The crucial takeaway is that none of the evaluation methods is sufficient on its own. Plus-minus statistics are often unreliable, individual statistics are incomplete and we can be fooled by what we see (or don’t see). We’ve long moved past the outdated notion that scouting and statistics are opposed to each other. Instead, at their best, they work together to help us form more robust evaluations that are more likely to prove correct.

In 75 words, Kevin Pelton tells you everything you need to know about advanced defensive stats. He may have been referring to player evaluation in general, but this idea also applies to how we think about defense. Video analysis and advanced stats work together: keeping track of numbers on defense sheds a light to certain aspects of the game that had previously gone unnoticed. This is what I suspected when I started charting Olympiakos’ Euroleague games, a process which forced me to pay closer attention to each possession. But how does one know what is worth measuring?

December 1, 2011, Instanbul, Turkey

Even the most basic boxscore provided enough information about Olympiakos’ horrible defensive effort against Fenerbahce. How did this happen, though? In the first game I charted (against Bilbao), pick and roll defense had not been my main focus on, as I found it difficult to come up with clearly defined categories or distinguish between individual tendencies and team defense. Curtis Jerrells forced me to reconsider.

The American guard had a field day coming off ball screens and Olympiakos had no idea how to contain him. Numbers certainly reflected the damage done by Jerrells (and Ukic). For that game, I only kept track of points and fouls per possession after a switch – i.e. points allowed and fouls committed on every possession where a defender switches to a player that he has not been assigned to. At the same time, I also counted how many assists were allowed after a switch. This is a “team only”-stat, designed to illustrate how many good looks the defense gives up after a mismatch has been created, by helping on the ball handler.

A switch between two defenders is a pretty straightforward concept, so there would be no doubt about the meaning of these numbers. However, they were not providing enough data. Fenerbahce picked apart Olympiakos’ switch defense, scoring 36 points on 25 possessions, but what were the alternatives? In other words, what can numbers tell us about different strategies when it comes to pick and roll defense?

December 7, 2011, Kaunas, Lithuania

The answer to these questions made me think about certain criteria that SJ had laid out in the comments of my original post:

the goal of this discussion is to find categories that

A) provide meaningful, statistically digestable results
B) have a very small error margin (…), and
C) are limited in number

Meaningful is the key word here. Pick and roll offense is perhaps the most crucial component on the playbook of most Euroleague coaches and there are basically three ways to defend it when using man-to-man:

• switch (see above)
• hedge out: the big man guarding the screener steps out on the perimeter to stop the ball handler and tries to get back to his man, who’s either rolling to the basket or popping out for a jumper. Another option on this play is to trap the ball handler.
• down: the big man guarding the screener stays in the paint and the player guarding the ball handler tries to go through the screen and get back to his man (also known as ‘level’).

In theory, the difference between these strategies is easy to grasp. But how can they be quantified, especially when it comes to assessing individual performance? What would these numbers mean? This is what I call the Milovan Rakovic paradox.

Against Panathinaikos, the Serbian center worked hard on defense, at least in the first half, when the game was still within reach. The problem is that very little of this work showed up in the boxscore, as Zalgiris conceded 41 points. Even worse, my first attempt at measuring PPP on hedge defense showed that Panathinaikos had scored nine points on the eight possessions where Rakovic had hedged out. Still, it is almost impossible to put the blame for these disastrous results on his shoulders.

For once, the numbers and what I saw did not work together. The footwork of Rakovic on these plays was more than adequate, as he would meet the Panathinaikos guard well beyond the three point line. Most of his problems started when the first pass was thrown – either to the screener rolling to the basket or a shooter stationed on the weak side. This tells us very little about his own defensive performance and much more about the Zalgiris team defense. As SJ noticed, ‘Things get complicated whenever a defensive rotation starts’. When it comes to individual defense this type of stats fails to capture this complexity. But it does say something meaningful.

December 8, 2011, Piraeus, Greece

Here was my dilemma, then: More data, or more clarity? Rob Scott had warned against putting too much emphasis on advanced defensive stats in that same comments section: ‘(D)efensive numbers are a little too abstract to be worthwhile’. Fair point, but with Mirza Teletovic in town, it was time to take an extra step. Two new statistical categories were added for the game against Caja Laboral: points per possession on hedge outs and on down.

Individually, each point would be attributed to the Olympiakos big man guarding the screener. This is neither fair nor accurate when it comes to assessing the defensive impact of Lazos or Hines, but I was more interested in team stats and the comparisons they allow. Rob had also asked a significant question regarding this issue:

‘I can see the usefulness of a coaching staff watching every play and providing breakdowns of how their players perform in a given situation – be it hedging on the PNR, making effective rotations, whatever. But (and finally we get to my question) to what extent do they need to know this as a precise *number* (ie. when Hines hedges the opposition scores .45PPP etc) if, as you say, there are so many other variables in play?’

By putting a number to each defensive strategy against pick and rolls, a team’s season becomes more accessible. It allows fans and coaches alike to monitor fluctuations in both defensive efficiency and strategy. Olympiakos were switching even more against Caja Laboral, and their defense was as bad as it was at Instanbul (even though switches worked better). The coaching staff did not need these numbers to realize that adjustments were in order, but statistical analysis would have certainly provided them with some quite useful guidelines, by assessing a player’s performance in each type of pick and roll defense.

But what about those pesky ‘other variables’? Caja Laboral scored 17 points on 13 pick and rolls defended by Lazaros Papadopoulos. Lazos’ struggles in this area are well known, but this number is not indicative of his effort – 13 of those points were scored against Olympiakos’ ‘down’ defense, where the defender on the perimeter must prevent the ball handler’s jump shot or pass to the cutter. The interior defender has to deal with drives to the basket. In other words, Caja Laboral’s efficient scoring in these possessions had a lot to do with how Olympiakos guards were (not) recovering after the screen.

Quantifying the responsibility of each defender requires an even more detailed chart, which will illustrate the spots on the floor where each of these baskets was scored. Therefore at this point these categories should not be used as indicators of individual defensive performance but they do tell us how a team defends pick and roll and how efficient each strategy is. They also reveal how its opponents attack, as it became evident on the next game.

December 15, 2011, Piraeus, Greece

With Olympiakos making only 29% of their mid-range shots, you can’t blame Andrea Trinchieri for keeping Cantu’s big men – especially 37-year old Dennis Marconato – in the paint. On 14 Olympiakos pick and rolls, the Italian coach went with some variation of down defense, but this time the home team went 9 of 18 on mid-range shots and scored 18 points overall in these possessions. Not all nine shots were made against this type of defense, but Spanoulis was quite effective pulling up in the open space between the screener and the Cantu big man.

Still, Trinchieri did the right thing, based on video analysis and advanced stats of Olympiakos’ previous games. Maybe he should have tried something different when the shots started falling, but the interaction between offensive and defensive stats does not make him look as bad as 86 points on 74 possessions would suggest; these new categories allow a better understanding of a team’s game plan on defense and lay out the context of their opponent’s offensive performance. If Huertas or Teodosic, point guards who possess a very dangerous mid-range game, have a big scoring night, it is worth measuring how they were defended, especially in pick and rolls, in order to get a better idea of how the defense attempted to prevent it.

December 22, 2011, Nancy, France

John Linehan was seriously getting on my nerves: five steals, with three of them leading to six transition points for his team, three disruptions – where, according to SJ, ‘the defender deflects a ball out of bounds or somewhere into the field, forcing the opposing team to reset the offense (but no direct turnover)’ – mayhem all around. Fortunately for Olympiakos, all this activity came at a cost, as the American guard fouled out. Hardly surprising, given the fact that among point guards only Jerel Blassingame committed more personal fouls in the regular season.

It is common sense that a good defender must avoid committing a lot of fouls. Dean Oliver’s emphasis on free throw to field goal attempts ratio (one of his ‘four factors‘) has further reinforced this point, but a more detailed analysis of defensive stats often tells a different story. Elite defenders like Khryapa, Sefolosha and Jamon Lucas foul a lot. Linehan’s impact on that game explains why this is not always a bad thing. His steals allowed Nancy to play efficiently at a high tempo – Dusan Ivkovic wants Olympiakos to push the ball as well, but this is not what he had in mind – and were one of the few bright spots defensively for the French champs: points per possession on pick and rolls and post ups for both teams indicate that when Nancy could not come up with a steal or a block, their chances of not allowing a basket were very limited, while Olympiakos’ half court defense was quite strong in this game. Simply put, Nancy had to run and Linehan helped them do just that.

The problem was that most of his steals were ‘on the ball’ – i.e. against the player he was guarding. This means that he was swiping at the ball quite often, which is exactly what got him into foul trouble. A less aggressive approach, however, would drastically limit his contributions – not to mention his + 16 on +/-. Each foul has a different meaning. This approach allows a better understanding of it.

Epilogue

On the original post we had asked you to take part in this initiative. From what I understand the response has not been overwhelming and understandably so: this project is time consuming and, if we’re honest, a bit geeky. But as new categories have been added, two things have become clear: first, there are so much more aspects of the game we could be monitoring, if we come up with workable definitions: for instance how does each team give up offensive boards? What happens when a player’s effort to steal the ball fails?

Second, thinking about basketball in these terms tells us more about the numbers on a conventional boxscore. There is no doubt that analysts, fans and even coaches (as CT commented on my first post) often spin numbers to fit their ideas or, even worse, predujices. This additional information offers a new perspective. So if you don’t want to monitor your team’s play in this way, it’s fine, but I really think you should try to look for these details that make all the difference in the world.

Thanks to everyone at www.redhoops.gr for their help in this project