Long&Short2s.

Spoiler

Figure 4

I started to look correlate the % of shots attempted as 3pters/Paint with the Winning%, but the relationship is too weak.
Even accounting for the shooting%’s it’s tough to eke out a clear relation.

Summary so far:
1) Against teams with DRtg<101: It’s tough to shoot >85% of shots as 3s&Paint against good defensive teams (no games where it was done).
2) Against teams with DRtg>101, if the Rockets shoot >85% as 3s&Paint, they have a pretty good chance of winning (18/26 games).
3) Against teams with DRtg>101, shooting <85% as 3s&Paint will give the Rockets a win as long as the shooting% is good.
4) When the Rockets win handily, they shot well from 3s&Paint and/or from Long&Short2s.
When the Rockets lost in blowouts, they shot badly from 3s&Paint and/or from Long&Short2s, with a generally lower fraction of shots from 3&Paint.
5) In close games, the Rockets shot below average either from 3s&Paint or from Long&Short2s.

Side note: the Home-Away splits don't seem to matter that much

 

Thanks - I edited the post and Summary point #1 in particular to answer your question. To elaborate:

- the Rockets haven't shot above 85% fraction of 3s&Paint shots against teams with DRtg<=100, so the conclusion for those teams is probably that it can't be done or that it is very difficult to shoot those type of shots.
- for teams right around 101 DRtg, i.e. the two Clippers games and one CHA game, we beat CHA by shooting high percentage on the Long&Short2s. Against LAC, I can't pinpoint exactly what the issue is. We stuck to our game plan against them and had a little below average shooting%, so I think our downfall there was just bad defense on our part.

 

Simply put I would say the following:

1) Against average to below average Def teams, we can win by shooting a high fraction of 3s&Paint shots versus Long&Short2s.

2) Against better defenses, we can't shoot enough 3s&Paint shots or hit them at a good enough clip to win, but we have a good shooting% on Long&Short2s so we should probably adjust the playbook to shoot more of those

 

Right; I don't specifically look into our defense and opponent offense here, but those would definitely help to capture the LAC matchup problems.

 

Definitely.

That's one of the counter-arguments to it, but the opposite could maybe be argued for other teams. For example, last year I and some other posters really wanted to face Memphis in the first round because we blew them out in a couple of regular season games and they were coached and practiced playing slow. If we had faced them, I was confident that our running would outdo their plodding.

But, it's a hypothetical now. I would say a similar thing applies with tailoring as you mentioned. Plus it won't just be the Rockets; every team will game plan differently for the playoff games to target their opponent's weaknesses.

If I had to look at prior data to show that it could be more difficult for the Rockets to have to play slower, I think I would struggle. I can't think of too many teams that play similarly enough to us. Maybe the loaded Suns teams from 8 years ago, the Spurs, or the Golden State team that beat Dallas. Those are all positive examples of teams doing decently in the post-season. Frankly, I think DEFENSE wins championships. If the Rockets play solid D and lose, then I would definitely agree that this offense style is a gimmick.

 

In fact, I tried to go back and compile the distribution of shots in games against us because I felt that would give a really good indication of the inconsistency of the team. I wanted to see how often we brought our best defense, and if it was good enough for the playoffs. I wanted to see that the Rockets played good defense even when their offense struggled against the other good defensive teams. Sure enough, each game was pretty different from any other and I'm real having difficulty finding any trends. I'll post if I ever find something worth noting

 

I fully agree with all of those conclusions. the Rockets are currently pretenders imo. The defense being bad is not fully endorsed by the numbers I gave here, but my eye test agrees with you. I really want the defense to pick up as well.

As I mentioned above, when I looked at the shot distribution for Rockets defense, it showed that the Rockets' randomly played good and bad defense. Not sure if that should worry or encourage at the moment. Dwight's presence has definitely improved it from close to league-worst last year, but I think it can get better still, and needs to.

Maybe contrary to you, I feel confident with the personnel. I think they have the talent to post top-10 Def Ratings, and they already post league high Off Ratings. There are obviously many many valid excuses for the team at the moment - injuries, incomplete roster, familiarity to name a few. The problem I have is that they haven't seemed to improved significantly since the start of the season, and for a playoff run they need to. I really want a good playoff run this year.