Statistics and the Human Element

 by Jon Nichols


Advanced basketball stats have come a long way.  If that point isn’t clear to you, check out the player pages at and the wealth of information contained within them.  Or head over to and get lost in the numbers.  Or just read any of the posts at the APBRmetrics message board.  Beyond this public data, teams are doing analyses that we probably have never even considered and know tons about players through simple box scores and play-by-play data.

How far can we go?  Can we truly measure everything that goes on in a basketball game?  More importantly, can we use those measurements to evaluate individual players and project how they will do on other teams?  As crazy as that may sound, I would bet there are a few people out there that think it’s possible.  Play-by-play and game tracking data open up many possibilities.  Throw in shot charts and you can come up with a whole bunch of different numbers.

But in my mind, the human element will always play a large role.  To many teams, it still may factor into 95% of their decision-making.  While I do think those teams are missing out on a great trend that’s now been around for a few years, there is some logic to trusting your eyes.  The big thing is knowing when to use your numbers and when to use your scouts.

My hypothesis is this: numbers do the best job of conveying how a player has performed in the past and how he will do in the immediate future, since this will most likely be dictated by his past performance.  If a guy has been shooting 45% on three-pointers this season, there’s a good chance he’ll make some long range shots against you if you leave him open. 

But what about projecting a guy’s performance five years from now when he’s on a different team with a different style of play and a bunch of new teammates?  Or how about projecting a college kid or European prospect who has never experienced NBA basketball?   Suddenly, a scout and his eyes have all the power. 

These ideas hit me hard at the MIT Sloan Sports Analytics conference this past weekend.  Ironically, even though one of the major goals of the conference was to discuss new analytical tools professional teams are using to do business, I became convinced that one of the oldest tools in the book still might be the most useful at times. 

The panel discussion on talent identification sure was persuasive.  According to the program, the purpose of the panel was to “discuss some of the existing, as well as new innovations in discovering the new talents of tomorrow.”  As it turned out, the discussion included some great analytical methods using complex data.  Kris Homsi of SPARQ and Aaron Schatz of are quantifying the seemingly unquantifiable in terms of training athletes and professional football, respectively. 

Despite their innovations, one of the main things that came out of their discussion was the importance of game intelligence, passion for the game, and hard work.  As someone said, the best players in any sport are usually some of the hardest workers.  Similarly, the more you like something, the better you will be at it. 

In addition, scouting isn’t as subjective and random as you may think.  Jack Mula, formerly of the New England Patriots, discussed how the team always used “systemic scouting” to establish their dominance.  With this system, the coaches made it clear what they needed from each position, and it was the scouts’ job to find players that met this criteria. 

Can stats attempt to replicate this?  Sure, they can try.  In fact, the PAC data I developed is supposed to be used almost like a scouting tool, allowing you to classify players using numbers (a basic version of systemic scouting).  But since so many things are literally intangible, the numbers often let things slip through the cracks.  That’s why experienced professionals using knowledge and game tape can sometimes paint a better picture.

Besides a player’s actions on the court, it’s very important to measure his character and other psychological aspects.  As I mentioned earlier, things such as hard work and game intelligence often lead to great success later on.  It would be foolish to think that statistics can somehow measure this type of information.  After all, it is the human element.

To sum it up, it’s all about balance.  Teams that are relying solely on subjectivity and scouting are probably making a big mistake, and the same thing goes for teams that only use numbers (although I doubt any of those teams exist).  The premise of Composite Score is that when multiple people have different ideas about the same thing, they’re all probably right about some of the issues, so you need to combine all of their knowledge.  The same thing goes for the interaction between scouting and statistics. 






Copyright © 2009