(An understanding of advanced stats is helpful in reading this post. I’ll try to explain them as best I can, but for a more complete and in-depth guide on advanced stats, read A Layman’s Guide to Advanced Statistics by Mike Kurylo.)
This season, the Memphis Grizzlies made one of the more interesting front-office hires as they named John Hollinger their Vice President of Basketball Operations. Hollinger was, of course, a noted number cruncher working for ESPN as an analyst and a writer before breaking into the NBA with the Grizzlies.
Fast forward to the present day, and Lionel Hollins is no longer the head coach of the Grizzlies. Supposedly, one of the reasons for his departure was that the Grizzlies wanted to incorporate more use of advanced stats into their approach, whereas Hollins didn’t. Hollins appeared on ESPN’s First Take, and said that the Grizzlies had already been using some level of statistics in their approach. However, Hollins preferred to use what he saw on the court in building his rotation, which included things that couldn’t be measured through advanced statistics–things like aggressiveness, heart and hustle.
In the most recent Beale Street Bears’ podcast, featuring yours truly, advanced stats was a topic of discussion for Hal, Daniel and I. We agreed that some level of advanced statistics would be helpful towards building a successful team, but there should be some level of balance between advanced stats and the product on the floor.
At the most basic level, statistics are just a count of what happens on the court–points, rebounds, assists, you know the deal. A coach could see what the statistics say about a player, game, or season without actually looking at the numbers on a piece of paper. That said, recording them as statistics make them infinitely easier to manage.
Advanced statistics are a little more difficult to interpret from just watching the court. Things like per-40s and pace-adjusted stats are one thing in that they allow us to see how, say, a player playing very few minutes in a low-tempo system could be a better rebounder than a player logging a lot of minutes for an uptempo team. Certainly, those things could be hard to judge when in most cases, players that play starter’s minutes are better than players sitting on the bench for most of the game. However, those are just extrapolations of the basic stats–they offer a different perspective, but it isn’t completely new information.
It’s when we delve into things like True Shooting Percentage (TS%) or Hollinger’s Player Efficiency Rating (PER) that we enter a completely new dimension of metrics. These numbers utilize basic statistics together to create a new number–one that isn’t a count of something, but is instead a function of those numbers created through a designed formula. They manipulate the numbers that are a statement of fact to reach a conclusion, similar to how I would manipulate different body paragraphs to reach a conclusion as a writer.
Advanced stats play an important role in the NBA. They allow us to see things that we may not be able to infer from basic stats or even per-40s/pace-adjusted stats. True Shooting Percentage allows us to identify the most efficient scorers by adding three-point percentage and free throw percentage to field goal percentage. Player Efficiency Rating shows the overall value of a player through combining all per-minute and pace-adjusted stats. And, stats exist for so many different things. Lineup stats, rebounding stats, usage stats, the list goes on and on.
With all that said, there’s a reason teams haven’t jumped to advanced stats as their be all and end all. Numbers can’t track everything, and this starts from the basic stats. When those basic stats are used to create advanced stats, there’s no way to account for the things left untouched by basic stats. Stats can’t show us which players can set a good screen or inject energy into their team. PER is flawed because it’s a measure of overall value, yet it can’t measure everything. And, because it’s a formula created from scratch, human error can lead to imperfections. PER has been criticized for overvaluing high-volume shooters.
If all we knew about the NBA were the stats, then Tony Allen would only be a NBA-level player for his steals. Yet, in reality, Allen is one of the most important players the Grizzlies have. He embodies their team’s identity, and brings out the best in his team in a way very few players can.
It’s by reading the court that we can see how Allen is valuable to the Grizzlies. By reading the court, Lionel Hollins sees that Allen carries more value than Jerryd Bayless or Austin Daye, players with a higher PER. And, it’s very likely that his decision to play Allen over those two correlates into a large number of wins for the team.
The question is, how do we take advantage of both advanced stats and reading the court so that it translates into as many wins as possible?
That equilibrium is not easy to find. Different people will have different answers, and for different teams, different answers will work.
The San Antonio Spurs have built a model of success through intelligent team-building. Advanced statistics came second to the team concept and intangibles, things that can’t be measured through stats. Many of their players aren’t necessarily elite players if you were to look at various advanced stats. However, it works for San Antonio because those players fit the puzzle.
Advanced statistics are hugely helpful in evaluating player talent. It allows you to see that few players are more efficient than Kyle Korver (TS% of 63.7%, 6th in the NBA). It allows you to see that Brandan Wright (PER of 21.03, 20th in the NBA) might be more helpful to the Mavericks if they played him more than just 18.0 minutes per game.
If you were the general manager of a team, advanced statistics could show that Kyle Korver or Brandan Wright might be players to target. That said, a coach can not and should not forgo the information they can glean from reading the floor. That’s going right back to the Tony Allen example. While advanced stats allow you to evaluate player talent, Tony Allen and the Spurs’ model of success show us that player talent isn’t everything.
In that regard, we could say that to strictly follow stats would be to go down a path that doesn’t necessarily include wins. Stats can be a trap, or a diversion from success. As a measure of information from the court at the most basic level, stats are contextual to what happens on the court–they are the dependent variable. At an advanced level, they tell us things that aren’t obvious. However, to mistake “y” as “x” would be to play Austin Daye over Tony Allen–not necessarily a recipe for success.
However, evaluations made from watching games aren’t absolute the way stats are. Where I might see good pick-and-roll defense, someone else might see a weakness that wasn’t exploited by the offense. Evaluation is subject to the evaluator, unlike mathematically proven metrics. If stats are the dependent variable, it may be safe to conclude that advanced stats should come second to good and complete basketball evaluation. That’s not easy, however, when “good basketball evaluation” can’t be explicitly defined.
Having a good coaching staff that can make astute evaluations is key in building a team. They understand how to develop players, how to distribute playing time between lineups, and how to get the most out of their team through different schemes and sets. Good scouts make sharp evaluations of which players to draft or trade for.
Reading the court has been the method for success ever since the NBA started. They are the best measure of the product on the court, in that they see things stats can’t and put it all together. So, what role should stats play in all of this?
The power of statistics lies in the perspective they offer. While they can’t track everything, they can track things even the best basketball evaluators miss. Even the best coaches and scouts can miss things the stats reveal. J.J. Hickson saw very little playing time during a brief tenure with the Sacramento Kings last season, backing up Jason Thompson at power forward. The Trail Blazers let Hickson loose, and he finished with the 33rd PER in the NBA this season (19.71) while Thompson ranked 157th (14.64). Would the Kings have been better served playing Hickson while they had him?
So, with the understanding that both statistics and human basketball evaluation have their flaws, the best method to take advantage of them both may be to use them both before making any conclusions. Because of what statistics miss, they come second to good old fashioned basketball evaluation. Trust statistics, and use them extensively to offer a different perspective from evaluations made by coaches and scouts. That said, be mindful of where stats can be misleading.
In firing Hollins, are the Grizzlies planning to start Austin Daye over Tony Allen next season? I doubt that’s the case, but it’s clear they want to use more statistics in their approach. Hollins was a good coach–he understood how to run his team, and it led to a successful season. It was factors unrelated to advanced stats that led to his team’s defeat in the playoffs, including a lack of three-point shooting.
If the new coach buys into advanced statistics and the Grizzlies begin to make more decisions based on advanced statistics over the intuition of the coach, they would be making a mistake. What Lionel Hollins did worked well and advanced statistics should not replace Hollins, but help him. If he was against the use of advanced statistics, it would be the front office’s job to persuade him.
A good system of long-term winning is hard to build up. There are a lot of things to balance, advanced statistics being just one of them. With all that said, and even with the fact that advanced statistics don’t necessarily come first in building a team, advanced statistics are part of a new age in basketball. As time goes on and teams learn how to make best use of them, advanced statistics will prove to be highly beneficial towards creating a true dynasty.