One of the biggest things that has risen during the analytics age is the notion of “clutch time.” That is when the game is within five points and there are five minutes or less left in it. Which players, then, are the best when it matters most?
At NBA.com/Stats, they track this, and that was the source of the data I used for this piece. As with any statistical analysis, however, it’s never as simple as combining stats. And the constant effort to try and find a single-number metric can be annoying. They give the semblance of “objectivity” because they have a “number” to measure, but there are all kinds of subjective evaluations in determining how those numbers are derived.
In other words, an “assist” might get a certain value that gets output in an “objective” way, but what was the value? The same goes with blocks, steals, rebounds and so on. How much each is “worth” relative to the other is based on the opinion of the author of the formula. So while the final number spit out might have the veneer of objectivity, there is a subjective aspect to it.
I don’t mean that it’s an uninformed opinion or even an illogical one—but it is still just that. And that’s why even with the myriad single-number metrics available to us, no two give the same results. And then when you try and pull it apart, you have to go back to the actual numbers.
So these have a sort of “smell-test” quality to them. They’re useful, but only to an extent.
So I’ve been considering: How can you evaluate one player to the next with the benefit of a single-number metric, but without the downside of trying to compare relative values. And the answer may lie in a visual representation more than a number.
Last year, Seth Partnow introduced the idea of using radio charts in comparing point guard types for Nylon Calculus. Ian Levy, writing for the same site, used the methodology to compare offenses in a playoff preview. Most recently, I used them to look at scoring types for BBallBreakdown.
The advantage to them is that they offer a more “three-dimensional” approach to analysis. It allows for you to “see” the different “shapes” a player can take while at the same time offering a comparison. It allows for more than just seeing “better”; it allows for illustrating “different” too.
For the purpose of this comparison, I used a three-point approach. In all these areas, they’re based on per-100 possession figures, with a minimum of 50 clutch minutes played to qualify. Here are the categories and how I came to the number for each one.
- Adjusted Scoring: As always with such things, the great conundrum here is distinguishing between the chucker and the efficient scorer.
- If you count just raw points, you reward the former; if you go by percentages, then it tends to undervalue the chucker. After all, the points they score are still real, and not all teams have the same quality of player around them.
- The distinction between regular season LeBron James and Finals LeBron James proves that there is a connection between quality of teammates and being a chucker vs. high-efficiency player. That said, there are plenty of chuckers who are going to be chucking regardless of who their teammates are.
- So, to balance things out, I looked at the number of true shooting attempts a player used (FGA+.44*FTA). From there, I determined how many points an average player would score using those attempts. Then I determined the difference between the two and split it in half, assuming that half the responsibility for said difference is on the player and half on his mates.
- I then subtracted that difference from the player’s scoring total. So, if a player “cost” his team 10 points by throwing up the ball with reckless abandon, I penalized him five points for it, thereby not rewarding him for bad behavior, but also not completely negating the positive effect his scoring had.
- I then added one point for each steal and subtracted one for each turnover.
- Points Created by Assists: For assists, I just estimated how many points a player added through his passing. Since the splits for whether dimes were for two or three during the clutch aren’t easily available, I just used the average points per assist based on the Sport VU data (Points Created by Assist/Assist). So, if a player averaged 2.5 points per assist and had five assists in the clutch, he had 12.5 “Points Created by Assist” in the clutch. I
- Rebounds and Blocks: This was just straightforward adding rebounds and blocks. Nothing fancy here.
Now, here’s the fun part: By determining the area of the radio charts, I can rank the players without having to compare points to assists or rebounds. Rebounds get compared with rebounds, points with points and assists with assists. You don’t have to worry about relative values because they are each their own unique values.
But, at the same time, we can conclude that the bigger the chart is, the better the player is. But it doesn’t just give you that, it gives you the texture of the player’s performance. For example, here is the chart for James Harden:
Compare him with Anthony Davis:
You may notice that Harden’s “area” is massively bigger, but don’t worry about that. It’s the nature of area to go up exponentially. It’s one of the reasons why I’m cautioning against reading too much into the actual number (along with the fact that the whole point of this exercise is to get away from idea of the numbers).
Rather focus on the shape of the triangle. Notice how it’s the near inverse of Harden’s.
Now look at Ty Lawson’s chart:
Lawson did very little in the way of rebounding, but he was huge in creating points by passing.
All three of the players above were tremendous in the clutch, but the charts show how they were big, not just that they were big, and it does so while maintaining a bit of ranking structure.
So here are some of the more popular players’ charts, starting with LeBron James:
Here’s Russell Westbrook’s triple-threatedness charted out:
And you have to include Kobe Bryant:
And here are the stats and respective areas for last year’s top-25 performers in the clutch. See how Tim Duncan, Dirk Nowitzki and the Mamba are still getting it done when it counts: