Not-So-Quick Thoughts on "Advanced Statistics"
by Gils_Keloids on May 2, 2009 10:13 AM PDT
25 comments
[Ed. Note: Gils_Keloids with an awesome discussion of the place and value of statistics in evaluating and comparing individual players. My favorite line: "Just because you are using numbers does not make it any less subjective..." A must read!]
I think you'll often hear me say (and others share this opinion) that while team statistics are useful and can be analyzed to give you insight into teams' strengths and weaknesses, what they do well and what they don't, the individual statistics that give player ratings and such are much less useful and are, in the end, subjective.
In fact, evaluating a player based on statistics derived from team statistics like basketball will always be a futile exercise.
Think of it this way.
You have a car, any car you like, a Prius for those environmentally conscious folk, a Ferrarri for those of you having a mid-life crisis, or a Mini-Van for the soccer moms and dads.
You can gauge how far your car has gone using the odometer, you can figure out the mileage per gallon, and sometimes if your car is fancy enough, you can calculate the average speed. Depending on how much time you are willing to invest, you can get all sorts of measurements of your car's performance.
Now, someone can take all those measurements, adjust them for city or highway, make all kinds of allowances. And then, someone could compare those numbers on paper and get a very good idea about which car was faster, had better mileage, etc.
Just because you are using numbers does not make it any less subjective
Now what if I told you to take those numbers and tell me: How much did work did the cylinders in the engine do? What about the pistons? Separate out the value of the sparkplug and valves, please. How much did the wheels add to the distance driven? Give me a numerical ranking of the steering versus the antilock brakes.
You would look at me like I'm crazy, right?
Well, most people would. Some people actually try to do it, and they come back with a calculation that splits all the contributions of the parts into a number, or rating. And they say, look, it adds back up to the mileage + avg speed! And when I do it for different cars, it adds back up! It must be right! And look, the parts of the engine always have the highest ratings, so it seems right. Now I can compare the cylinders of one car versus the sparkplug of another and tell you which is better! I'm simplifying or course, but the point is that if the concept of the task is absurd, any calculation you come up with is moot.
It all sounds silly doesn't it? But that is exactly what people try to do with basketball and individual ratings. It's a futile task. Evaluating individual players in a flowing sport with a large team dynamic component is, in the end, a subjective task. Just because you are using numbers does not make it any less subjective, due to the nature of the numbers.
Arguing which player is better is a bar fight, and will always be a bar fight. I've seen more and more people introducing individual statistics as if they were bringing objectivity to the debate. Really, It's just more subjective judgments and opinions disguised in numbers (and I love numbers).
4 recs |
25 comments
|
Email
Print
Comments
Thank you.
Like PER. I hate it when someone uses PER as a criteria of judging who’s better (and it has nothing to do with where Kobe ranks on it), because it’s only one person’s formula opinion of what statistics he feels is relevant to him.
I totally agree with you about PER
according to PER, David Robinson was a better player than Wilt Chamberlain. LOL at the bullshit.
I hate when people use stats to argue they never look at the intangibles
4 wins down.... 12 more to go
this article
would make john hollinger cry. have you ever read an article he did that didn’t bring up PER and/or some other stat(s)? i haven’t. the man is a complete stat freak.
True
There are so many things that go on in a basketball game that don’t show up in statistics and I don’t play that ohh this ones better because his stats are nice, stats are useful to a point and people shouldn’t start using then to discount other players ex. John Hollinger ESPN.com
When judging or trying to judge a player against others, subjectivity is always going to come into play. Statistics themselves are not subjective. But it’s a dangerous road to go own once you start weighing statistics as absolute truth. The mistake that I feel dog many fans is the fact that when making comparisons between players or among certain teams, they tend to use numbers as proof of their claims. And while numbers do tell us a lot about a player or a team, even the most sophisticated formula cannot entirely reflect everything that player does on the court.
John Hollinger obviously puts a lot of effort in his work, but even he had the Bucks as high as the fifth best team in the NBA at one point. Any reasonable fan would discount that immediately, but it’s sad to see that such a smart analyst would back up such a bold statement with just a single formula.
I consider basketball closer to art than hard science...
People can provide pseudo-proofs for almost anything with hyper-mathematization, but ultimately it is about the things you see, the matchups that are exploited, and the random events which make sports so interesting and ultimately unpredictable.
Pontiff of the Pryz for Prez Posse...
An outsider thought
I thought I would highlight ATS (at the hive) point about Byron Scott regarding something he mentioned regarding Brandon Bass:
Want to guess how I’d assess Byron Scott’s treatment of Brandon Bass? Here, I’m going to throw some numbers at you:
Off. Reb. % Def. Reb. % Tot. Reb. %
Player A 12.2 22.2 17.3
Player B 7.8 22.8 15.1
Player C 9.1 17.1 13.1
Player D 10.5 23.8 17.6
Now, that’s pretty subjective if you ask me. The idea of an opinion is just that. It’s subject to the person giving it. We all know that.
Numbers aren’t the end all be all of anything. They can be misleading or take a point that’s not necessarily worthwhile. This is all true.
Numbers can also highlight important points when the proper light is shed. That’s why I brought up this point about Bass. It helps illustrate an argument that Byron Scott did not spend enough time or considering why Brandon Bass could have helped the Hornets long term.
That’s just one of many examples. I’ve seen bad examples where PER is used (like in some of the ways that John Hollinger does it), but that’s also because I’ve never been a big fan of PER really. Numbers can tell you anything you want them to, and there is no question this. I can’t wait for the day people embrace these numbers, and look at them as an alternative rather than replacement.
Evil Cowtown Inc: Screwin' Suckaz over since Nineteen Eighty-Five.....
No mistakes in the tango, darling. Not like life. Simple. That's what makes the tango so great. If you make a mistake, and get all tangled up, you just tango on.....
there is still a difference
between the “cocktail” stats, like PER of Wages of Wins, and rebound percentage. The isolated stats can prove to be pretty useful.
draft dejuan blair
The problem with the isolated stats
Are that they aren’t “isolated” or independent. A basket scored is sometimes the result of the efforts of all ten players on the court. Even a blocked shot is the dependent on other people beside the blocker and the blocked. So you have to keep that in mind even when using the most basic of stats. If Jonny and Mikey played for the same team, and Jonny scored 12 points a game, and Mikey scored 15 a game, that would not automatically make the latter the better scorer. You would actually have to take a look at their skill set, and what roles they played on the team. It would still come down to a subjective evaluation.
Not to mention when other guys block out the opposition in order for someone on their team to get a rebound…
Even then, it's subjective
I’m quite certain this is not the only issue at play, but just one way in which this is still not an objective measure of a player is expectation.
For example, try comparing a power forward to a shooting guard. On offense, it is often the case that the power forward’s role is to go after an offensive rebound if possible, while it is the shooting guard’s role is to get back down the court and set up defensively. Why? Because if the shooting guard also goes in for the offensive rebound, it is true that the team is more likely to get a second chance opportunity — but it is even more likely that the opposing team will secure the rebound, make the outlet pass, and have an opportunity at a fast break or a 3-on2 or 2-on-1… all because the shooting guard didn’t get back defensively.
On defense, it is again the role of the power forward to rebound, preventing any second chance opportunities for the other team. Meanwhile, depending on the coach’s offensive system, it may be the role of the shooting guard to leak out to receive the outlet pass, possibly setting up a fast break or a 3-on-2 or 2-on-1.
In late game situations where one team is down, this can all change. Recognizing the need to take an approach that is riskier, but has higher potential for rewards, a coach may decide to go harder after offensive rebounds (leaving the door open for the other team to rebound and get a quick score), and gamble more to set up fast break opportunities (meaning rebounding with fewer guys on the defensive glass, which could set up an easy offensive opportunity on one side if they succeed, but risks an offensive rebound and putback for the other team if they don’t).
In these scenarios, the positional expectations on the player, based on the coaching style and the offensive and defensive systems the team runs, directly affects what kind of role each player plays in rebounding. For a team that really wants to get out and run a lot, the coach may ask the shooting guard not to worry about defensive rebounds at all, except perhaps on long rebounds. Game situations also may affect the role that each player plays in the team’s overall rebounding effort.
So you see, the needs of the team, the expectations of the coach, and the roles that different players play in their respective offensive and defensive systems are all significant factors that affect each player’s overall rebounding rate.
It’s much more applicable to use these stats to evaluate an entire team as a unit, rather than individual players within the team. There are often still multiple factors at play, but if you say that one team is a better rebounding team than another, based on rebounding rate, you’re much more likely to be correct than if you make the same statement about two players, using the same statistic.
16...15...14...13...12...
Strength & Honor
In Houston, our GM Darryl Morey, is a "stat guy"
but from many of the thing’s he’s said, I think very likely agree with you. He’s generally very vague in talking about some of the things he and his crew do, but they create their own stats by trying to look at and evaluate everything that goes on in any specific play; did a player set a good pick, did someone tip a rebound out to another player, or did the team recover a lose ball after a blocked shot? Morey would also agree (and he has said so) with your disdain for the idea that any player’s absolute value can be simplified into a single metric.
One big difference between Morey and writers like Hollinger is that Morey is paid to be correct, whereas people like Hollinger are paid to please the people who read their columns. Simplifying value to a single number panders to people who want to be able to easily compare different players (ie. most people).
Time to galvanize
Yes I'm not devaluing statistics as a whole
Just that creating a single individual “worth” metric that trumps context, eras, and any other variables is implausible, and that even the most basic individual stats require subjective interpretation to determine the quality of the performance of a player.
by Gils_Keloids on May 5, 2009 10:31 AM PDT up reply actions
I read an article before
I think it was the Bill Simmons’ article. Morey basically said that he does some statistical analysis for certain situations, but doesn’t really rely on them. It’s more for “convenience”(For the lack of a better term) in certain situations where it could prove to be useful. Such as assessing a player for that could feel a certain need that needs to be filled. He also asserted that the most important/telling stats are the 4 factors (EFG, FTr, Oreb%, TOr).
Personally, I don’t like it when people limit themselves to using advanced stats to determine whether Player A or Player B without actually “making meaning” out of the numbers they pull. I think it was JE Skeets from BDL who said it best: (paraphrasing, not exact) As long as these stat analysts understand how it feels like to pick up a basketball then we’ll be fine.
these fanposts are starting to kill me
by misterterrific on May 9, 2009 10:11 AM PDT up reply actions
agh, what the hell
assessing a player for that could *fill a certain need that needs to be filled
these fanposts are starting to kill me
by misterterrific on May 9, 2009 10:13 AM PDT up reply actions
Agreed
Limited usage, where the context is specific, can be helpful.
"This is not a game for boys. This is a game for men." - Phil Jackson
by Gils_Keloids on May 9, 2009 12:32 PM PDT up reply actions
Two thoughts on the post
1. Even if advanced metrics aren’t infalliable, they’re still a hell of a lot better than “traditional” stats. A couple of examples. One is TS% and eFG% as opposed to regular FG%. Another would be Points scored and points allowed by teams vs. efficiency adjusted through pace.
In a game filled with stats, you have to use SOMETHING to measure a player or a team’s worth. I’d rather use some of the new stats than just boxscore stats which are usually much more misleading.
2. Just because you don’t understand something doesn’t mean it’s bad. I mean, even if I don’t like PER because I think it tries too hard to sum up a player’s worth. That doesn’t mean there does not exist such a number which COULD work. You can’t just dismiss all advanced statistics just because you don’t like some of them.
Oh, I understand the statistics
I have a degree in mathematics, which is not a boast, but I am saying that to support my claims that because I understand the statistics, as well as the nature of basketball, I see just how subjective the individual metrics are.
And I am not advocating individual box score metrics by any means. I’m talking about parsing out the stats to individual players, which is an art, not a science (yes, even the box score stats you see, and even TS% and eFG%).
About the only independent stat I can think of is FT%, where I can say with a great degree of certainty that if player X shoots a higher percentage than player Y, then player X is a better free throw shooter than player Y. To apply that type of comparison to any other individual statistic without any contextual and subjective analysis would be folly.
Just reread your point 1
And you missed that I actually like the team stats, especially the new ones.
Because with the team you are looking at the unit at which the organism is operating. In my example, it’s a car. The purpose of the car is to transport you in some manner. So you look at how the car performs as a whole, not the individual parts. In the same way, a basketball team works as a whole, so looking at team statistics can be very valuable. Every player plays a different role, just as the parts of a car do. No one is “better” or “more valuable”, but some people’s interpretation of “valuable”, is equivalent to the way that the engine is “more valuable”, but the car still wouldn’t go anywhere without the wheels! That’s fine to do (say one part is more valuable than the other), but “valuable” is a subjective evaluation, even if you decide to attach a number or rating to it. (Stretching the anology further, is the part that costs the most also the most valuable to the car? Think about that.)
Put it this way:
Two players could have the exact same statistical line in a game (points, rebounds, assists, even plus/minus) and still have had a completely different effect on the outcome of the game, in magnitude and direction.
What does that tell you about the reliability of comparing two individual’s statistics to evaluate performance?
"This is not a game for boys. This is a game for men." - Phil Jackson
Comments For This Post Are Closed
Secondary Sidebar
FanPosts
Community blog posts and discussion.
Recommended FanPosts
Recent FanPosts
Featured Stories
FanShots
Quick hits of video, photos, quotes, chats, links and lists that you find around the web.
Recommended FanShots
Recent FanShots
Certain photos copyright © 2010 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
