Consistent play is a concept that is incessantly talked about by broadcasters and pundits in the world of sports. How often do we hear an announcer talk about how a certain player needs to be more consistent for their team to win a game?
Other than “heart”, consistency is the most talked about aspect of basketball that no one has yet to measure. That being said, it’s actually fairly easy to come up with a basic way to look at consistency.
The following metric attempts to measure consistent effectiveness and ineffectiveness in players using John Hollinger’s game score and PER.
Like any stat, it’s not the end all, be all of all numbers, nor is it intended to be. However, it gives fans a gist of which players are consistently reliable (or unreliable). This can prove to be a valuable tool in terms of player evaluation.
John Hollinger’s game score is essentially the PER of a player’s single game performance. It has its critics, but it’s one of few statistics that measures a player’s single game performance.
By taking the standard deviation of a given player’s game scores for the entire season, one gets a number that tells us how much variation there was in the game scores. The more consistent a player was, the closer to zero the standard deviation of his game scores become.
However, simply doing this to measure a player’s consistent effectiveness is flawed. For instance, let’s take Kobe Bryant and Robert Sacre. Bryant’s standard deviation of game scores was 7.83 in 2012-13. Sacre’s was 1.83. Sacre’s standard deviation is closer to zero, so this means that his game score values had less variation than Kobe’s. This may be true, but it does not tell us which player was consistently effective.
So how do we separate the consistently effective players from the consistently ineffective ones?
We simply divide the standard deviation by the season PER. This gives us (in mathematical diction) the coefficient of variation. It tells us how consistent a player is RELATIVE to his own average.
After doing this operation, it becomes clear and obvious that Bryant is far more consistently effective than Sacre. Bryant’s coefficient of variation is 0.34, while Sacre’s is 0.54.
Here are the consistency ratings for the Los Angeles Lakers from this past season.
2012-13 CONSISTENCY RATINGS (LA LAKERS):
- Once again, the higher the coefficient of variation is, the more variable a player will perform relative to his average.
- Jordan Hill is the most consistently effective player on the Lakers. He has the lowest coefficient of variation. Again, this does not mean he is the most EFFECTIVE player on the team. It only means that one can expect Hill to play at a level near his 18.5 PER on a consistent basis.
- Steve Blake, Jodie Meeks, and Metta World Peace are great examples of players who would put in an amazing performance one night, then follow it up with a horrible game. We never knew what to expect from these guys. All of their coefficients of variation exceed 0.44.
- Another interesting observation is Dwight Howard’s coefficient of variation. He was fourth on the team this year with a 0.377 coefficient of variation. That’s great, but it’s a huge rise from his final year with Orlando, when his coefficient of variation was 0.322.
- To add more perspective to all of this, though, LeBron James’ coefficient of variation this season was a miniscule 0.19. Kevin Durant’s was even better at 0.18. Both of these guys weren’t consistently effective – they were consistently dominant.
- The fact that the Lakers didn’t have a single player below 0.30 shows that the team was plagued with inconsistencies this season, which was obvious for anyone who watched the team this year. However, now it’s quantifiable.
- This metric works well for lesser quality players who don’t see much floor time, too. Because the denominator (PER) is small for these types of players, the coefficient of variation often turns out to be large for guys like Sacre and Devin Ebanks. This tells us that they’re consistently ineffective when they play. If they DID play well, their denominator would be larger causing their coefficient of correlation to decrease.
Taking the standard deviation of game scores for guys like Earl Clark who sat in the back of the bench and rarely played early on but then became a regular can be a bit convoluted. Clark rarely played early in the season and as a result, received low game scores. In the second half he played well and became an integral component for the Lakers. As a result, up went his game scores. The variety of numbers can inflate his standard deviation and make him look like an inconsistent player when really he just wasn’t playing much in the beginning of the season. It may be more effective to only include games where a player plays at least 12 minutes. But then we ignore performances from guys who played less than a quarter and made an impact on the game. Although rare, it’s unfair to ignore this. Therefore, we include the garbage time games for Clark early in the season and penalize him for not being a regular. Again, no stat is perfect, especially in this analytics era.
This metric helps us break down players into four different groups:
- High PER, High Standard Deviation: This is Dwight Howard and Kobe Bryant. They have high PERs, but can be inconsistent at times.
- High PER, Low Standard Deviation: This is Kevin Durant and LeBron James. This is where everyone wants to be. These are guys who are consistently dominant.
- Low PER, High Standard Deviation: This is Jodie Meeks and Metta World Peace. They can put up big games, but more often than not they’re going to be average or below average.
- Low PER, Low Standard Deviation: This is Robert Sacre and Devin Ebanks. They are consistently ineffective.
The following table illustrates this notion:
This season, the Lakers had a lot of guys who had a PER below 15 (which is the league average) and a game score standard deviation greater than five. They were an inconsistent ball club.
Now while Durant, LeBron, and Kobe all had game score standard deviations greater than five, they made up for it because of their efficiency on the court. In other words, a standard deviation of five or six is not bad for super stars like LeBron or Kobe. Unfortunately for Kobe and Howard, their game score standard deviations hovered over seven. They would like theirs to be below six.
It’s all relative. That’s why it’s important to look at the coefficient of variation at the end of the day to see who consistently plays up to their own level and as previously mentioned, the Lakers didn’t have a single guy who had a coefficient of variation below 0.30.