TS% vs FG%

Moderator: Doctor MJ

mysticbb
Banned User
Posts: 8,205
And1: 713
Joined: May 28, 2007
Contact:
   

Re: TS% vs FG% 

Post#61 » by mysticbb » Thu Nov 22, 2012 12:20 pm

Hendrix wrote:How do you know that we can derive this information?


By determining a pattern for certain player types.

Hendrix wrote:How do you know how accurate the information is?


Direct testing.

Hendrix wrote:You didn't bother to comment on any of my comments about the strength of the correlation or anything


Because it made no sense whatsoever to the points made. It seems like you have the wrong impression that someone argued that with just one single number (in that case FG%), some sort of other quantifiable value would be generated, which then can be compared to a direct measurement of the wanted value. If that would be the case, such regression analysis would make sense. But that is not the case. The information we can derive by the compilation of shooting numbers can be used to put players into certain categories.

Hendrix wrote:For example: player 'x' shoots 45fg% in 1993 in the ACB.


Nobody ever argued that this would be enough information about a specific player to say more about him than how many of his field goal attempts he converted. Because that's what FG% is telling you. And it is very precise at that too. ;)
tsherkin
Retired Mod
Retired Mod
Posts: 78,346
And1: 19,900
Joined: Oct 14, 2003
 

Re: TS% vs FG% 

Post#62 » by tsherkin » Thu Nov 22, 2012 4:44 pm

FG% says something very specific. In what it tells you, it's not flawed, but to use it in any meaningful capacity for player comparison is pointless.

It is useless without also knowing volume and doesn't paint a good picture of total efficiency (either scoring efficiency, like TS%, or offensive efficiency, like ORTG).

As I said, it's useful when broken down by zone and compared to positional league average, which is readily available. It's useful compared to that player's previous performance. In a direct P2P comparison, though, it offers too little info to be of value. And inferences made about style of play through raw FG% don't tell much. You could be a slasher who takes lots of floaters and pull-ups, relying on FTA for efficiency, or a pure mid-range shooter, or a guy who bombs a lot of 3s and you would have an equally low FG%, rendering the inference uselessly imprecise and variable.

Zone-specific FG%, TS%, eFG%, ORTG, game film... As ever, a comprehensive approach is better than relying on a single number. FG% isn't terribly useful, and is more flawed than most raw box score stats. It would be simpler to look at your typical NBA/ESPN box score and be able to see useful data, but we don't. That's why those other numbers exist. As long as we evade Tendex-style BS metrics, there are plenty of numbers out there that (when combined with film) do a great job of describing things.

I don't see how this debate has gone on any length. Looking at raw FG% has limited value, there are too many things it doesn't say, questions unanswered.
mysticbb
Banned User
Posts: 8,205
And1: 713
Joined: May 28, 2007
Contact:
   

Re: TS% vs FG% 

Post#63 » by mysticbb » Thu Nov 22, 2012 5:13 pm

tsherkin wrote:FG% says something very specific. In what it tells you, it's not flawed, but to use it in any meaningful capacity for player comparison is pointless.


That is true for every boxscore entry.

tsherkin wrote:And inferences made about style of play through raw FG% don't tell much. You could be a slasher who takes lots of floaters and pull-ups, relying on FTA for efficiency, or a pure mid-range shooter, or a guy who bombs a lot of 3s and you would have an equally low FG%, rendering the inference uselessly imprecise and variable.


FG% as a single number alone can't tell you anything about that, which is understandable because it has a very precise meaning. But in combination with the other numbers, it is helpful. Especially when you have 3p%, FT% , eFG% and TS% also. Either one of those alone will also not tell you anything, but the combination of those numbers will rather limit the possibilities how a player acts on offense. For example:

Player A: 50 FG%, 45 3p%, 90 FT%, 55 eFG%, 56 TS%
Player B: 46 FG%, 40 3P%, 70 FT%, 58 eFG%, 60 TS%
Player C: 56 FG%, 0 3P%, 58 FT%, 56 eFG%, 57 TS%
Player D: 43 FG%, 43 3P%, 90 FT%, 53 eFG%, 60 TS%

I honestly think, that you have at least an idea what kind of players can produce such numbers.
User avatar
EvanZ
RealGM
Posts: 12,640
And1: 3,181
Joined: Apr 06, 2011

Re: TS% vs FG% 

Post#64 » by EvanZ » Thu Nov 22, 2012 5:33 pm

So, I took a quick look to see the correlation between FG%, eFG% and ORTG and DRTG at the team level for 2012 (last season).

Predictably, eFG% has a higher correlation with ORTG (R^2 of 0.66 vs. 0.55). Note that FG% is obviously highly correlated as well, and it's not that much different. Not as different as you might be led to believe after reading this thread.

Anyone care to guess which stat (eFG% or FG%) correlates better with DRTG?
I was right about 3 point shooting. I expect to be right about Tacko Fall. Some coach will figure out how to use Tacko Fall. This movement towards undersized centers will sweep ng back. Back to the basket scorers will return to the NBA.
tsherkin
Retired Mod
Retired Mod
Posts: 78,346
And1: 19,900
Joined: Oct 14, 2003
 

Re: TS% vs FG% 

Post#65 » by tsherkin » Thu Nov 22, 2012 5:38 pm

mysticbb wrote:That is true for every boxscore entry.


That was part of my point, yes.

FG% as a single number alone can't tell you anything about that, which is understandable because it has a very precise meaning. But in combination with the other numbers, it is helpful.


It can be, but you're now arguing something I wasn't discussing.

Player A: 50 FG%, 45 3p%, 90 FT%, 55 eFG%, 56 TS%
Player B: 46 FG%, 40 3P%, 70 FT%, 58 eFG%, 60 TS%
Player C: 56 FG%, 0 3P%, 58 FT%, 56 eFG%, 57 TS%
Player D: 43 FG%, 43 3P%, 90 FT%, 53 eFG%, 60 TS%

I honestly think, that you have at least an idea what kind of players can produce such numbers.


See but now you're including TS% and eFG%, as well as FT% and 3P%, rendering the inclusion of FG% moot, and my response is "so what?"

Once you start including all the data requisite in making FG% somewhat useful, it's invalidated by overlap. Again, zone-specific stuff, all day. I love it, those numbers are highly illustrative.

But just overall FG%? Not so useful, no.
User avatar
EvanZ
RealGM
Posts: 12,640
And1: 3,181
Joined: Apr 06, 2011

Re: TS% vs FG% 

Post#66 » by EvanZ » Thu Nov 22, 2012 5:54 pm

Another useful thing about FG% is that its variance is lower than eFG%. If you take a 2-pt shooter and a 3-pt shooter both having the same eFG%, the variance from game to game will be higher for the 3-pt shooter. Dean Oliver wrote about how if you're an underdog, you might use that variance to your advantage, but if you're the favorite it works against you.

Making an analogy to finance, if you have two stocks with the same return, but one has lower variance than the other, which do you invest in?
I was right about 3 point shooting. I expect to be right about Tacko Fall. Some coach will figure out how to use Tacko Fall. This movement towards undersized centers will sweep ng back. Back to the basket scorers will return to the NBA.
mysticbb
Banned User
Posts: 8,205
And1: 713
Joined: May 28, 2007
Contact:
   

Re: TS% vs FG% 

Post#67 » by mysticbb » Thu Nov 22, 2012 5:55 pm

EvanZ wrote:Anyone care to guess which stat (eFG% or FG%) correlates better with DRTG?


FG%, but me saying that is some kind of cheating, I guess.

But it should be rather clear, why. A possession after a defensive rebound is more efficient than after an inbound. And given the fact that 3pa and 2pa are similar likely to end up with a defensive rebound (yeah, that 3pa are more likely to end up as offensive rebound is a myth), the added point for the 3pt made will make the correlation lower.
User avatar
EvanZ
RealGM
Posts: 12,640
And1: 3,181
Joined: Apr 06, 2011

Re: TS% vs FG% 

Post#68 » by EvanZ » Thu Nov 22, 2012 5:58 pm

mysticbb wrote:
EvanZ wrote:Anyone care to guess which stat (eFG% or FG%) correlates better with DRTG?


FG%, but me saying that is some kind of cheating, I guess.

But it should be rather clear, why. A possession after a defensive rebound is more efficient than after an inbound. And given the fact that 3pa and 2pa are similar likely to end up with a defensive rebound (yeah, that 3pa are more likely to end up as offensive rebound is a myth), the added point for the 3pt made will make the correlation lower.


This is an important point, and the difference in R^2 (0.21 vs. 0.15) is not trivial.

I think the efficiency effect and the variance I mentioned above are two "useful" things to know about FG%, wouldn't you say? 8-)
I was right about 3 point shooting. I expect to be right about Tacko Fall. Some coach will figure out how to use Tacko Fall. This movement towards undersized centers will sweep ng back. Back to the basket scorers will return to the NBA.
mysticbb
Banned User
Posts: 8,205
And1: 713
Joined: May 28, 2007
Contact:
   

Re: TS% vs FG% 

Post#69 » by mysticbb » Thu Nov 22, 2012 6:11 pm

EvanZ wrote:I think the efficiency effect and the variance I mentioned above are two "useful" things to know about FG%, wouldn't you say? 8-)


There are. Favorites can actually adapt to the variance "problem" by putting the player on the bench. But well, if you disregard any possible value for FG%, you might miss such things.
mysticbb
Banned User
Posts: 8,205
And1: 713
Joined: May 28, 2007
Contact:
   

Re: TS% vs FG% 

Post#70 » by mysticbb » Thu Nov 22, 2012 6:17 pm

tsherkin wrote:Once you start including all the data requisite in making FG% somewhat useful, it's invalidated by overlap.


As we both agreed upon, other numbers are running into the same issue. And if you remove the FG% from my example, it becomes harder to differentiate between the players. Why? Because the differences between FG% and eFG% are giving an estimate of the distribution between 2pt shots and 3pt shots.

tsherkin wrote:But just overall FG%? Not so useful, no.


Well, I hope the example Evan brought up change your mind a bit here.
User avatar
EvanZ
RealGM
Posts: 12,640
And1: 3,181
Joined: Apr 06, 2011

Re: TS% vs FG% 

Post#71 » by EvanZ » Thu Nov 22, 2012 6:50 pm

Here's something else to chew on this Thanksgiving Day (in America anyway):

Between FG% and eFG% which has a higher correlation with PyWins? :wink:
I was right about 3 point shooting. I expect to be right about Tacko Fall. Some coach will figure out how to use Tacko Fall. This movement towards undersized centers will sweep ng back. Back to the basket scorers will return to the NBA.
tsherkin
Retired Mod
Retired Mod
Posts: 78,346
And1: 19,900
Joined: Oct 14, 2003
 

Re: TS% vs FG% 

Post#72 » by tsherkin » Thu Nov 22, 2012 6:55 pm

mysticbb wrote:As we both agreed upon, other numbers are running into the same issue. And if you remove the FG% from my example, it becomes harder to differentiate between the players. Why? Because the differences between FG% and eFG% are giving an estimate of the distribution between 2pt shots and 3pt shots.


Sure, but I never agreed that eFG% was a good substitute for FG%, I said using multiple data points when evaluating a player is more important and valuable than using FG%.

Well, I hope the example Evan brought up change your mind a bit here.


Not at all, it wasn't related to my point (unless I misread) because I was talking about player evaluation and he's talking about projection of team-related data.

I don't find FG% useful in player comparisons unless it's zone-specific. There are too many other points of data that are necessary. At a team level, what you're trying to see and learn and project, it's a little different.
mysticbb
Banned User
Posts: 8,205
And1: 713
Joined: May 28, 2007
Contact:
   

Re: TS% vs FG% 

Post#73 » by mysticbb » Thu Nov 22, 2012 7:48 pm

tsherkin wrote:Not at all, it wasn't related to my point (unless I misread) because I was talking about player evaluation and he's talking about projection of team-related data.


How a player actually fits into those "team-related data" is something which should be part of a good player evaluation. We need to know how useful a player is in different settings.
User avatar
EvanZ
RealGM
Posts: 12,640
And1: 3,181
Joined: Apr 06, 2011

Re: TS% vs FG% 

Post#74 » by EvanZ » Thu Nov 22, 2012 8:37 pm

tsherkin wrote:
mysticbb wrote:As we both agreed upon, other numbers are running into the same issue. And if you remove the FG% from my example, it becomes harder to differentiate between the players. Why? Because the differences between FG% and eFG% are giving an estimate of the distribution between 2pt shots and 3pt shots.


Sure, but I never agreed that eFG% was a good substitute for FG%, I said using multiple data points when evaluating a player is more important and valuable than using FG%.

Well, I hope the example Evan brought up change your mind a bit here.


Not at all, it wasn't related to my point (unless I misread) because I was talking about player evaluation and he's talking about projection of team-related data.

I don't find FG% useful in player comparisons unless it's zone-specific. There are too many other points of data that are necessary. At a team level, what you're trying to see and learn and project, it's a little different.


Correct me if I'm wrong, but aren't teams composed of players?
I was right about 3 point shooting. I expect to be right about Tacko Fall. Some coach will figure out how to use Tacko Fall. This movement towards undersized centers will sweep ng back. Back to the basket scorers will return to the NBA.
jman3134
Forum Mod
Forum Mod
Posts: 19,459
And1: 1,308
Joined: Apr 17, 2005
Location: Follow me on Twitter: JTMBasketball
Contact:
 

Re: TS% vs FG% 

Post#75 » by jman3134 » Wed Dec 12, 2012 8:43 pm

As someone who puts these statistics into practice for player comparison/scouting purposes, I want to chime in here. I have historically utilized eFG% as a metric to normalize frontcourt and backcourt players in terms of their shooting percentages from the field. (obviously factoring in the additional weight to the three pt shot) I have no access to Synergy Sports data telling me where certain players are best shooting from different areas of the floor, but I watch these guys often enough to be knowledgeable about their sweet spots. Frankly, for player comparison purposes, I think that those Synergy numbers are largely irrelevant if you are looking back and trying to evaluate one player vs. another ala the player comparison board. (they can be used, but value is determined apart from these statistics) I say this because you can use Synergy stats to break down which shooter is best 20 ft out, but you are still going to have to make a judgment call as to whether or not that was more valuable to his team than someone who shot well from around the basket. Maybe that player's team had a wealth of shooters and only one guy who could finish inside? Also, perhaps the other player in the comparison hypothetically played alongside a big man like Shaq that forced the defenses to collapse and left him with open shots. So, you need to understand circumstance in order to gauge value. And, those stats (FG% from each available range) are useful when comparing players on that same basis, but not in determining value.

However, for assessing where improvements can be made etc., those Synergy stats are useful, particularly when it comes to scouting talent and filling free agency needs. That's why they are so valuable to both NCAA and NBA teams.

And, while I will admit that my knowledge of basketball statistics is not to the level of some here, I do feel that something is flying in the face of common sense here. When someone is evaluating a player(s) for player comparison or scouting purposes, why would one look purely at a shot chart as opposed to FG%? No one performs analysis on each individual play. While it says nothing about a player in terms of his tendencies, doesn't FG% say something about overall shot selection when used in coordination with other stats/film? That has always been my take on it. So, I agree with mystic and evan here.
User avatar
EzeDoesIt
Junior
Posts: 461
And1: 10
Joined: Jun 01, 2012

Re: TS% vs FG% 

Post#76 » by EzeDoesIt » Fri Jan 11, 2013 3:16 am

If you want to know the percentage of field goal attempts that a player makes, you should go with FG%.

If you want to know anything about a player's efficiency, look to TS%.
User avatar
EvanZ
RealGM
Posts: 12,640
And1: 3,181
Joined: Apr 06, 2011

Re: TS% vs FG% 

Post#77 » by EvanZ » Fri Jan 11, 2013 6:34 am

EzeDoesIt wrote:If you want to know the percentage of field goal attempts that a player makes, you should go with FG%.

If you want to know anything about a player's efficiency, look to TS%.


Indeed.
I was right about 3 point shooting. I expect to be right about Tacko Fall. Some coach will figure out how to use Tacko Fall. This movement towards undersized centers will sweep ng back. Back to the basket scorers will return to the NBA.

Return to Statistical Analysis