# Breaking down the 2012 NBA Offenses: Looking for Floor Stretch

*It is easier to perceive error than to find truth, for the former lies on the surface and is easily seen, while the latter lies in the depth, where few are willing to search for it.*

*-Johann Wolfgang Von Goethe *

You’ll know it when you see it.

For the last while we have been having an ongoing conversation amongst the contributors for this site about the specific value of position (You can read Ari’s thought provoking piece on this here). As is typical of the truly great pieces that we are lucky to have in this space, it has inspired some interesting avenues of thought for me.

One such avenue was also inspired by the videocast with our own Andres, Vivek and Hoospeak‘s Beckley Mason. In it, Beckley asked a question about the Memphis Grizzlies in the 2011-12 NBA Season. The Grizzlies led the league in offensive rebounds plus steals per game and both of these generally create easy buckets. So, why then are the Grizzlies only at 19th in Offensive ranking (as measured by points per 100 possessions)?

The easy answer seems to be that they shoot poorly as evidenced by their poor Effective Field Goal percentage and it matches what we see for Memphis and Sacramento. The problem is that it doesn’t completely hold for Utah, Milwaukee and Indiana who are not great shooters but do actually rank in the top ten in offense. This lead me to take a deeper look.

Thanks to the wonderful Hoopdata site I was able to look at shotcharts for every team and take this a lot further. Let’s start with the breakdown by position:

That gives you the shot percentage of total by distance from the rim. The picture is a little more clear once you add in the actual efficiency.

The Worst return at the rim was 57% for the Nets (1.14 points per shot), the worst from three was 44% for the Bobcats (.88 points per shot). The best from 3-9 feet,10-15 feet or 16 to 23 feet was 43% for Miami (.86 points a shot). It seems then that to be efficient we want to shoot from the rim or from three and minimize the % of shots from 3-23 feet (or as a colleague once called them “Shit” shots).

How does this help our quest? Turns out the Grizzlies despite stealing and rebounding their way to 22.2 easy opportunities a game tend to lose the plot. I am reminded of something I wrote after their Game #4 loss versus the clippers in the playoffs last year:

*Let’s pick up the Action with 10:40 to go in the game:*

*71-69 Memphis 10:40 to go.***Long 2 pt. Jumper**alert for Memphis. Clips get ball back.*Whoops no. Jump ball and foul.**73-69 and the clips run a set that ends in a Kenyon dunk. 73-71 Memphis.**Another Speights***Long 2 pt. Jumper**. Zbo fouls on the rebound. Bad theme for Memphis.*Bad clippers turnover on drive to hoop. Memphis Sets for Zbo on other end. First smart set in replay for Memphis. 75-71 Memphis**Another bad clippers possession into TO. Another Memphis iso***Long 2 pt. Jumper**possession. They get rb and Gasol nails a jumper on the 2nd go around. 77-71*Foye jumper. Memphis 3 from Conley. Zbo bails them out on the boards. 78-73 Grizz.**Griffin just wiffs in the lane. Memphis in halfcourt. Griffin makes it up by drawing the charge on Gasol. I am hating the 2012 Grizz in halfcourt.**Blake just wiffs again followed by a crazy steal sequence and another bad Memphis halfcourt poss ending in a Gay moonshot. Refs bail them out**Clippers keep attacking the basket on every possession and it pays off. Bledsoe layup followed by the Evans putback on the ft miss. 79-77**Memphis in halfcourt gets bailed out by refs . 3:49 . Gay on the line. 80-77. Important to point out that***Vinnie is out coaching Hollins.***Everything LAC is doing is about five feet from hoop or from three. Foye ties it with a three. 80-80**Another Memphis possession that does not get within 10 feet of the hoop. Lac attacks bucket again. Charge call saves them.**Conley moonshot miss. CP3 bucket in paint. CP3 steal. CP3 drives and sets up Blake. All your bases belong to CP3. 84-80 LAC**Grizz set for Gay off balance jumper. CP3 bad iso. Conley moonshot. Mayo almost steal.Mayo foul. CP3 to the line. 86-80 LAC with 23 left.**Gay makes the first moonshot. Bledsoe and Evans make 1 of 4 fts. Gay makes another. LAC by one 9 ticks left.**Bledsoe is choking, somebody help him. Memphis hc set to the rescue and the clippers win.*

*Timeline over. What did we learn? Memphis is atrocious in 4th quarter play calling. LAC goes for high value offense. This is important.*

Let’s recap, Memphis tends to focus on low value offense as their default. They rank 5th in shit shots % and 21st in efficiency from that range and they also rank 28th in threes as a % shots and 25th in efficiency from that range. Practically, this means that if you clog the lane and force them into a halfcourt offense, they’ll effectively beat themselves by taking contested jumpers from inside the three point line. This then neutralizes their outstanding post bigs and turns their offense into a series of head-shaking heroball possesions.

You could say that their lack of floor stretch kills what should be an effective offense.

Let’s lay this out in another fashion.

When we look at teams by offensive rank and by % by region and efficiency we start to see some real evidence of the floor stretch phenomenon. I don’t think it an accident that the top three teams in offensive efficiency are all towards the bottom of the Shit shot takers and at top of shots at the rim and from three. Interestingly, San Antonio, OKC and the Clipps all are efficient at the shit shots they do take perhaps as a function of the space created. The bottom of that list is also top heavy with shit shot takers and not three point shooting teams.

The lesson to be learned? A simple one is that no team,I repeat **NO team**,should be taking less than a quarter of their shots from three. Let’s take Memphis as an example. If they exchanged shit shots for threes until they had 25% from long range they would increase their point per 100 possession by 2.1 without accounting for the probability that it would help their percentages at the rim and other ranges by stretching the defense and giving them additional open looks. 2.1 pts is good enough for about 2.1/31.1*82= 5.5 wins. So five and a half extra wins just from taking more threes.

And before anyone starts about how shooting more from three is going to drive down field goal efficiencies, I looked (see here for the really cool interactive chart). In the last twenty seasons, Three point attempts per game are positively correlated with three point field goal percentage, effective field goal percentage, points per 100 possessions and two point field goal percentage.

The actual results are shown below:

**Trend Lines Model**

A linear trend model is computed for sum of 3PA per game given sum of 3pt Fg %. The model may be significant at p <= 0.05.

Model formula: |
( 3pt Fg % + intercept ) |

Number of modeled observations: |
582 |

Number of filtered observations: |
0 |

Model degrees of freedom: |
2 |

Residual degrees of freedom (DF): |
580 |

SSE (sum squared error): |
8180.07 |

MSE (mean squared error): |
14.1036 |

R-Squared: |
0.199361 |

Standard error: |
3.75547 |

p-value (significance): |
< 0.0001 |

A linear trend model is computed for sum of 3PA per game given sum of EFG%. The model may be significant at p <= 0.05.

Model formula: |
( EFG% + intercept ) |

Number of modeled observations: |
582 |

Number of filtered observations: |
0 |

Model degrees of freedom: |
2 |

Residual degrees of freedom (DF): |
580 |

SSE (sum squared error): |
8158.18 |

MSE (mean squared error): |
14.0658 |

R-Squared: |
0.201504 |

Standard error: |
3.75044 |

p-value (significance): |
< 0.0001 |

A linear trend model is computed for sum of 3PA per game given sum of 2P Fg%. The model may be significant at p <= 0.05.

Model formula: |
( 2P Fg% + intercept ) |

Number of modeled observations: |
582 |

Number of filtered observations: |
0 |

Model degrees of freedom: |
2 |

Residual degrees of freedom (DF): |
580 |

SSE (sum squared error): |
9506.8 |

MSE (mean squared error): |
16.391 |

R-Squared: |
0.0695048 |

Standard error: |
4.04859 |

p-value (significance): |
< 0.0001 |

A linear trend model is computed for sum of 3PA per game given sum of Pts per 100 possesions. The model may be significant at p <= 0.05.

Model formula: |
( Pts per 100 possesions + intercept ) |

Number of modeled observations: |
582 |

Number of filtered observations: |
0 |

Model degrees of freedom: |
2 |

Residual degrees of freedom (DF): |
580 |

SSE (sum squared error): |
9125.1 |

MSE (mean squared error): |
15.7329 |

R-Squared: |
0.106865 |

Standard error: |
3.96648 |

p-value (significance): |
< 0.0001 |

This will leads to all sorts of interesting things.

-Arturo

P.S. Guess who’s teams take threes at an historic rate?