Pages Navigation Menu

A Chart Inspired by Stan Van Gundy

USA Today Sports

USA Today Sports

At the Sloan Sports Analytics Conference last weekend, Stan Van Gundy aptly replayed his role as anti-numbers curmudgeon, turning the verbal sprinklers on the audience of sports stat geeks and their affiliates, much to said audience’s delight.

At this point I am nearly Sloan-ed out, and skipping most of the #SSAC14′s in my timeline. But, a couple of things related to the Basketball Analytics panel, hosted by Zach Lowe and thoroughly stolen by Van Gundy, have stuck in my mind.

Van Gundy rightly mocked the stat blurb in the ESPN magazine Analytics issue that charted how many miles Paul George had run to that point in games this year. Without context or further analysis that’s barely a factoid, much less ‘analytics.’ The distance stat comes from SportVU player tracking which dominated the basketball analytics sessions to the point that play-by-play data started tanking.

While the SportVU-based papers and presentations were interesting, most of them just aren’t there yet. With the possible exception of the winning paper on the three dimensions of rebounds by Second Spectrum, none of the player tracking papers were ready for an analyst or team to dig into them and make any sort of cogent analysis or decisions.

A good part of that is the contrast between the breadth of player tracking data coming in and the depth in terms of years with less than one season of data in every NBA arena. There are many, many possibilities, but not enough cases of each to tell if the new data is actually superior to the existing.

And that brings me to the chart I wanted to share with Coach Van Gundy and everyone else:

Data and Opionion

Ok, don’t get too hung up on the shape of the curve or the size of the respective areas, this an example only.

Statistically the slope of the trade off between data and subject expertise would be determined by the variance in the data, the more variance the longer subject expertise would retain its relative importance. A third axis could also be added for complexity and ineffability.

It’s a chart I’ve had in mind ever since reading, Everything Is Obvious, a book I’ve referenced before. In that book author Duncan Watts relates how we have a tendency to fill in the story after an event has happened in ways that makes it seem as though that was the only possible outcome. In many instances in business or history, events happen only once, but in retrospect take on an inevitability mythology. When in fact if time ran one hundred different times the business may only be a success 72 times or that battle would only have been won 47 times.

None of that is a surprise to sports fans or coaches who get to observe players shoot free throws hundreds of times a season, a most basic play that has a variable outcome.

Watts points out that in dealing with small numbers expert subject matter opinion, especially when focused on process, can help us a great deal in sorting out the true probability of an event.

For example, if a player shoots and misses two free throws a coach with knowledge of shooting mechanics can give you a much better idea of the player’s likely ability than that small sample. On the other hand, while no coach teaches anyone to shoot free throws the way Shawn Marion does, he’s a career 81% free throw shooter over 2,800 attempts. After that many attempts no shot doctor would try to change it.

%d bloggers like this: