why ppl use ts% so much?

[mysticbb]

how does that make sense?

So, we have told you multiple times already that a player is not getting just one free throw attempt when he is fouled, but somehow you have managed to not get that. Incredible, really incredible. :)

[Doctor MJ]

lets compare both with actual examples:

player a = 10/10 ft
palyer b = 10/10 2fg
player c = 10/10 3fg

ts%:

player a = 113%
player b = 100%
player c = 150%

"my" pps:

player a = 1
player b = 2
palyer c = 3

which method makes more sense?

"my" pps tells u exactly how many points a player produces every time he shoots the ball. 2fg is worth more than ft therefore player b with 10/10 2fg produces more than player a with 10/10 ft. likewise palyer c produces more than palyer b with same amount of shots.

now lets look ts%

somehow with ts% player a (10/10 ft = 10 pts) produced more, or is more efficient, than player b (10/10 2fg = 20 pts) on same amount of shots.

how does that make sense?

I've just pointed out to you a case where TS% works, and yours doesn't. If you don't understand the example ask for clarification.

You pointing out a scenario where yours works and TS% doesn't basically just puts us at a superficially level playing field where neither metric is flawed.

How do we decide between them? We go by what happens most typically in the NBA. Can't think of another way to do it. When all you can do is approximate the truly correct answer, then you have to give weights that best approximate that answer in a typical scenario.

Which metric does better? Well TS%. The one actually designed recognizing that only an approximation is possible, and thus weighted to come as close to what actually happens as possible.

[Doctor MJ]

the reason why billups might have higher ts% is cause ts% tends to overvalue ft.

That's not true though. Dude, I'm sorry but do you have experience with scientific problems solving. You seem like all you do is think of one absurd scenario, and assume that's the truth. You need to think this through thinking about all the scenarios that stat needs to handle before you design it.

If I'm a player who takes a greater portion of his free throws two at a time than normal because I don't make my FGA as much as normal when I got fouled, then TS% overrates the cost of my FTs.

If I'm a player who takes a greater portion of his free throws in an AND 1 fashion than normal because I do make my FGA more than normal when I get fouled, tthen TS% underrates the cost of my FTs.

The weight given to FTs is based on exactly what the average player does. It's screwing over player in roughly equal amounts on both sides - and THAT is the issue with it, not that it favors one side too much.

[grimballer]

how does that make sense?

So, we have told you multiple times already that a player is not getting just one free throw attempt when he is fouled, but somehow you have managed to not get that. Incredible, really incredible.

so what u saying is:

10/10 ft for 10 pts > 10/10 2fg for 20 pts?

intresting.

[mysticbb]

That's not what I said, but you can very well believe that I said it. ;)

[grimballer]

lets compare both with actual examples:

player a = 10/10 ft
palyer b = 10/10 2fg
player c = 10/10 3fg

ts%:

player a = 113%
player b = 100%
player c = 150%

"my" pps:

player a = 1
player b = 2
palyer c = 3

which method makes more sense?

"my" pps tells u exactly how many points a player produces every time he shoots the ball. 2fg is worth more than ft therefore player b with 10/10 2fg produces more than player a with 10/10 ft. likewise palyer c produces more than palyer b with same amount of shots.

now lets look ts%

somehow with ts% player a (10/10 ft = 10 pts) produced more, or is more efficient, than player b (10/10 2fg = 20 pts) on same amount of shots.

how does that make sense?

I've just pointed out to you a case where TS% works, and yours doesn't. If you don't understand the example ask for clarification.

You pointing out a scenario where yours works and TS% doesn't basically just puts us at a superficially level playing field where neither metric is flawed.

How do we decide between them? We go by what happens most typically in the NBA. Can't think of another way to do it. When all you can do is approximate the truly correct answer, then you have to give weights that best approximate that answer in a typical scenario.

Which metric does better? Well TS%. The one actually designed recognizing that only an approximation is possible, and thus weighted to come as close to what actually happens as possible.

your case was 1 fga vs 1 fga + 1 fta.

u really think thats nuff to make a legit case?

thats like saying

howard is 1/1 (100%) from 3, steve nash is 3/4 (75%),

100% > 75% therefore howard is better from 3.

how about both guys have around 10 attempts before coming up to a conclusion?

in my example i show u how ts% overrates fts. instead of 10 u can put 1000 n ull still get the same result. ft > 2fg. which im sure everybody agrees is not the case.

[grimballer]

That's not what I said, but you can very well believe that I said it.

well ts% does that n thats what was being addressed.

u jumped in convinced im wrong, so plz elaborate.

hows the example i gave not true?

[grimballer]

the reason why billups might have higher ts% is cause ts% tends to overvalue ft.

That's not true though. Dude, I'm sorry but do you have experience with scientific problems solving. You seem like all you do is think of one absurd scenario, and assume that's the truth. You need to think this through thinking about all the scenarios that stat needs to handle before you design it.

If I'm a player who takes a greater portion of his free throws two at a time than normal because I don't make my FGA as much as normal when I got fouled, then TS% overrates the cost of my FTs.

If I'm a player who takes a greater portion of his free throws in an AND 1 fashion than normal because I do make my FGA more than normal when I get fouled, tthen TS% underrates the cost of my FTs.

The weight given to FTs is based on exactly what the average player does. It's screwing over player in roughly equal amounts on both sides - and THAT is the issue with it, not that it favors one side too much.

this is what ur not getting with "my" pps.

1 pps is acceptable number.

as long as the player is producing 1 pps, hes doing a good job.

so instead of looking at just one possession or one fga. look at the entire season.

so if

player a has 700 fga n 300 fta

player b has 300 fga n 700 fta

both players scored 1000 pts.

ts%

player a = 60%
player b = 82%

"my" pps

player a = 1
player b = 1

now we dont know how they scored those 1000 pts or better yet what were their fg n ft %s. all we know player b has a better ts% cause he shoots more fts.

with "my" pps we get straight to the point without exaggerating ft value.

both players took 1000 shots, both scored 1000 pts. therefore they both score 1 pps.

[mysticbb]

grimballer, my last try: TS% measures scoring efficiency. Not some shooting accuracy in particular nor does it tell you exactly how a player scores. The important point is that players have to score as efficient as possible. And the basis for this is 2pt per possession as measuring stick. If a player scores 2 pts and is using 1 scoring possession for this, his TS% is 1 (or 100%). Because a player also attempts to score when he is attempting a free throw, we have to figure that in. To account for the different amount of free throws a player can get per scoring possession the factor 0.44 is used. This factor is derived by analyising multiple years of game data. And it always showed that 0.44 gives the best approximation for the player's amount of scoring possession. Basically: FGA + 0.44 FTA = scoring possessions. We now want to know how efficient a player is using his scoring possessions, thus we are deviding the amount of points by twice the amount of scoring possessions to get TS%.

In your example in average a player who gets 10 FTA used up 4.4 possessions. Obviously the 0.4 makes not much sense in a specific example, but in average it makes a lot of sense. In your particular example the 10 FTA might be the result of a player gets fouled 2 times at the 3pt line and twice for 2 FTA. That makes overall 10 FTA in just 4 possessions. The formula tells us he has a 113.6 TS%, but in reality he has even 125 TS%, because he only used 4 possessions. TS% is now UNDERRATING the scorer from your example. But over the course of a couple of games that will equal out and the 0.44 as a factor works.
Your formula doesn't address that problem at all. Which is the reason it has a lower correlation coefficient to ORtg than TS%.

That's what TS% is measuring, and the correlation coefficient shows it is important.

[grimballer]

grimballer, my last try: TS% measures scoring efficiency. Not some shooting accuracy in particular nor does it tell you exactly how a player scores. The important point is that players have to score as efficient as possible. And the basis for this is 2pt per possession as measuring stick. If a player scores 2 pts and is using 1 scoring possession for this, his TS% is 1 (or 100%). Because a player also attempts to score when he is attempting a free throw, we have to figure that in. To account for the different amount of free throws a player can get per scoring possession the factor 0.44 is used. This factor is derived by analyising multiple years of game data. And it always showed that 0.44 gives the best approximation for the player's amount of scoring possession. Basically: FGA + 0.44 FTA = scoring possessions. We now want to know how efficient a player is using his scoring possessions, thus we are deviding the amount of points by twice the amount of scoring possessions to get TS%.

In your example in average a player who gets 10 FTA used up 4.4 possessions. Obviously the 0.4 makes not much sense in a specific example, but in average it makes a lot of sense. In your particular example the 10 FTA might be the result of a player gets fouled 2 times at the 3pt line and twice for 2 FTA. That makes overall 10 FTA in just 4 possessions. The formula tells us he has a 113.6 TS%, but in reality he has even 125 TS%, because he only used 4 possessions. TS% is now UNDERRATING the scorer from your example. But over the course of a couple of games that will equal out and the 0.44 as a factor works.
Your formula doesn't address that problem at all. Which is the reason it has a lower correlation coefficient to ORtg than TS%.

That's what TS% is measuring, and the correlation coefficient shows it is important.

again lets play your possessions game

6 possessions

player a = 12 fts, 12/12 for 12 pts
player b = 6 2fg, 6/6 for 12 pts

ts%

player a = 113%
player b = 100%

so again same amount of possessions n points yet somehow player with more ft is more efficient?

really?

[Doctor MJ]

grimballer,

I'm sorry. I really don't know where to go from here. I really don't want to be an ass here, but we keep laying out for you the fundamentals of how people go about designing and evaluating stats, and you come back not by pointing out a subtle flaw about what the collective intelligence of a ton of extremely intelligent people missed (which happens from time time to time), but fundamentally never seeming to understand what we're talking about.

I'm not saying you are incapable of understanding this stuff, but I think you're going about your approach in an unproductive manner. You need to first seek to understand exactly what we all see in TS%. And make no mistake there is plenty to like about the stat. So first figure that out, and once you see the good, you will see the more nuanced problems with it, and can consider trying to improve it.

All of the tools you need to do this have been laid out before you. Now you need to spend some time thinking, and doing so objectively. If by some chance you think that you are so incredibly smart we just can't grasp what you are talking about, you need to back up. Communities of smart people make errors, but they don't fumble around blindly with everyone accepting what one person has done. Many of us have come into basketball statistics, looked at the stats available, and determined TS% is the best approximation we have for shooting efficiency despite the fact that it isn't available on ESPN, and it is actually slightly less promoted than eFG% by basketball-reference. You need to figure out how we came to that conclusion - and again, everything you need is in this conversation.

[EvanZ]

so if

player a has 700 fga n 300 fta

player b has 300 fga n 700 fta

both players scored 1000 pts.

ts%

player a = 60%
player b = 82%

"my" pps

player a = 1
player b = 1

now we dont know how they scored those 1000 pts or better yet what were their fg n ft %s. all we know player b has a better ts% cause he shoots more fts.

with "my" pps we get straight to the point without exaggerating ft value.

both players took 1000 shots, both scored 1000 pts. therefore they both score 1 pps.

grimballer, I would call you a troll, but few trolls are clever enough to actually try to pass this off as a legitimate argument.

For the moment, I'm going to assume you are not a troll, and perhaps, just hopelessly confused. Sometimes, my brain goes on autopilot, and I've had occassions where I didn't realize how silly sounding I was coming off in a forum, because I kept saying something that was obviously a complete error. I'm going to give you the benefit of the doubt, for now, that this is what is going on with you. So I will try this in good faith.

You say that both players in your example took 1000 shots. Player A clearly had at least 700 shots, and player B clearly had at least 300 shots. Each FGA is a shot. I think (hope, uh, pray) that we can all agree there.

The question is how many shots does 700 or 300 FTA count for? Now, my understanding is that you are saying each free throw attempt counts as a single shot. Is that correct? I will wait for your reply.

[D-Wags#13]

[
so if

player a has 700 fga n 300 fta

player b has 300 fga n 700 fta

both players scored 1000 pts.

ts%

player a = 60%
player b = 82%

"my" pps

player a = 1
player b = 1

now we dont know how they scored those 1000 pts or better yet what were their fg n ft %s. all we know player b has a better ts% cause he shoots more fts.

with "my" pps we get straight to the point without exaggerating ft value.

both players took 1000 shots, both scored 1000 pts. therefore they both score 1 pps.

well, i usually never come to this board and have never posted here, but reading through the thread I felt the need to do so.

let's just go to your example.

let's assume all 300 fts of player a come on fouls on 2pointers (no and1s, no fouls on threes), to score that 1000 points, he needed 850 posessions.
player b on the other hand would only need 650 posessions for the same amount of points.

if all those free throws come on and1s, player a would need his 700 posession for those 1000 points. player b would need 500 posessions (his 300 fgs being and1s and the remaining 400 fts split by 2 assuming all are fouls on twos that he didn't convert.) for 1000 points.

if all the fts come from fouls on threes, player a would need 800 posessions for his 1000 points, player b 533.

assuming you do have 1000 posessions, player b gives your team more remaining posessions to put on top of that 1000 points in all scenarios.

if you had to choose between these two players. which one would you take? and could you tell if all the information you have pps as the only information?

btw.: why do you start a discussion when you're convinced your truth is the only truth even if proven wrong?

[azuresou1]

TS% needs to not have a set coefficient for And-1 rate. There's no reason why, in this day and age, we should have to use a league-average And-1 rate when we could easily calculate the actual And-1 rate of individual players.

[Fencer reregistered]

TS% needs to not have a set coefficient for And-1 rate. There's no reason why, in this day and age, we should have to use a league-average And-1 rate when we could easily calculate the actual And-1 rate of individual players.

That's my concern, except that I'll confess to not knowing whether those statistics are collected accurately yet, and in particular whether they're well-collected historically. (My guesses are yes for the present day, no for 20+ years ago.)

What you really want is a measure for each player's points per possession, where a possession can be either a shot (with or without an And-1) or a set of multiple free throws (2 or 3 as the case may be). Technical free throws are probably rare enough to be safely ignored.

It would be nice to also count possessions that end in TOs, but since there are several things you might have been trying to do before the TO, and only one of those is score, including TOs doesn't seem realistic.

OK -- all that is obvious, and TS% is only a proxy for the real figure. Why is that proxy used?

[Chronz]

Don't mean to bump an old thread but I just have to applaud the amount of patience you guys have shown the op. If 3 pages aren't enough to get a concept as easy as ts% across I would've started flaming the guy (intentional or not).

[batmana]

The classic example is Chauncey Billups. He is a very low percentage shooter, .417 for his career. However, because he shoots a very large percentage of 3 point shots AND draws a lot of fouls, he is actually a very efficient shooter in terms of points created for each possession, ts% of .583. More efficient than, say, Tony Parker who shoots .498 (very good), but shoots 3's very rarely and is only average at drawing fouls for a ts% of .548. Assuming the multiplier for shooting fouls is reasonably consistent for both, that means that if Parker takes 100 shots, he will score about 55 points whereas if Billups takes 100 shots he will score 58 or more points on average -- where shots includes shots where a foul is called which the league doesn't count as a shot for some reason.
Does that mean Billups is a more explosive scorer than Parker? No, it says nothing about scoring volume, role on a team, ability to create your own shot, playmaking (either the player's or his teammates', etc.). It only says that on the average for the same number of possessions that end in a shot (or shooting foul), Billups will score a bit more points -- nothing else.

A bit off-topic (nothing to do with ts%, particularly I don't like it and find little use in it but I don't want to say others might not find it useful) but the bolded part really shocked me... Are you serious? For some reason? How about being hacked while trying to shoot. If you want to keep a stat about how often someone scores WHILE being fouled and have shooting percentages around .100, fine. But you can't count those towards the normal shots, sorry.

[EvanZ]

It's funny to see discussion of Chauncey Billups. He came in 10th in my total scoring metric, and just as penbeast suggested, it was based entirely on the strength of 3-pt and foul shooting. Parker came in 19th on the strength of inside shooting, but average 3pt/foul shooting.

http://www.goldenstateofmind.com/2011/1 ... le-scorers

[Elden Payton]

Knowing how many points a player produces every time he shoots the basketball is useful information for people who like basketball.

This made me laugh because it's true.

[Dr Positivity]

Because it's simple and because its meaning (points per shot) seem mildly important

I do think though that like most stats, it is applied in the wrong places far too much - and like most stats that are a additive of smaller ones, you can get much better information from looking at the smaller ones separately and drawing conclusions from that, it just takes longer for people

One stat to rule them all is usually how these things work. No greater evidence than in baseball where it's painfully obvious that a single is worth more than a walk and that'd it'd be quite easy to assign rough values to walks in all the different situations (no bases, man on 1st, man on 2nd, man on 1st and 2nd, man on 2nd and 3rd, bases loaded) from a value ranging from 1.0x single for no bases down the list to something much closer to .1x a single for a man on 2nd w/ 1st empty - which would drastically change the results - but plain old flawed OPS remains. Nor in hockey where 95% of the important parts of a game aren't captured in the scoresheet, but ppg and GAA is all anyone can seem to talk about

So TS% is pretty powerful in that sense. One stat to tell you how good a scorer someone is. Finding that is a myth like the holy grail but people will latch onto the best they can get.