The Lie That's Called Shooting Efficiency

[permaximum]

I sometimes wonder why do people give too much credit to shooting efficiency? Especially, it's like there's a magical number somewhere that after it gets exponentially better ( Bad below 45%, Average between 45-50% and great after 50%)

For example let's take two guys that each shoots 20 times a game. One has 45% FG, the other has 50% FG.

How can we know if one takes considerable amount of those available open shots that a team receives each game, while the other player takes the contested shots and leaving most of the available open shots to his teammates thus incresing their respective eFG%?

There's a signifcant chance that even if we consider the shot efficiency alone, a 45%-shooting guy may help his team much more than a 50%-shooting guy.

Going forward, let's say contested/open shot balance is equal for each of them... Does anyone honestly think all shots are valued the same throughout the game? One shot may help the team much more to win the game than an another shot. How can we know which player is better at these crucial shots?

Let's make this a perfect example for shooting efficiency fans. Both players take the same amount of contested, open shots and both players take part in games that for some reason each shot is valued equal. And let's assume these players are completely equal in skill at everything else. How many more wins will 50%-shooting player add compared to 45%-shooting player? Let me tell you. (Let's assume they don't take 3 point shots or they take very less that it's not significant enough to make the example simple.)

0.0134% chance to add more wins.
0.0243% chance to take it to OT.

To make it even simpler and better for shooting efficeny fans let's assume they both play full 82 games a year. In that best-case scenerio, 50%-shooting guy will add 1.1 more wins compared to 45%-shooting guy in 82 games. And this is if we assume their contested/open shot ratio is equal and all shots in those 82 games are valued equal to eachother.

Someone had to call the lie at some point. When you miss so many variables while you are are trying to make controlled observations, tests and comparisons it will be laughable at best especially if you compare such shooting efficiency margins below 10%. And even in that case, you forget the point that 5% increase in FG for "one player" is almost nothing.

It's really disappointing for serious analytic pros to do such mistakes knowingly or not because casual fans WHO like and depend on advanced stats think TS% and it's variations are the end of everything. And worse, they don't know how to value player defenses and they go by common bias for player defenses in the media.

[Dr Positivity]

For example let's take two guys that each shoots 20 times a game. One has 45% FG, the other has 50% FG.

How can we know if one takes considerable amount of those available open shots that a team receives each game, while the other player takes the contested shots and leaving most of the available open shots to his teammates thus incresing their respective eFG%?

There's a signifcant chance that even if we consider the shot efficiency alone, a 45%-shooting guy may help his team much more than a 50%-shooting guy.

Well in your example they both shoot the same amount. However I would argue the difference in style of play between these two shooters is llikely to affect their FGA volume, with the player taking contested shots being the higher volume FGA taker, so it ends up being the long standing debate between volume vs efficiency. If both players take 20 shots per game but one guy is taking hard shots and the other one is taking good ones without sacrificing value, there's probably something important going on in the skillset that makes the latter more valuable.

Going forward, let's say contested/open shot balance is equal for each of them... Does anyone honestly think all shots are valued the same throughout the game? One shot may help the team much more to win the game than an another shot. How can we know which player is better at these crucial shots?

From my perspective a possession in the 1st quarter is worth the same as in the 4th. But there are also ways to evaluate which players have the best crunch time stats.

Let's make this a perfect example for shooting efficiency fans. Both players take the same amount of contested, open shots and both players take part in games that for some reason each shot is valued equal. And let's assume these players are completely equal in skill at everything else. How many more wins will 50%-shooting player add compared to 45%-shooting player? Let me tell you. (Let's assume they don't take 3 point shots or they take very less that it's not significant enough to make the example simple.)

0.0134% chance to add more wins.
0.0243% chance to take it to OT.

To make it even simpler and better for shooting efficeny fans let's assume they both play full 82 games a year. In that best-case scenerio, 50%-shooting guy will add 1.1 more wins compared to 45%-shooting guy in 82 games. And this is if we assume their contested/open shot ratio is equal and all shots in those 82 games are valued equal to eachother.

Someone had to call the lie at some point. When you miss so many variables while you are are trying to make controlled observations, tests and comparisons it will be laughable at best especially if you compare such shooting efficiency margins below 10%. And even in that case, you forget the point that 5% increase in FG for "one player" is almost nothing.

It's really disappointing for serious analytic pros to do such mistakes knowingly or not because casual fans WHO like and depend on advanced stats think TS% and it's variations are the end of everything. And worse, they don't know how to value player defenses and they go by common bias for player defenses in the media.

I don't really follow what you did here. Here is my attempt. 20 shots on 50% shooting is 10 points, 20 shots on 45% shooting is 9 points. A point is worth about 2.7 wins over the course of a season.

Shooting efficiency is definitely not a perfect measure, and there are examples of higher volume, lower efficiency scorers who were probably more valuable than lower volume, higher efficiency players by taking hard shots and opening up easy one for teammates. Virtually all boxscore stats have a spotty record of predicting season to season because shooting efficiency is unpredictable year to year, most likely because of contextual factors making players take easier or harder shots. The unpredictability of shooting efficiency is one of the many reasons why people have found success with advanced +/- stats which are an attempt to sift through things like the context drivenness of shooting efficiency to understand which players are helping their team win in ways not captured in the boxscore

[permaximum]

Well in your example they both shoot the same amount. However I would argue the difference in style of play between these two shooters is llikely to affect their FGA volume, with the player taking contested shots being the higher volume FGA taker, so it ends up being the long standing debate between volume vs efficiency. If both players take 20 shots per game but one guy is taking hard shots and the other one is taking good ones without sacrificing value, there's probably something important going on in the skillset that makes the latter more valuable.

I don't agree. There are more than 20 contested and/or open shots in the game and somebody has to take contested shots. You can't have a perfect game with lots of open shots. It's very hard to create an open shot by one individual player's skillset alone and that means the latter simply takes lots of open shots created by the team as a whole. The one WHO takes the contested shots is a lot more valuable for the team given the fact that someone else can replace the other player WHO takes open shots.. Shooting volume is not the issue here. A player can find 15 open shots while an another can find 15 contested and eventually reaching their respective FGA of 20.

From my perspective a possession in the 1st quarter is worth the same as in the 4th. But there are also ways to evaluate which players have the best crunch time stats.

Even two different shots in the same minute can worth differently depending on the score, margin, momentum, before-or-after break and the shot itself (power dunk, miracle shot, buzzer beater etc.) I thought this would be obvious for any basketball lover let alone for someone like you.

I don't really follow what you did here. Here is my attempt. 20 shots on 50% shooting is 10 points, 20 shots on 45% shooting is 9 points. A point is worth about 2.7 wins over the course of a season.

It doesn't work like that. There had been 2460 games last season. 33 of them ended with a margin of +1 and 60 of them ended with a margin of +2. You can reach added 1.1 wins per 82 games with those numbers when we compare two players WHO both shoot 20 times a game with their FG% of 45 and 50. The theory and the practice is different and that's one of the reasons I posted this thread.

Shooting efficiency is definitely not a perfect measure, and there are examples of higher volume, lower efficiency scorers who were probably more valuable than lower volume, higher efficiency players by taking hard shots and opening up easy one for teammates. Virtually all boxscore stats have a spotty record of predicting season to season because shooting efficiency is unpredictable year to year, most likely because of contextual factors making players take easier or harder shots. The unpredictability of shooting efficiency is one of the many reasons why people have found success with advanced +/- stats which are an attempt to sift through things like the context drivenness of shooting efficiency to understand which players are helping their team win in ways not captured in the boxscore.

Completely agreed. However I stand by my claim that RAPM is too noisy, I mean "a lot". On the other side RPM is less noisy but very biased.

[Crow]

The new shooting data with degree of contest can help, if you have it.

Overall efg% is a starting point and, yes, should not be an end point. The two new tools by nyloncalculus authors that compare shooting results to "expected" results are helpful for assessing shot difficulty, contribution to team.



The degree of "noise" and "bias" in non RAPM / RPM impact assessment approaches is high too. Especially where one metrics are rejected and the evaluator just synthesizes all the discrete and varied data without any weights or formulas. Final answers just arise in the mind and either accepted or ruminates on and adjusted by "experience" and "judgment", usually without documentation. So while RAPM / RPM deserve critique / skepticism, so does everything else, if not more so.

[Nate]

....

It's really disappointing for serious analytic pros to do such mistakes knowingly or not because casual fans WHO like and depend on advanced stats think TS% and it's variations are the end of everything. And worse, they don't know how to value player defenses and they go by common bias for player defenses in the media.

Who are these "serious analytic pros", and how do you get access to their analytic product?

For me, it's always important to start with a question. It's nonsense to say "EFG is a red herring" without first establishing what the objective is. The same thing that tells us that single-number metrics are unlikely to always be appropriate, also tells us that single number metrics are unlikely to always be inappropriate.

In the mass media, statistics are trivia for entertainment. That's the reason that silly stuff like streak metrics are popular. We can't really expect the sort of clarity or nuance that would be necessary for sophisticated discussion when everything has to be packaged into 30 second stand alone sound bytes.

[KwameBeanJordan]

I sometimes wonder why do people give too much credit to shooting efficiency?

I agree to an extent. It's easy to see how comparing the shooting efficiency of two players can be misleading:

(1) If one player takes a higher number of "bail out" shots.
(2) If one player is moreso the recipient of quality playmaking by his teammates.
(3) If one player's misses produce much better rebounding opportunities.

I tend to look for (1), (2), and (3) very closely when watching film. I take notes on these instances. In what footage I have recorded, however, I haven't found any of (1), (2), or (3) to be very different between superstars.

[permaximum]

Who are these "serious analytic pros", and how do you get access to their analytic product?

For me, it's always important to start with a question. It's nonsense to say "EFG is a red herring" without first establishing what the objective is. The same thing that tells us that single-number metrics are unlikely to always be appropriate, also tells us that single number metrics are unlikely to always be inappropriate.

In the mass media, statistics are trivia for entertainment. That's the reason that silly stuff like streak metrics are popular. We can't really expect the sort of clarity or nuance that would be necessary for sophisticated discussion when everything has to be packaged into 30 second stand alone sound bytes.

1. You. (Plural)

2. Here, there.

3. The objective is the word itself. FG% or EFG% or TS% doesn't really capture the meaning of "shooting efficiency". When we think of shooting efficiency to compare individual players we tend to think of them which lead us to potentially wrong answers.

4. How is that different from two words that consist of 5 letters "better and worse"? Those metrics claim they give the final answer of "better or worse" without getting deep into "how".

[Mike G]

You may be confusing the bathwater with the baby here. A metric doesn't "claim" anything. Will a Ford get a more beautiful woman in your car than a Toyota would? An ad may suggest it to you, but the car itself is neutral on the matter.

A shooting% is generally a partial tool when you want to evaluate a player. If you want a guy who can finish with a high% but doesn't need to create shots for others, you combine his TS% with his Usg% and TO%, and whatever else you find interesting. If you need a shot creator, you may include Ast% and whatever.

I am sympathetic with the idea that by itself, it's not telling us much. Yet I can't remember when a player's shooting% was the only available stat for him.

[permaximum]

You may be confusing the bathwater with the baby here. A metric doesn't "claim" anything. Will a Ford get a more beautiful woman in your car than a Toyota would? An ad may suggest it to you, but the car itself is neutral on the matter.

A shooting% is generally a partial tool when you want to evaluate a player. If you want a guy who can finish with a high% but doesn't need to create shots for others, you combine his TS% with his Usg% and TO%, and whatever else you find interesting. If you need a shot creator, you may include Ast% and whatever.

I am sympathetic with the idea that by itself, it's not telling us much. Yet I can't remember when a player's shooting% was the only available stat for him.

Pardon me, but what's the purpose of those metrics then? Perhaps I used the word "claim" without knowing it's true meaning and that's why you answered like that. Otherwise, what you said sounds absurd to me. Without any claim why does that metric exist anyways? There has to be a thesis for a metric to exist.

OFC there are lots of other available stats for players. But we are talking about shooting efficiency in particular here. All around the web you will spot people who believe in stats and think they are everything but all they do is just using TS% to compare two players that have TS differential like 3-4%. So I don't get why you pointed out other stats too...

[Nate]

... For me, it's always important to start with a question. It's nonsense to say "EFG is a red herring" without first establishing what the objective is. ...

3. The objective is the word itself. FG% or EFG% or TS% doesn't really capture the meaning of "shooting efficiency". When we think of shooting efficiency to compare individual players we tend to think of them which lead us to potentially wrong answers.

If you're complaining that "True Shooting Percentage" is a pretentious name, then I'll freely agree with you, but should like to point out that sports statistics are rife with other, similarly pretentious terms. (What's real about "Real Plus Minus"? What kind of efficiency does the "Player Efficiency Rating" represent?) None of the statistical metrics we have for players do a particularly good job of controlling for context. It's challenging to do that, and the public just doesn't have access to enough data for applying brute force methods. (I imagine that teams don't either, but insight like that is limited for a dilletante like me.)

Maybe I'm misunderstanding. Can you provide an example of a popular statistical metric that doesn't have the issues that the `shooting efficiency` stats have, and elaborate a bit on how to distinguish a "lie" stat from a "not lie" stat?

[permaximum]

If you're complaining that "True Shooting Percentage" is a pretentious name, then I'll freely agree with you, but should like to point out that sports statistics are rife with other, similarly pretentious terms. (What's real about "Real Plus Minus"? What kind of efficiency does the "Player Efficiency Rating" represent?) None of the statistical metrics we have for players do a particularly good job of controlling for context. It's challenging to do that, and the public just doesn't have access to enough data for applying brute force methods. (I imagine that teams don't either, but insight like that is limited for a dilletante like me.)

Maybe I'm misunderstanding. Can you provide an example of a popular statistical metric that doesn't have the issues that the `shooting efficiency` stats have, and elaborate a bit on how to distinguish a "lie" stat from a "not lie" stat?

I'm just saying FG%, eFG%, TS% don't necessarily reflect the actual shooting efficiency of an individual player because of the reasons I mentioned in the OP. And even if it did like some assume (some may not belong here), 5% increase in FG% for a high usage player WHO takes 20 shots a game results in only 1.1 added wins given the scenerio that he plays all 82 games.

APM and it's variants. But like I said before, it has the problem of noise in it's pure form or bias in it's varied forms.

[Nate]

...I'm just saying FG%, eFG%, TS% don't necessarily reflect the actual shooting efficiency of an individual player because of the reasons I mentioned...

Suppose you had infinite resources and infinite time. Can you propose an experiment that produces an "actual shooting effiency" measurement that doesn't also suffer from the same sorts of issues that you're complaining about for FG%. (Ideally, this would also be a measurement that other people agree shows the "actual shooting efficiency" of the player.)

...
APM and it's variants. But like I said before, it has the problem of noise in it's pure form or bias in it's varied forms

But what makes APM "honest" and EFG% a "lie"?

[permaximum]

Suppose you had infinite resources and infinite time. Can you propose an experiment that produces an "actual shooting effiency" measurement that doesn't also suffer from the same sorts of issues that you're complaining about for FG%. (Ideally, this would also be a measurement that other people agree shows the "actual shooting efficiency" of the player.)

But what makes APM "honest" and EFG% a "lie"?

I can. But in reality, it's impossible. Each single NBA game has to be watched in order to capture the contest degree and ratio of every shot and also each of those shots should be valued differently by analysing the situation it was taken in.

Let me get this straight. EFG% is not a lie but using it or it's variants to compare players' shooting efficiency is the lie when you deal with such differentials. If one player has 50% EFG and the other has 43% it doesn't automatically mean the former is a more efficient shooter. Perhaps he is perhaps he is not. And even if he is a better shooter, with such low margins, it doesn't mean he helps his team considerably more than the latter. It's not like we're comparing 20% of eFG% differentials between NBA players. With such margins, it's hard to know who's more efficient without analysing the shots taken.

APM captures everything. That's why it's honest but I listed it's own cons.

[Nate]

APM captures everything. That's why it's honest but I listed it's own cons.

In the original comment, there's a critique is that eFG doesn't control for things like "difficulty of shot", but APM doesn't control for all the factors for "difficulty of pointn" either. (Sure, the opposing players matter, but it also matters if the team is at the end of a long road trip or on a back-to-back, rubber banding means that teams who are behind tend to score net more, and player performance is typically weaker at the start of a stint.)

APM certainly assumes that all points are created equal. The original comment criticizes eFG for considering every shot to be equal, but the passage seems just as apropos with a little search and replace:

Does anyone honestly think all points are valued the same throughout the game? One point may help the team much more to win the game than an another point. How can we know which player is better at these crucial points?

Like most NBA stats, APM is something that measures performance, rather than potential. We're basically stuck trying to find the best we can with the limited data that we have.

[permaximum]

In the original comment, there's a critique is that eFG doesn't control for things like "difficulty of shot", but APM doesn't control for all the factors for "difficulty of pointn" either. (Sure, the opposing players matter, but it also matters if the team is at the end of a long road trip or on a back-to-back, rubber banding means that teams who are behind tend to score net more, and player performance is typically weaker at the start of a stint.)

APM certainly assumes that all points are created equal. The original comment criticizes eFG for considering every shot to be equal, but the passage seems just as apropos with a little search and replace:

Like most NBA stats, APM is something that measures performance, rather than potential. We're basically stuck trying to find the best we can with the limited data that we have.

APM doesn't have to. The result eventually includes everything that's happenning on the court. Including the sacrifice of a player who takes difficult shots and increases teammates' efficiency by giving them open shots, generally. APM will reward and punish players accordingly. That's the basis of APM.

As for the things you mentioned such as back-to-back, score diffential during the game; there are APM variants which adjust for those type of situations. J.E.'s xRAPM and his current RPM adjust for those type of things. In return, it introduces bias with each adjustment like I pointed out before.

I think you got me wrong. I'm not saying APM variants are the answer for our questions. On the contrary, I'm saying they are not enough at the moment. Besides I'm not sure it's logical to compare a captured statistic such as eFG with complex metrics like RPM. What I wanted to achieve with this thread was to come to the conclusion that;

1. FG% and it's variants don't really draw the whole picture when it comes to shooting effiency and players' help towards wins.
2. Small FG% differentials like 5% between players (even high-usage ones) are almost irrelevant in the real wold scenerios.
3. What I forgot was there's also a difference between a guy who shoots 50% from the floor every night and an another guy who shoots 50% from the floor on average but in a very inconsistent manner. The latter has potential to add more wins or loses. Some great players shoot more efficienty in close games and can go really off when it doesn't even matter if he shoots 70% from the floor.

But I'm not suggesting an alternative or blaming anyone with ignorance. I'm just stating these just in case some people may forget and try to simulate real games with TS% and eFG%. Also, it would help if some stat-lover NBA fans understand that TS% is nothing as a sole stat and stop comparing players with it. I thought you people would be some help. If you forget situations like this, how can I blame a typical realgm stat "guru" (btw I'm not a member in any basketball related forum or site but this one).