Player DNA Part I: Roy and Rudy

When the basketball video game NBA Live 09 was released last fall, there were two additions to the game that reviewers were excited about (NBA2k9 has similar features): What the developer Electronic Arts (EA) calls Live 365, a year-round update of rosters and statistics via an online connection. And connected to this a feature named Dynamic DNA, a graphical breakdown of a players tendencies that among other things show on what percentage of possessions the player acts in isolation, as the pick ball handler, as the pick setter, as a post up player, as a spot up shooter, setting off-ball screens, and cutting. You might want to watch this nice trailer video to get a first taste of the feature before I'll try to explain more. 

The system comes with a default set of data deducted from the previous season(s), but only gets really interesting once the actual NBA season is under way so updates can provide realistic information on how the player behaves in real action on the court. Previously clumsy roster and player strengths updates have come a long way.

What can we learn from this data?

Honestly this is a post where I have to admit I'm not sure how useful it is. It shows tendencies what a player likes to do, not really how well he does it - that data costs extra :)
I'm also not sure DNA is the perfect term for this other than that it looks a bit like a visualized DNA sequence, since the "makeup" of a number of players changes tremendously over the year from the default state the developers thought they would have. Tendencies can change as a player progresses. They also wanted this feature to reflect how a trade to a new team would influence a player. Their example in the trailer linked above was Pau Gasol coming from Memphis to the Lakers.

[Little sidenote: Their pre-season simulation done for ESPN based on the default data from previous years straight out of the shop were not the most accurate - especially regarding the Blazers. All predicted league awards were wrong, Oden should lead the league in blocks together with Yao, the Blazers to come out on 10 while the Clippers make 7th place, etc. But those errors could be expected and they got most other playoff teams right. The product only gets really interesting when it gets "Live" with year-round updates of the ongoing season.]

Who collects and provides this data?

EA has licensing rights from the NBA, and they have partnered with ESPN to present/sponsor this information in the game. But the force behind this is Synergy Sports Technology, a venture capital funded sports analytics company that has specialized on indexing and deducting statistics from (live) video. It's quite an interesting shop: It's founded by a former member of the Phoenix Suns coaching staff and a man who developed video solutions that were acquired by Microsoft, and funded among others by Mark Cuban.

They provides many analytical tools, among them one called Digital DNA offered as a "fan experience solution".  You can be absolutely sure that teams know about these sources and many use them extensively, likely even much more sophisticated ones. For us it's just a handy interesting way to visualize this information. For them such data influences real decisions, simulating how plugging in a different player can change the makeup of a team or how to play against a certain team or player.

What does it show?

What percentage of possessions the player acted in isolation, as the pick ball handler, as the pick setter, as a post up player, as a spot up shooter, setting off-ball screens, and cutting. The same is also available to show the makeup of a team as a whole/sum of the parts as seen in the trailer above. I have not included that here in this first post. We might look at team differences later in the offseason.

And In the game there is also information if a player likes to go left or right, which might be handy for a team ("DNA scouting report". Hint: Von Wafer goes almost exclusively left dear Blazers). Unfortunately that data can't be accessed online/exported out of the game as far as I can see and I didn't have the time to start taking screenshots of those. Here is a brief video on NBA.com that describes the tendencies of Brandon Roy using this data.

What it does it not cover?

How good a player is at any of these things. Just how often he does it. I know there is information available on this (e.g. claiming Roy's effectiveness in isolations is 87%, which is near the absolute top in the league), but I don't have access to it.

D-E-F-E-N-S-E
Yes, that is either too complex for the common fan, or not even these stat geeks have figured out how to measure it. And if they did, it's probably better to commercialize it first with the teams and not the fans.

How accurate is it?

Since I have limited access I also can't verify how accurate the information is. I have seen somewhat different data. Example: With the NBA Live engine, Roy is supposed to have an isolation usage of 36% and Deron Williams of 28%. Another source also fed by Synergy Sports claims those isolations to be 29% and 22% respectively (and in general is lower than the Live data by about 6-7%). That's quite a difference. Maybe one database includes the playoffs and the other doesn't, though both claim to be "up to date" and that wouldn't explain the continuus difference. Probably one product has a finer granularity and differentiates more actions while they are lumped together here? I don't know.

Also, the team and player stats simulation aspect to predict outcomes is interesting, and needed in these games if you don't want to play the whole season - and nobody plays all teams. But like all statistics it hugely depends on the available data (see slightly off pre-season prediction described above), so it gets better and better as a season progresses.

On to the real stuff

Now that we have this introduction out of the way and set out the limitations of this data, let's take it for what it's worth and look at some of our players. Since I didn't want to overload this first post with dozens of graphics for the whole team or even just the starters, I'll focus here only on our shooting guards Roy and Rudy.

The difference in usage is very marked between those two players. Roy thrives in isolations and as the ball handler in pick and rolls (or in the Blazers system rather pick and pops, another granularity that is not reflected in the data), while Rudy predominantly works as a spot up shooter and uses screens more than Roy. We know that he is the most active off-ball player on the Blazers.

If you compare this graphic to the one above, Rudy is also a fairly extreme example of changing tendencies as opposed to the default data. The developers apparently expected him to be much more of a pick-and-roll player than a spot up shooter like he is now mostly used in the Blazers system. (Source: EA, Operation Sports game review)

External comparisons:

A few posters have hinted that Roy and Deron Williams are similar players only on different positions. And the data seems to support this claim. While Roy works a bit more in isolation Deron uses the pick and roll more frequently, which is understandable in coach Sloan's system and also playing together with Carlos Boozer and Paul Millsap a lot.

The same seems true for Dwayne Wade, also using the pick and roll slightly more but still being similar in tendencies.

A somewhat different shooting guard is Kobe Bryant. He posts up significantly more (a tendency that has increased as the season progresses - and likely will continue to do so as his legs get a bit tired after a long career) and plays even more in isolation (but has a slightly lower efficiency there at 84% according to Synergy Sports). In those regards he is more similar to Joe Johnson (whose graphic I won't show here. His percentages are 38, 19, 0, 13, 16, 9, 5) than to Brandon Roy.

Conclusion

So as I said initially I don't know yet what to make of this data. Just that it looks interesting to me when trying to find out about tendencies and playing style, when comparing players for trade decisions, and after the fact to see how they have changed. That's why I wanted to share this especially for people who don't play these games and have so far never been in contact with this new way to look inside of what a player does.

Sorry for following up my post about the draft lottery with this in short order. But I won't have as much time to write here over the next few days since I'm travelling, and wanted to get this Player DNA discussion started to see if other people find it interesting, and if yes to continue this with other players on our team and compare them internally and externally (LMA is an interesting case, so are our small forwards and point guards).

Have at it in the comments.