Maturity of Roboscout
by Dean Oliver, 03/28/05What you look for are tendencies.
And what you want from those tendencies are the most important ones. It’s easy to say that you need to stop everyone and here’s how, but simplicity in coaching dictates identifying what is most important and what is vulnerable. That prioritization is critical because in basketball you almost always give something up in order to take something away. Knowing how to take the most while giving away the least is what gives you an edge.
So this article is about scouting. I was a scout for years, doing it the way that it’s traditionally done – watching tape over and over, going to games, and writing out 40 page reports for the Bertka Views scouting service. But as I worked on even the fourth page of those reports, I knew that I had gained a sense for the more important things in the report (which were almost never the side inbounds play that we also documented). And I always wanted some statistical measure from all games to check on what I was seeing from the one or two games I watched. That statistical measure ultimately came about 8 years later from a tool I call Roboscout, an artificial intelligence program I don’t sell or give away (which I have to say because people have repeatedly asked). It is a tool that suggests what is important to shut down in an opponent. For instance, should you stop Michael Jordan or his teammates? Should you double-team Shaquille O’Neal or single team him or foul him a lot? Roboscout provides some insight into these things.
Roboscout is a useful tool that pulls information from as many sources as asked for. One of those sources currently is 82games, Roland Beech’s collection of unique data and, of course, home to this article. One key to understand about the 82games data is that, though Roland shows the stats as associated with lineups or with players, the data actually is about a lot more than lineups and players. They reflect style and strategy.
So what specifically does this tool say? One of the more interesting ways of demonstrating what it says is to look at the past, especially a team that most people have some memory of. In this case, that will be the 1998 Chicago Bulls, the last of their championship dynasty teams. The Bulls went 62-20 and beat the Utah Jazz in six games in the NBA Finals. The team was vulnerable, unlike the 1996 version that went 72-10, as unbeatable a team as there ever was. Behind Jordan, they had Scottie Pippen (for 44 games due to injury), Toni Kukoc (who took some of Pippen’s time), Ron Harper, Dennis Rodman, and Luc Longley. Coming off the bench were Steve Kerr, Scott Burrell, and Randy Brown.
The basics of the 1998 Chicago Bulls were their offensive and defensive ratings, 107.7 and 99.8, versus a league average of 105.0. (If you don’t know what a rating is, it is points scored or allowed per 100 possessions, the most fundamental measure of team efficiency.) They struggled a bit offensively, but made up for it with a fine defense. Not surprisingly, Roboscout sees more holes in the Bulls offense than in their defense. (This is also due to better stats on the offensive side than on the defensive side, something that will hopefully be remedied in the near future with improved defensive stats.) So we’ll start with the offense.
The key things that Roboscout saw in the Bulls’ offense are as follows. I’ll list them then talk about a couple things afterwards.
Rather, the individual indicators are a little stronger. First, the importance of different shooters was pretty distinct. On this team, it was very clearly Jordan who was most important. You might say, “Duh, that’s obvious.” But it isn’t always true that a team’s top scorer is the most important to the success of the team. It is true fairly often, but there are always several teams per year for whom it’s not true. In the Bulls’ case, it was universally true across time that Jordan was the most important person for field goal percentage. It was true even in 1996 with the record-setting team, but it was more balanced than here. On this team, stopping Jordan was about 60-65% more important than stopping the next most important. On the 1996 team, it was only about 30% more important. This reflects somewhat on whether you stop Jordan or the Jordanaires. This first indicator says to stop Jordan. The second indicator is whether Jordan should be taking more shots, which is reflected in the third point…
The third point emphasizes that it was Jordan and not the Jordanaires that drove the offense here. Take away his shots and make sure he is taking tough ones. There was a clear case for a defensive stopper and/or double-teaming. This situation doesn’t happen too much in the NBA (or even in college). Usually offenses show an optimal balance because, over the course of a season, defenses will dictate a kind of equilibrium scoring distribution. But on this Bulls team, there was a clear advantage for the Bulls to having Jordan take a higher percentage of shots. When teams took him away, it hurt that team. In contrast, the 1996 Bulls team did not have this tendency. If you tried to keep him from shooting, other guys would pick up the slack. (This is frankly why I chose to write about the 1998 Bulls – they actually had weaknesses that the 1996 team didn’t have. The Roboscout report for that team was “Spike their Gatorade and pray that they get bored.”)
Also under the third point is something I said so often back in those days as a Bulls-hater – don’t let Steve Kerr shoot!!! You didn’t have to do too much with Kerr, just keep him from getting an easy look. If Kerr was shooting, you had a problem because the guy only took smart shots. (Because the importance of his field goal percentage was so low, you didn’t have to worry about them going in so much – they usually would go in – but you didn’t want to give him many looks.)
On the other hand, you could give the extra shots to Rodman and Longley. There was some concern in the 1990’s about double-teaming off of Rodman because he’d go to the offensive boards. By the 1998 version of the team, this was less of a concern.
Longley was actually an interesting case. You wanted him to take more shots. His field goal percentage wasn’t of much importance and, from point number four, his turnovers hurt the team more. This basically meant to double off him, but if he got the ball, go for the strip and, if he managed to get a shot up, no big deal. That was a nice little hole in the offense to exploit… that teams didn’t exploit enough!
Mentioning Kukoc under turnovers in the fourth point reminds me of something. Kukoc was occasionally called a “key” to that team. He wasn’t as “key” as Jordan, but he was probably second in, uh, keyness. His shooting was as important as Pippen’s and his turnovers were more important. He wasn’t as good as Pippen, but he was more “key”.
Finally, the last point about their offense was that two guys were more important to keep off the line – Jordan and Pippen. And it was much more important with Jordan than Pippen. The importance of keeping Jordan off the line is probably the largest I’ve seen. Yike, it was worth a couple points a game to do better than average at keeping him off the line. He was getting old and tired (in a way we all wish we could be old and tired), so getting calls was much more important for him by this season.
Against the Bulls’ defense, the keys were not as clear:
So that’s what Roboscout says these days. His vocabulary is growing. His skills are broadening. He has no girlfriend to distract him. He’ll be taking summer classes with 82games and doing all his homework. I’m so proud. I have actually checked his work by watching games myself to see if his judgment is, well, like mine. But I’ve also passed his recommendations for teams I’ve never seen (a few college teams) on to scouts who have seen the teams and they have come back to say they were thinking the same things. So he’s a good boy.
Even with Roboscout, I always think back to what I first heard about UCLA coaching legend John Wooden. Coach Wooden believed in maximizing his own team’s talents over scouting an opponent, something I believe strongly in many cases. This is especially a good idea when you are a favorite, as Wooden’s teams often were. It’s also a really good idea early in an NBA season when you’re building your team. You have to establish your own identity first before trying to adapt to others. But if you’re fairly well established as a team and you are an underdog, you better know your opponent. I hope Coach Wooden would approve.
Copyright © 2005 by 82games.com, All Rights Reserved