Pages Navigation Menu

More Offensive Geometry

Two weeks ago, I wrote a piece for Hardwood Paroxysm on the Lakers’ offense and how they use and distribute offensive possessions. The heart of that piece was a set of radar graphs comparing the offensive distribution of the Lakers to the Spurs, Thunder and Bobcats. After that post went up I received quite a few requests to put similar graphs together for other teams, so I just went ahead and covered the entire league. Here’s the explanation of how the graphs are created and set up:

I began by combing mySynergySports and identifying every offensive outcome for the Lakers that had occurred at least 100 times this season. By offensive outcome I mean both possession type and the specific player who ultimately used that possession. I was focused on deliberate offensive choices so I left out transition possessions and offensive rebounds, who’s frequency may have more to do with opportunity than deliberate design. The radar graph below shows two different pieces of information for each outcome – the points per possession that outcome netted the Lakers on average, and the total number of times it occurred this season. The yellow line represents the points per possession, the purple represents the number of occurrences. (Each vertical segment of the graph represents 100 occurrences. Try as I might I couldn’t get Excel to display the scales for both data sets without having them overlap.)

Below are the graphs for all 30 NBA teams, covering this regular season. The graphs are in order of team Offensive Rating from highest to lowest.

In looking at these graphs there are a few important things to keep in mind. First off, not all offensive possession types are created equal. Some possessions are inherently more likely to be successful. For example, players don’t generally receive the ball in a cutting situation unless they’re in a position to score. Therefore, across the league, players average more points per possession on cuts than they do in isolations or post ups. Here are the average points per possession numbers league-wide from each offensive possession type, according to mySynergySports. Again, I’m only looking at structured half-court offense so I’ve left out transition opportunities and offensive rebounds.

  • Cut – 1.18 PPP
  • PnR Screener – 0.97 PPP
  • Spot Up – 0.94 PPP
  • Off Screen – 0.87 PPP
  • Hand Off – 0.87 PPP
  • Post Up – 0.82 PPP
  • PnR Ball Handler – 0.78 PPP
  • Isolation – 0.78 PPP

There is a clear separation in efficiency between cuts, shots for the screener in pick-and-rolls, spot ups, and everything else. However, creating a more efficient offensive balance is not as easy as just running more cuts. Each of those three possession types are created by some other action that takes place with the ball. An offense can’t be built around those three possession types, without some sort of initial set to create those opportunities. Each team and coaching staff is looking at their personnel and trying to find the most effective recipe for combining and creating these different scoring opportunities.

Keeping this in mind we would expect to see a few similarities in good offenses. Teams that score very efficiently generally tilt the balance towards more effective scoring opportunities. They also score at an above average rate on some of the possession types that are generally less effective. The Spurs are a perfect example. Looking at just their offensive outcomes which occurred at least 100 times this season, we are left with a total of 3558 possessions. 1952 of those possessions, or 54.8% were used by either cutters, spot up shooters or screeners in the pick and roll. Across the entire league just 45.9% of offensive outcomes came from those three possession types. The Spurs have found a way to significantly tilt that offensive balance in their favor. They also have managed to be more efficient than average with those less efficient opportunities. The Spurs had 7 different offensive outcomes that occurred at least 100 times that were on the bottom half of that efficiency list (Gary Neal – PnR Ball Handler, Tony Parker – PnR Ball Handler, Manu Ginobili – PnR Ball Handler, Tony Parker – Isolation, Tim Duncan – Post Up, Tiago Splitter – Post Up, DeJuan Blair – Post Up). Of those 7 different offensive outcomes, only three (Tim Duncan – Post Up, Tiago Splitter – Post Up, DeJuan Blair – Post Up) scored at a rate below the league average for that possession type.

The Thunder are a team that found an offensive recipe completely different from the Spurs. They don’t rely on an efficient balance, but instead on individual excellence. Of their offensive outcomes that occurred at least 100 times, just 26.2% of their possessions were used on those three most efficient opportunities. The Thunder put together the second best offense in the league this season because they made the most of less efficient possession types. 55.5% of the Thunder’s possessions were used in isolations or as the ball handler in the pick-and-roll by Kevin Durant, James Harden, and Russell Westbrook. Across the league these were some of the least efficient offensive outcomes, but Durant, Harden and Westbrook were all well above average in both possession types. Durant and Westbrook get most of the attention, but Harden’s efficiency blew the other two away. On possession types where the league average is just 0.78 points per possession, Harden scored 1.07 in isolations and 1.04 in the pick-and-roll.

In these graphs we also see examples of teams struggling to find success both because of an inefficient balance, and an inability to exploit specific opportunities. The Hawks had some terrific offensive options at their disposal this season. They had strong spot up shooting from Kirk Hinrich, Willie Green, Joe Johnson, Jeff Teague and Marvin Williams. They also had two strong ball handlers in the pick-and-roll in Johnson and Teague. The performance of Teague was particularly impressive, averaging 0.93 points per possession in the pick-and-roll, well above the league average. Unfortunately 29.7% of their possessions in our data set went to Josh Smith in either isolations, spot-ups or post ups. Smith was an atrocious spot up shooter, scoring just 0.81 points per possession, well below the league average of 0.94. He was right around average on post ups and isolations, but again, nearly a third of the Hawks’ offensive possessions were being used on some of their least efficient outcomes.

Then we come to the poor Kings. It certainly could have been worse overall, they had the 21st most efficient offense this season, but at a micro level things were really bad. Their problems ran deeper than balance and raised some questions about the abilities and potential of the players they have on their roster. First off they couldn’t shoot. Other than Isaiah Thomas and Jimmer Fredette, every King who used at least 100 spot up possessions scored at a below average rate. That covers six different players and the span goes from Marcus Thornton and his 0.92 points per possession, all the way down to Tyreke Evans and his 0.70 points per possession. Evans used at least 100 possessions in isolations, the pick-and-roll and as a spot-up shooter. He scored significantly below average in all three. DeMarcus Cousins used at least 100 possessions as a screener in pick-and-rolls, post ups, spot ups and isolations. He was above average in the pick and roll, but significantly below average in the other three. Finding a solution to the Kings’ woes may involve more than just the passage of time and some minor tweaks.

I know that basketball analytics have generally had a hard time sussing out the impact of coaching on team performance. All though we haven’t specifically look at these numbers through a coaching lens, that’s the issue we’re dancing around. Players ultimately make the decision whether to shoot or pass, but the do so within a framework established by the coach. An efficient offense doesn’t just require good players, it requires good players to be used in the right ways.

  • Pingback: Tuesday Bolts – 5.22.12 | Daily Thunder.com

  • krishna

    man this is incredible. And cool looking. Despite people saying that numbers don’t work in situations like this, I think this works

  • Vile Sir Vile Duke Vile Count Vile

    The Rocket’s graph actually looks like a rocket. There’s a conspiracy going on in the NBA and I believe it starts at the top.

  • Pingback: Official Spurs vs Thunder Series thread! - Page 2 - SpursReport.com

  • Justin

    I’m wondering what the regression is between percentage of possessions and PPP for each type of play? We know that USG% leads to lower TS%, for example, so that balance should lead to understanding of what mix of plays can create the highest PPP overall, without going to extremes that are unrealistic.

    • Ian Levy

      I haven’t done any regression analysis on the numbers. Though it seems like their may not necessarily be an ideal mix. The Spurs and Thunder were 1-2 and did it in completely different ways. I’d be happy to do some more digging this summer and see if I can find some trends.

      Thanks for reading and commenting Justin.

      • Justin

        No problem. As an engineer, I kind of think like this. I guess my more general question is how sensitive the effectiveness of a type of play is to the percentage of time that type of play is run. For example, a simple test would be to see what the relationship is between the percentage of plays that are Kobe iso’s and the PPP on those plays, for each game? When he’s hot, does he do it more? When he’s hot, is it because he’s being more selective?

        • Ian Levy

          Unfortunately Synergy doesn’t have individual possession data available on a per game basis. I could do some hand coding by watching the video but would need some time to make it happen. It’s an interesting thought and definitely something I’d like to take a look at closer when I have some more time this summer.

          • Justin

            It is an interesting question. It is much more global than on a per game basis and that would be a lot of work for an answer that might be in the overall data. I think the global answer of how usage of particular play types generally affects the effectiveness of those play types ie how quickly does increasing usage leads to decreasing effectiveness is a very interesting one. Specific example would be how many times can Noah try a back cut without drawing the double from the other side of the lane.

          • Ian Levy

            But don’t you think that sort of resets with each game? Noah’s back cut trickery becomes obvious the more times he does it, but the next night he has a chance to start fresh against a new defense with all eyes on Rose?

  • Pingback: Corporate Knowledge: Spurs appreciation, charts | 48 Minutes of Hell

  • Pingback: Macho Toughness vs. Silent Strength of a Stone – The Downbeat – #756 | | Utah JazzUtah Jazz

%d bloggers like this: