Please stop characterizing analytics as a misuse of simple statistics

By Stephen Shea, Ph.D.

February 13, 2015

It has now been several days since Barkley went on his anti-analytics rant on TNT.  In case you missed it, among other comments, he referred to Houston General Manager Daryl Morey as “one of those idiots who believes in analytics” and suggested that the use of analytics in basketball is “crap.”

Barkley’s comments reignited the analytics vs. traditional eye test debate.  Many working in analytics felt the need to again justify their existence, while those on the other side have used the opportunity to restate their position against the modern statistical movement.

To be honest, I don’t really care that Barkley and others around basketball believe analytics is crap.  What bothers me is that what they are criticizing and the examples that they use in their criticisms are not analytics.

During the TNT segment that included Barkley’s rant, Kenny Smith offered an example of what he believed indicated that analytics is flawed.  He talked about how he didn’t take very many 3s in his first five years in the NBA.  Then, after joining the Rockets, his 3-point attempts increased dramatically.  He suggested if you looked at his analytics in his pre-Houston seasons, you would say he’s not a 3-point shooter.  Then, he “became a 3-point shooter because [he] had a big beast-mode in the middle that drew two men,” which opened him up on the perimeter to take more 3s.  At this point, Shaq chimes in with “that’s why analytics don’t work.”

Kenny Smith’s 3-point attempts in a given season are a simple statistic.  That number alone is not analytics.  Looking at his 3-point attempts in his seasons before joining Houston and claiming he’s not a 3-point shooter is a misuse of simple statistics.

Smith’s description of how his production changed when the context in which he played (his teammates and his usage) changed is not an argument of why analytics does not work.  It is a description of the types of lessons learned through analytics.

Simple statistics, such as a player’s points or 3-point attempts, have been around for quite some time.  In the past, certain individuals used those statistics to make claims like “Kenny Smith is not a 3-point shooter.”  Those were flawed conclusions on incomplete information.

Sports analytics is a subject that grew out of frustration with the misuse of simple statistics.  It is a field that attempts to discredit exactly the types of claims that Kenny Smith and others are now classifying as analytics.

Barkley’s rant has reinvigorated the analytics vs. eye test debate.  Barkley, Smith and Shaq seem to see the situation as in the following diagram.

chart 1 2 13 15

However, I have never viewed analytics as opposing the eye test.  In fact, I’ve written, “Analytics are best used in conjunction with more traditional means of evaluating players, teams and strategy.”  (Basketball Analytics, 2014)  I see the eye test as an ally and the misuse of simple statistics as the enemy.

chart 2 2 13 15

Hearing comments like those made by Kenny Smith are like hearing an NBA fan say they rooted against the Lakers in the 80s because they didn’t like Larry Bird.

In the last few days, I’ve heard many individuals question why we use “analytics” and “advanced stats” as opposed to just “stats.”  During a recent Clippers/Rockets broadcast (where Mark Jackson discussed Ekpe Udoh and plus-minus), Jeff Van Gundy questioned the use of these newer terms when coaches have been using stats for years.  I can’t speak for the entire analytics community.  The reason I use these terms is to distinguish the current use of statistics from the misuse of simple statistics in the past.  Apparently, it isn’t working.

Of course, those on the TNT crew are not the only individuals with this flawed perspective.  Charissa Thompson, Brian Scalabrine, Doug Stewart, and Peter Schrager came together on Fox Sports to discuss the issue.  Scalabrine’s analogy to 21 Jump Street was amusing, but the conversation deteriorated after that.  Schrager brings up a 4th and 1 NFL scenario and implies that the analytics community would say, “We must run.”  He then adds that the play call might depend on which player is in the backfield (whether it is Marshawn Lynch, for example) as an argument against analytics.   Thompson contributes, “By the way, how many times in the season when they gave Lynch the ball on the 1-yard line had he scored?”  Stewart continues, “That’s exactly what we’re talking about.  You have to throw numbers out sometimes.”

The fact that Marshawn Lynch (I believe) scored 1 TD on 5 attempts at the 1-yard line this season is a simple statistic.  Looking at that statistic and deciding Seattle should not run Lynch in the same situation again (say at a crucial point in the Super Bowl) would be a misuse of that simple statistic.  Analytics sees that simple statistic and first notices the exceptionally small sample size.  It then tries to dig deeper.  Who were those attempts against?  What were the offensive and defensive formations?  What was the game situation?  What happened in prior years?

I suppose I shouldn’t be too tough on the TNT or Fox Sports crews for not knowing that analytics studies how statistical production (such as Kenny Smith’s 3-point attempts) varies according to context (such as game situation, teammates, opponents, and player usage).  It’s not like the analytics community has been publicly preaching these concepts for years.  No, we guard our ways with the secrecy of the Freemasons.  The anti-analysts would have to hire symbologist Robert Langdon to crack our Da Vinci Code.  And just like a Dan Brown novel, when they do find our treasures, the location will be mockingly obvious, like in Chapter 1: Production and Context of a book titled “Basketball Analytics.”

Analytics looks at the Houston team that Kenny Smith joined and sees a great interior offensive presence in Olajuwon.  Analytics then understands that Olajuwon will draw double-teams more so than other centers around the league.  This implies that Houston’s guards on the perimeter will get more open looks from 3.  Analytics then scans the league to find guards that don’t currently get to play with a player like Olajuwon.  Analytics realizes that these players aren’t going to see as many good 3-point attempts as they would see with Houston.  In other words, their simple statistics will not reflect their true 3-point shooting ability.  So, analytics suggests finding these players and putting them with Olajuwon and then enjoying the increase in 3-point production.

Analytics won’t tell us everything about the game of basketball.  Often, analytics agrees with what those around the game already knew.  Occasionally, analytics makes incorrect claims or predictions.  The true merits of analytics can be debated.  But, on prom night, when us nerds took to our parents’ basements to do analytics (and play a little “Dungeons and Dragons”), our objective was to vanquish the very thing that you now claim us to be.  So, PLEASE don’t confuse the misuse of simple statistics for analytics.

Leave a comment

4 Comments

  1. Nate

     /  February 15, 2015

    Everyone and anyone can claim to be doing analytics, and I’ve seen plenty of questionable reasoning under that banner.

    Reply
  2. Steve

     /  February 15, 2015

    I loved both the books and your blog too. Furter, I appreciate your level headed response to Sir Charles. Keep up the great work.

    Reply
  3. Matt

     /  October 26, 2016

    What I think some teams have looked at are bad teams from the past (Midrange shots). You have to be very careful in looking at that. Well, let me restate that. That tells you nothing about what wins and doesn’t win. That tells you that those bad teams had no scoring high or low. Midrange shots are not what made them bad. And here’s the key on that. Those teams were better off doing that. Why? The 2016 Philadelphia 76ers. They were NOT built to run a Moreyball-like format. Yet they did, and we land at 10 wins in an era where 10 wins are basically handed to you before the season starts. You can cherry pick wins (if you will). Golden State did so in their pursuit of 73 wins. It’s a low minute era, and teams rest players all of the time. Philly? I’ll go ahead and tell you they were the rest game of all rest games for all teams. Speaking of Golden State. They rested players in both of their games against Philly

    It’s not a one-size-fits-all deal. You may not be the right team for it. You may be facing a team that it doesn’t work against. There can be all sorts of variables from all different angles. It doesn’t work against everyone. It didn’t work against Golden State. I have charted their entire season.

    Reply
  4. Matt

     /  October 27, 2016

    On that note. I get that you don’t want to get too far outside of the percentages. At the same time, I see what is in my opinion going to heavy on the percentages in certain cases. In terms of Moreyball, nothing really has changed. The big play, close to the basket play, the more efficient still comes from the more dynamic players. It’s just no longer the 7 foot to shaq, or Duncan. It comes in a variety more from all positions now. Part of the problem with that, some of these bad teams are trying to simulate that which is dynamic. That’s a recipe for disaster at the professional sports level. Being close to the basket won’t make you a modern version of shaq. I posted on this and another one of your articles. And like I said in that passed article, those layup/3 percentages carry so much more weight over 82 games than they do 7. It’s just a matter of percentages, simple science.

    Reply

Leave a Reply

Your email address will not be published.