Does Anyone Understand Advanced Stats?
This past week we saw…
1. Moneyball the movie comes out and depicts owners as fools that don’t get numbers.
2. Henry Abbott wrote a piece confirming this.
3. Zach Lowe commented on Henry Abbott’s piece with the following line
That last one bugs me, because blocks per minute is not an advanced stat. It is the same thing as blocks per game, except you are putting minutes in the denominator instead of games.
Wait, what? Blocks per game is not the same as blocks per minute because player’s minutes vary by game. Some players play more minutes per game than others. Here’s a simple example: Serge Ibaka and Dwight Howard have the same blocks per game but not per minute!
- Serge Ibaka: 82 Games, 2.4 Blocks per Game, 0.089 Blocks per Minute
- Dwight Howard: 82 Games, 2.4 Blocks per Game, 0.063 Blocks Per Minute
The obvious key issue here is our ability to compare players. In discussing whether or not Ibaka or Howard are superior with respect to blocked shots, one has to control for the fact Howard gets more minutes. Consequently, the per game metric will favor Howard. But when we remove the bias we see in per game numbers (i.e. use the per-minute numbers) we see that Ibaka is clearly superior at blocking shots (Scorecasting further complicates this by showing that various blocks are more important than others).
Stats aren’t always simple, and it’s very easy to get into weird debates about which stats qualify as “advanced”, which stats to use, and even if people are using them correctly. When I read articles like these though, I think it is important to take a step back and look at something else.
What do Stats Tell You?
Zach and Henry Abbott (and indeed others) have scoffed at GMs when they consider simple things like per minute stats “advanced”. Basketball-Reference has been showing these for years and it is the first hit in Google for “Basketball Stats”. Rather than squabble over stats being easy or advanced though, we should ask what they tell us. Per minute stats are great. They tell you how well a player is doing in an easy defined range that makes it easy to make comparison across players. In fact one of my first stat scans for any article is to use Basketball-Reference’s Player Comparison finder and look at the Per 36 Minutes Table (for example Howard vs. Ibaka).
When you use a stat it’s not important if it is “advanced” or not, it’s important if it tells you useful information.
- Does blocks per minute give us useful information? Yes!
- Is it a better measure than blocks per game? Yes!
- Does a player’s Adjusted Plus Minus or Horoscope give you any useful information? No!
Yes, some people think APM is “advanced” (and someone might feel the same about a horoscope). But who cares if a subset of a community has deemed APM “advanced”?
Summing Up
My fear when I read the Lowe piece was two fold. The first was I didn’t see a heated comment trail debating if Lowe meant blocks per game for a single player or a population (let’s give him the benefit of the doubt). I didn’t see claims saying “so what if blocks per game is simple, it might be useful”. I did see a quip about using simple stats vs. advanced stats though. The important question in our analysis should always be “What do my numbers tell me?” and if that is useful information than use it! If it turned out the solution to making a good NBA team was a single number that you could calculate, I’d say use it! Who cares if there are other metrics that are more complicated and require more data? Making something “complicated” doesn’t necessarily result in something that is useful.
I like to think of myself as a member of the advanced stats movement in basketball. That being said, though, I think I might prefer to relabel myself as being part of the useful stats movement.
-Dre