Better Adv. Stat: TS% vs PER vs VORP vs BPM vs WS

[madmaxmedia]

Its like asking whats the better measurement between length, height and width. They all measure different things.

Length, definitely length (that's what she said.)

[JonFromVA]

Statistics (and all information really) are useless if you don't understand what they mean. TS% isn't measuring who the best player is. It's measuring a player's combined shooting percentages by taking into account 3FGAs and FTs -- that's it. TS% isn't pretending to answer any questions about overall value or shot difficulty or anything else, just shooting efficiency for the shots a player takes. The other stats are trying to convince you that they have answers to questions that they don't have. They're too ambitious and because of that frequently incorrect (or correct for the wrong reasons).

Keeping in mind that the version of TS% commonly used is an estimate of shooting efficiency ...

I get how a correct TS% would account for 2 shot fouls vs 3 shot fouls vs and1s. Would you mind clarifying how it treats things like free throws from technicals, or defensive three in the key? Any reason this isn't being used more on at least a going forward basis since this is all being tracked

From basketball-reference ...

TS% - True Shooting Percentage; the formula is PTS / (2 * TSA). True shooting percentage is a measure of shooting efficiency that takes into account field goals, 3-point field goals, and free throws.

TSA - True Shooting Attempts; the formula is FGA + 0.44 * FTA.

That 0.44 factor tries to account for all the free-throws that don't come from 2 shot fouls and reportedly does a very nice job of it. I can't say why sites like nba.com haven't switched to "true true shooting %" or whatever they'd call it, other than that TS% has become well established.

Really this is the tip of the iceberg when it comes to advanced stats. Fudge factors and estimations are very common in them. Individual offensive and defensive rating, for instance assumes that team performance can be divided up to the individual players based on their box score stats. In both cases there are certain classes of players the stat will cheat.

[Hungry]

Keeping in mind that the version of TS% commonly used is an estimate of shooting efficiency ...

I get how a correct TS% would account for 2 shot fouls vs 3 shot fouls vs and1s. Would you mind clarifying how it treats things like free throws from technicals, or defensive three in the key? Any reason this isn't being used more on at least a going forward basis since this is all being tracked

From basketball-reference ...

TS% - True Shooting Percentage; the formula is PTS / (2 * TSA). True shooting percentage is a measure of shooting efficiency that takes into account field goals, 3-point field goals, and free throws.

TSA - True Shooting Attempts; the formula is FGA + 0.44 * FTA.

That 0.44 factor tries to account for all the free-throws that don't come from 2 shot fouls and reportedly does a very nice job of it. I can't say why sites like nba.com haven't switched to "true true shooting %" or whatever they'd call it, other than that TS% has become well established.

Really this is the tip of the iceberg when it comes to advanced stats. Fudge factors and estimations are very common in them. Individual offensive and defensive rating, for instance assumes that team performance can be divided up to the individual players based on their box score stats. In both cases there are certain classes of players the stat will cheat.

Is there a consensus on how a true(er) shooting percentage would handle technical fouls/defensive three in the keys? It seems to me you could,

1) Treat them the same as an and1 (I personally don't like this)
2) Discard points scored off of these shots
3) Discard an expected points scored based on number of shots taken and ft%

I fully get that the .44 factor is close enough and I'm geeking out over something completely minor

[JonFromVA]

I get how a correct TS% would account for 2 shot fouls vs 3 shot fouls vs and1s. Would you mind clarifying how it treats things like free throws from technicals, or defensive three in the key? Any reason this isn't being used more on at least a going forward basis since this is all being tracked

From basketball-reference ...

TS% - True Shooting Percentage; the formula is PTS / (2 * TSA). True shooting percentage is a measure of shooting efficiency that takes into account field goals, 3-point field goals, and free throws.

TSA - True Shooting Attempts; the formula is FGA + 0.44 * FTA.

That 0.44 factor tries to account for all the free-throws that don't come from 2 shot fouls and reportedly does a very nice job of it. I can't say why sites like nba.com haven't switched to "true true shooting %" or whatever they'd call it, other than that TS% has become well established.

Really this is the tip of the iceberg when it comes to advanced stats. Fudge factors and estimations are very common in them. Individual offensive and defensive rating, for instance assumes that team performance can be divided up to the individual players based on their box score stats. In both cases there are certain classes of players the stat will cheat.

Is there a consensus on how a true(er) shooting percentage would handle technical fouls/defensive three in the keys? It seems to me you could,

1) Treat them the same as an and1 (I personally don't like this)
2) Discard points scored off of these shots
3) Discard an expected points scored based on number of shots taken and ft%

I fully get that the .44 factor is close enough and I'm geeking out over something completely minor

I'd assume the same way they treat points per possession. In other words, is a technical foul a possession?

I haven't bothered to look it up, but intuitively, I would think not. So, I'd guess option 1.

But we could compute a derived stat that would exclude points due to non player initiated actions if we had access to the raw data.

[SwatLakeCity527]

Torture the numbers enough and they'll tell you whatever you want.

[Capn'O]

I always start with a player's DEMBO and then measure it against their CONDOR ranking.

[NBAFan93]

Which one says my favorite players are the best? That’s the one I like . Haha, just kidding. I don’t really like these stats cause I think they inflate the value of players who aren’t always playing against the other teams best players and don’t account how a player performs in more important moments of a game like at the end of a close game - unless they do and I don’t know it? Maybe I should get a book on them all so I understand them better outside of the little summaries one can google.

A stat I like, even though it’s not “advanced” in any way, is how many points, rebounds, assists, etc. did they get for their number of minutes played.

[Capn'O]

Aaaaand people here are again fooled by useless stat. TS% is useless to evaluate a player. By that metric, the best player ever is...
DeAndre Jordan
Yep.

No. Jordan has been the most efficient at converting the shots he has taken.

Nobody except Mark Cuban and maybe this guy dk4rth that used to post on our board would extrapolate that into meaning he should have had a significantly expanded offensive role that would place him among the all time greats.

https://www.basketball-reference.com/leaders/ts_pct_career.html

PER is much better at saying which player had the most impact in the game. Just check the top players in PER, they mostly won the MVP that season, or were in contention most of the time.

https://www.basketball-reference.com/leaders/per_season.html

And career overall.
1st Jordan
2nd Lebron
(what most people have in their GOAT lists)
3rd is Davis but he's still entering his prime, while Jordan dragged on the two seasons with the Wizards and Lebron has started to slow down. Just check this list of all time greats:

https://www.basketball-reference.com/leaders/per_career.html

PER is much maligned. It's pretty good at saying - hey this guy is good at putting up counted stats. One of the biggest limitations is that it favors bigs that get a lot of rebounds and have a limited offensive role but convert at high percentages. Do you think, for example, that Ivica Zubac is better than Bradley Beal? PER sure does! It does, however, correctly identify Mitchell Robinson as a god among men.

[Lalouie]

is vorp based on a baseline established by the whole league or just the players who replace a you on YOUR team. because if the latter is the case, players on teams with bad benches or maybe has no backup will have a better vorp...yes?

in other words is there a hard baseline or is it fluid?

i can see, of course, the irreplaceable players like giannis like harden, but how do you value say redick or williams who are tremendous shooters who are better than the players they're replacing.

[valrond1]

Aaaaand people here are again fooled by useless stat. TS% is useless to evaluate a player. By that metric, the best player ever is...
DeAndre Jordan
Yep.

https://www.basketball-reference.com/leaders/ts_pct_career.html

PER is much better at saying which player had the most impact in the game. Just check the top players in PER, they mostly won the MVP that season, or were in contention most of the time.

https://www.basketball-reference.com/leaders/per_season.html

And career overall.
1st Jordan
2nd Lebron
(what most people have in their GOAT lists)
3rd is Davis but he's still entering his prime, while Jordan dragged on the two seasons with the Wizards and Lebron has started to slow down. Just check this list of all time greats:

https://www.basketball-reference.com/leaders/per_career.html

Statistics (and all information really) are useless if you don't understand what they mean. TS% isn't measuring who the best player is. It's measuring a player's combined shooting percentages by taking into account 3FGAs and FTs -- that's it. TS% isn't pretending to answer any questions about overall value or shot difficulty or anything else, just shooting efficiency for the shots a player takes. The other stats are trying to convince you that they have answers to questions that they don't have. They're too ambitious and because of that frequently incorrect (or correct for the wrong reasons).

This thread is about which stats is more defining of the quality of the player. As per the OP:

Which advanced stat is better judging a player?

I have just proven that PER is better for judging a player than TS%. Unless, of course, you think DeAndre Jordan is the best player ever, Cedric Maxwell the 2nd and Tyson Chandler the 3rd. The only good one in that top TS% is Curry.

However, PER has proven to be a much better stat for jugding a player.

Now, you can pretend to say other things, but the question at hand is that. Which is better for judging a player. You take the better TS% team, I take the better PER team.

Your top 15 all time is:

1.
DeAndre Jordan
.6366
2.
Cedric Maxwell
.6294
3.
Tyson Chandler
.6258
4.
Stephen Curry
.6232
5.
Artis Gilmore*
.6227
6.
Karl-Anthony Towns
.6194
7.
Dave Twardzik
.6184
8.
James Donaldson
.6177
9.
Adrian Dantley*
.6166
10.
Reggie Miller*
.6139
11.
Kevin Durant
.6127
12.
Charles Barkley*
.6120
13.
Magic Johnson*
.6095
14.
James Harden
.6093
15.
John Stockton*
.608


My top 15 is:
1.
Michael Jordan*
27.91
2.
LeBron James
27.58
3.
Anthony Davis
27.43
4.
Shaquille O'Neal*
26.43
5.
David Robinson*
26.18
6.
Wilt Chamberlain*
26.13
7.
Bob Pettit*
25.34
8.
Chris Paul
25.24
9.
Kevin Durant
25.20
10.
Neil Johnston*
24.68
11.
Charles Barkley*
24.63
12.
Kareem Abdul-Jabbar*
24.58
13.
James Harden
24.52
14.
Tim Duncan
24.22
15.
Magic Johnson*
24.11

[valrond1]

PER is much maligned. It's pretty good at saying - hey this guy is good at putting up counted stats. One of the biggest limitations is that it favors bigs that get a lot of rebounds and have a limited offensive role but convert at high percentages. Do you think, for example, that Ivica Zubac is better than Bradley Beal? PER sure does! It does, however, correctly identify Mitchell Robinson as a god among men.

So, it favours bigs, so Michael Jordan is the top guy all time. Lebron is number 2.
Number 1 in TS% is Jordan too, but Deandre.

The thing is, only great players have careers with great PERs. Some pretty average players have good TS%.

VORP is also pretty good:

https://www.basketball-reference.com/leaders/vorp_career.html

No bad players there either. But TS% ALONE is pretty miss and hit.

1.
LeBron James
130.90
2.
Michael Jordan*
104.43
3.
Karl Malone*
102.52
4.
Kevin Garnett
93.97
5.
Charles Barkley*
93.53
6.
Tim Duncan
89.31
7.
Kareem Abdul-Jabbar*
86.03
8.
David Robinson*
80.87
9.
Julius Erving*
79.90
10.
Larry Bird*
79.74
11.
Chris Paul
78.64
12.
Jason Kidd*
78.19
13.
Magic Johnson*
77.40
14.
Hakeem Olajuwon*
77.14
15.
Clyde Drexler*
75.55
16.
Shaquille O'Neal*
73.97
17.
Kobe Bryant
72.07
18.
Scottie Pippen*
69.56
19.
Dirk Nowitzki
66.29
20.
John Stockton*
65

[TheNG]

I used the following link to extract some date:
https://www.basketball-reference.com/leagues/NBA_2020_advanced.html
Here is the table of the ranking of these 10 players in each metric (1 means 1st in the NBA, 2 means 2nd etc.):



Most metrics show correctly that Giannis is the best player and Doncic 2nd.
TS is completely not a good indication.
WS puts Lillard as number 1, and puts Kawhi, Jokic and Embiid far behind which obviously is wrong.
PER shows correctly that Doncic is much better than LeBron (2nd vs 9th) but shows incorrectly that Embiid is better than Jokic (7th vs 38th).
So we're left with BPM and VORP. It's hard because both put LeBron in Top5 which is obviously wrong. VORP puts Embiid at 28th which is probably too low, so we're left with BPM.

So BPM is the least bad way from all the alternatives you mentioned.
The best way for career evaluation btw is by counting rings

[valrond1]

I used the following link to extract some date:
https://www.basketball-reference.com/leagues/NBA_2020_advanced.html
Here is the table of the ranking of these 10 players in each metric (1 means 1st in the NBA, 2 means 2nd etc.):



Most metrics show correctly that Giannis is the best player and Doncic 2nd.
TS is completely not a good indication.
WS puts Lillard as number 1, and puts Kawhi, Jokic and Embiid far behind which obviously is wrong.
PER shows correctly that Doncic is much better than LeBron (2nd vs 9th) but shows incorrectly that Embiid is better than Jokic (7th vs 38th).
So we're left with BPM and VORP. It's hard because both put LeBron in Top5 which is obviously wrong. VORP puts Embiid at 28th which is probably too low, so we're left with BPM.

So BPM is the least bad way from all the alternatives you mentioned.
The best way for career evaluation btw is by counting rings

Thank you. You have just proven that the WORST is TS%, by far. Any of the other metrics are a lot more reliable to determine how good a player is.

[Lockdown504090]

one more time for the people in the back : DONT USE 1 STAT BY ITSELF

[JonFromVA]

I used the following link to extract some date:
https://www.basketball-reference.com/leagues/NBA_2020_advanced.html
Here is the table of the ranking of these 10 players in each metric (1 means 1st in the NBA, 2 means 2nd etc.):



Most metrics show correctly that Giannis is the best player and Doncic 2nd.
TS is completely not a good indication.
WS puts Lillard as number 1, and puts Kawhi, Jokic and Embiid far behind which obviously is wrong.
PER shows correctly that Doncic is much better than LeBron (2nd vs 9th) but shows incorrectly that Embiid is better than Jokic (7th vs 38th).
So we're left with BPM and VORP. It's hard because both put LeBron in Top5 which is obviously wrong. VORP puts Embiid at 28th which is probably too low, so we're left with BPM.

So BPM is the least bad way from all the alternatives you mentioned.
The best way for career evaluation btw is by counting rings

I just find it strangely backwards to hold up a stat, rank players by it, then judge whether it's any good by the eye test. I recall Hollinger even ran with this concept with PER tweaking his numbers so the list of the highest rated players "looked right" to him.

If the "eye test" is god ... why bother with the stat?

otoh, if we think a stat is designed to tell us something, why don't we listen to it rather than dismiss it?

And most of all, why do we waste our time looking at regression based stats that aren't accompanied by an error?

[TheNG]

I used the following link to extract some date:
https://www.basketball-reference.com/leagues/NBA_2020_advanced.html
Here is the table of the ranking of these 10 players in each metric (1 means 1st in the NBA, 2 means 2nd etc.):



Most metrics show correctly that Giannis is the best player and Doncic 2nd.
TS is completely not a good indication.
WS puts Lillard as number 1, and puts Kawhi, Jokic and Embiid far behind which obviously is wrong.
PER shows correctly that Doncic is much better than LeBron (2nd vs 9th) but shows incorrectly that Embiid is better than Jokic (7th vs 38th).
So we're left with BPM and VORP. It's hard because both put LeBron in Top5 which is obviously wrong. VORP puts Embiid at 28th which is probably too low, so we're left with BPM.

So BPM is the least bad way from all the alternatives you mentioned.
The best way for career evaluation btw is by counting rings

I just find it strangely backwards to hold up a stat, rank players by it, then judge whether it's any good by the eye test. I recall Hollinger even ran with this concept with PER tweaking his numbers so the list of the highest rated players "looked right" to him.

If the "eye test" is god ... why bother with the stat?

otoh, if we think a stat is designed to tell us something, why don't we listen to it rather than dismiss it?

And most of all, why do we waste our time looking at regression based stats that aren't accompanied by an error?

Once you have a stat you trust more than your "eye test" you should follow it.
However, what you call the "eye test" is actually a very sophisticated neural network trained by millions years of evolution...
I usually use the following algorithm:
When evaluating which player has the best TS% - I use the TS% stat.
When evaluating which player has the best BPM - I use the BPM stat.
When evaluating which player has the best PER - I use the PER stat.
When evaluating which player has the best VORP - I use the VORP stat.
When evaluating which player has the best WS - I use the WS stat.

When evaluating which player is the best- I use my own neural network - giving it the inputs of TS%, BPM, PER, VORP, WS and multiple thousands of other inputs - and let it output the correct answer.

[valrond1]

I used the following link to extract some date:
https://www.basketball-reference.com/leagues/NBA_2020_advanced.html
Here is the table of the ranking of these 10 players in each metric (1 means 1st in the NBA, 2 means 2nd etc.):



Most metrics show correctly that Giannis is the best player and Doncic 2nd.
TS is completely not a good indication.
WS puts Lillard as number 1, and puts Kawhi, Jokic and Embiid far behind which obviously is wrong.
PER shows correctly that Doncic is much better than LeBron (2nd vs 9th) but shows incorrectly that Embiid is better than Jokic (7th vs 38th).
So we're left with BPM and VORP. It's hard because both put LeBron in Top5 which is obviously wrong. VORP puts Embiid at 28th which is probably too low, so we're left with BPM.

So BPM is the least bad way from all the alternatives you mentioned.
The best way for career evaluation btw is by counting rings

I just find it strangely backwards to hold up a stat, rank players by it, then judge whether it's any good by the eye test. I recall Hollinger even ran with this concept with PER tweaking his numbers so the list of the highest rated players "looked right" to him.

If the "eye test" is god ... why bother with the stat?

otoh, if we think a stat is designed to tell us something, why don't we listen to it rather than dismiss it?

And most of all, why do we waste our time looking at regression based stats that aren't accompanied by an error?

Again, this thread is about which stats better tells how good a player is. Most people voted for TS%. It has been proven than TS% is the worst by far.

Last season TS% leaders:

1.
Rudy Gobert • UTA
.682
2.
Dwight Powell • DAL
.682
3.
DeAndre Jordan • TOT
.674
4.
Thomas Bryant • WAS
.674
5.
Clint Capela • HOU
.658
6.
Joe Harris • BRK
.645
7.
Giannis Antetokounmpo • MIL
.644
8.
Stephen Curry • GSW
.641
9.
Montrezl Harrell • LAC
.636
10.
JaVale McGee • LAL
.634

[PistolPeteJR]

Per 36 is bezt in Europ bro.

[Clyde Frazier]

Ban PER from the basketball lexicon forever.

[Cavsfansince84]

For me personally it would go
vorp
WS
bpm
per
ts%