the case against TS%

[INKtastic]

People like to use TS% to determine the most efficient shooters. I generally prefer advanced stats, but TS% is a flawed stat. The problem is to attempt to adjust for all of the ways people get foul shots, every foul shot count as 0.44 FGA.

Example 1:

Player 1 takes a 3 and makes it - the TS% for that possession is 150%
Player 2 drives, gets fouled, makes their shot, makes their foul shot - the TS% for that possession is 104%

They should be identical. They both used one possession and scored 3 points. Instead player 1 gets a better TS%.

Example 2:

Player 1 drives, gets fouled, makes their shot, misses their foul shot - TS% 69.4
Player 2 drives, gets fouled, misses their shot, makes their two foul shots - TS% 113.6

They both had 1 possession, they both scored 2 points, they should both have identical TS% of 100%. Instead the guy who missed the shot when fouled and converted both shots in the and-1 gets a better TS% than even player 2 in the first example, who converted a 3 point play, which makes absolutely no sense at all, the better player who scored more points gets punished.

Example 3:
Player 1 shoots a 3 and makes it - TS% 150%
player 2 gets fouled attempting a 3, misses the shot, makes all 3 foul shots. - TS% TS% 113.6

They both scored 3 points off of 1 possession, they should have the same TS%.

Example 4:
Player 1 shoots a 3 and makes it - TS% 150%
player 2 gets fouled attempting a 3, makes shot, makes the foul shot. - TS% 139

The player who makes the 4 point play gets a worse TS% for that possession than the player who made the 3. His TS% should be 200% for that.

the 0.44 is supposed to be a league average to make sure all possessions are accounted for. The idea was to make sure possessions that didn't count as FGA end up being counted as FGA. Either fouled while shooting and miss the shot or fouled in the penalty. Instead it doesn't account for changes in the game (the percentage of 3 point attempts is increasing) or the strengths and weakness of different players.

There should be enough data these days to calculate a real TS%. (and-1s shouldn't count as a FGA, 3 foul shots fouled while shooting a 3 counts a 1FGA, all other foul shots count as 0.5 FGA)

If that was done, I would prefer that stat over eFG% for measuring efficiency.

[Slava]

There's already a points per shot, if that is what you want to use.

[rpa]

There's already a points per shot, if that is what you want to use.

Isn't points per shot merely points per FG attempt? If so that's an equally flawed metric.

[Slava]

There's already a points per shot, if that is what you want to use.

Isn't points per shot merely points per FG attempt? If so that's an equally flawed metric.

The problem the OP suggests is that TS% is being used to measure scoring efficiency, which is not its intended use but its the best measure of shooting efficiency. PPS is a better measure of scoring efficiency. They should each be used for different purposes.

[MarcusBrody]

There's already a points per shot, if that is what you want to use.

Points per shot is generally calculated using FGA, correct? [Edit: other replies while I was typing thing makes it look like it is].

If so, it would make people who score a lot of their points on foul shots look better than a player who used the same number of possessions to score the same amount of points without getting fouled. And in fact, it would again penalize players for making a shot after being fouled as the made shot becomes a FGA even though the scored more points on the same number of possessions.

[batmana]

I absolutely support the OP's points and couldn't have given better examples even if I tried. My basic problem with TS% has always been that it puts FGs and FTs together when they are completely different and ultimately TS% doesn't tell you how likely player X is to make a shot or a FT since it is a stat that somehow tries to encompass both.
Also, I understand how you wouldn't want to change the formula because play by play stats aren't available for earlier eras but the .44 coefficient IMO makes TS% not a reflection of reality (as shown by the OP's examples) but an arbitrary estimation. I personally only look at EFG% and FT% separately, they tell me how likely a player is to make a shot and how likely he is to make a FT.

[Slava]

There's already a points per shot, if that is what you want to use.

Points per shot is generally calculated using FGA, correct? If so, it would make people who score a lot of their points on foul shots look better than a player who used the same number of possessions to score the same amount of points without getting fouled. And in fact, it would again penalize players for making a shot after being fouled as the made shot becomes a FGA even though the scored more points on the same number of possessions.

Yes ofcourse, drawing fouls is a crucial part of offensive efficiency and it should be measured as such.

The part about and1s being counted as an extra shot attempt while FTs aren't is a good one but I don't think the proportion is too high (I haven't checked, just an assumption) to create a big disparity.

[rpa]

The problem the OP suggests is that TS% is being used to measure scoring efficiency, which is not its intended use but its the best measure of shooting efficiency. PPS is a better measure of scoring efficiency. They should each be used for different purposes.

I think the OP's argument is that due to the way it weights FTs (as an exercise in laziness at this point) makes it far inferior as a shooting efficiency metric than it would otherwise be with correct data.

[INKtastic]

There's already a points per shot, if that is what you want to use.

Isn't points per shot merely points per FG attempt? If so that's an equally flawed metric.

The problem the OP suggests is that TS% is being used to measure scoring efficiency, which is not its intended use but its the best measure of shooting efficiency. PPS is a better measure of scoring efficiency. They should each be used for different purposes.

no, it has flaws too. a real TS% would do toss out the 0.44 factor for foul shots and do the following

and-1s don't count as a FGA
fouled while shooting a 3 counts a 1/3 FGA
all other foul shots count as 0.5 FGA

The data is in the game logs. And I imaging the latest tracking systems have the data (like is used for stats.nba.com), but am not sure if the required details are made available or not.

[og15]

Just use multiple things, include eFG% which takes FT's out of the equation

[Chinook]

Finding extreme examples to break a model isn't hard. The idea behind the stat is that on the aggregate, it will bear out values that are informative while also conforming to our qualitative observations. Demonstrating that TS% is unable to do this requires significantly more work, like a dissertation's level of math.

Practically, TS% is supposed to be an estimate, not an actual calculation. That's why it has a .44 coefficient for FTA rather than actually parsing out when the attempts should count as possessions. That assumption allows for the stat to be generated from a box score and not take a bunch of extra work. That way, we can hop on over to BBR and get the numbers the next day rather than having to wait for periodic releases by some stat wizard that may cost more money.

[kabstah]

The differences between PbP calculated TS% and estimated TS% with the 0.44 coefficient are very, very small.

https://elgee35.wordpress.com/2011/03/04/errors-in-true-shooting/

[Rapcity_11]

The OP is right, but with a large sample the issue basically goes away. It's been shown the difference between "real" TS% and the TS% with the 0.44 estimate is insignificant.

[TMU]

Great post though it has been discussed numerously in the past. It's a flawed measure and it's quite unfortunate when people nitpick a very small difference to argue that Player A is a better scorer than Player B.

[The Real Dalic]

My problem is that I feel like people only judge players based off of stats in general without even watching the players anymore. I feel like their are just too many stats that make some players look better/worse than they actually are.

[HoopsterJones]

Personally I think TS% is a flawed stat.

With the game on the line when your team needs a bucket via a jumpshot, is it better to have a 38% FG shooter to take the shot with a higher TS% or a 50% FG shooter with a lower TS%?

A player can shoot 38% from the field, 30% from 3, and 80% from FT and have a TS% around 52%. Carmelo Anthony has a TS around 53%, but is vastly superior to the other player. The advanced statistic incorrectly demonstrates that the difference between them is minuscule.

Also lower FG% equates to more missed shot attempts, which turns into defensive rebounds, transition defense, and a proabability of fast break points. But TS% masks lower FG% shooters.

[Amare_1_Knicks]

People like to use TS% to determine the most efficient shooters. I generally prefer advanced stats, but TS% is a flawed stat. The problem is to attempt to adjust for all of the ways people get foul shots, every foul shot count as 0.44 FGA.

Example 1:

Player 1 takes a 3 and makes it - the TS% for that possession is 150%
Player 2 drives, gets fouled, makes their shot, makes their foul shot - the TS% for that possession is 104%

They should be identical. They both used one possession and scored 3 points. Instead player 1 gets a better TS%.

Example 2:

Player 1 drives, gets fouled, makes their shot, misses their foul shot - TS% 69.4
Player 2 drives, gets fouled, misses their shot, makes their two foul shots - TS% 113.6

They both had 1 possession, they both scored 2 points, they should both have identical TS% of 100%. Instead the guy who missed the shot when fouled and converted both shots in the and-1 gets a better TS% than even player 2 in the first example, who converted a 3 point play, which makes absolutely no sense at all, the better player who scored more points gets punished.

Example 3:
Player 1 shoots a 3 and makes it - TS% 150%
player 2 gets fouled attempting a 3, misses the shot, makes all 3 foul shots. - TS% TS% 113.6

They both scored 3 points off of 1 possession, they should have the same TS%.

Example 4:
Player 1 shoots a 3 and makes it - TS% 150%
player 2 gets fouled attempting a 3, makes shot, makes the foul shot. - TS% 139

The player who makes the 4 point play gets a worse TS% for that possession than the player who made the 3. His TS% should be 200% for that.

the 0.44 is supposed to be a league average to make sure all possessions are accounted for. The idea was to make sure possessions that didn't count as FGA end up being counted as FGA. Either fouled while shooting and miss the shot or fouled in the penalty. Instead it doesn't account for changes in the game (the percentage of 3 point attempts is increasing) or the strengths and weakness of different players.

There should be enough data these days to calculate a real TS%. (and-1s shouldn't count as a FGA, 3 foul shots fouled while shooting a 3 counts a 1FGA, all other foul shots count as 0.5 FGA)

If that was done, I would prefer that stat over eFG% for measuring efficiency.

Tremendous post and information. I think taking multiple forms of shooting efficiency into account along with TS% may be the best way to combat its "flaw(s)".

[yosemiteben]

Good stuff in this thread, good job all.

[Rapcity_11]

Personally I think TS% is a flawed stat.

With the game on the line when your team needs a bucket via a jumpshot, is it better to have a 38% FG shooter to take the shot with a higher TS% or a 50% FG shooter with a lower TS%?

A player can shoot 38% from the field, 30% from 3, and 80% from FT and have a TS% around 52%. Carmelo Anthony has a TS around 53%, but is vastly superior to the other player. The advanced statistic incorrectly demonstrates that the difference between them is minuscule.

Also lower FG% equates to more missed shot attempts, which turns into defensive rebounds, transition defense, and a proabability of fast break points. But TS% masks lower FG% shooters.

That's not a flaw in TS%. That's a flaw in interpreting it.

A lower FG% also leads to more OREB chances. It evens out.

[PurpleTrees]

Ts% in a nutshell = shoot 3 pointers

Demarre Carrol hitting an open 3 that was created by defensive attention on Demar Derozan = 150%

Demar Derozan drawing a foul after breaking the defense down, getting fouled, hitting the shot, hitting the freethrow = 104%


Stupid stat that needs to be thrown out.