Winshares,winshares/48 and on/off metrics

[f4p]

I think if you were to just go by like a summation of career value I kind of get that, but I also Think it just is a tad bit absurd thinking what’s gone on over the past 7-8 years, even if it’s cheesy or narrative driven I find it hard to compare someone thats probably had the most success in the nba since he hit his stride vs a guy who gets memed for the past few years lol

I think cp3 has been a victim of unfortunate circumstance to an extent for sure though

That is actually like someone in 2013 saying that Dwight had surpassed Garnett because of the past five years. There was much before that.

Curry literally started an offensive revolution and won 4 rings and been to 6 finals in the last 7 years, comparing that to dwight is almost as ridiculous as saying Curry is just a shooter.

which revolution did he start? only he and dame really shoot deep 3's.

[f4p]

You may have seen this before, but for posterity, this study builds off ben's base(or at least claims to):
https://dunksandthrees.com/blog/metric-comparison

so before i go any further, am i reading this table correctly?




they put in the metric of choice, use it to predict, and the RMSE represents the average error on the team net rating?

so like they use EPM, guess +2.48 net rating and then on average, since the RMSE is 2.48 (let's just assume RMSE and MAE are close enough for government work here), the team ends up being either a 0.0 (41 wins) or a +4.96 (54 wins)? so if they plug in WS48 and guess +2.48, it could be -0.37 (40 wins) or 5.35 (55 wins)?

Seems about right, yeah. Though maybe someone whose good at this stuff like unibro can check we're interpreting this properly

so this seems to get at the point i was talking about in our other back and forth about all of these metrics:

...but is still going to be a struggle where i think the error bars (i.e. who actually gets the credit from the lineups) probably exceed the differences we are talking about...

setting aside PER for a second, which isn't even trying to predict anything, the error on EPM is 2.48 and the error on BPM is 2.7 and for WS48 is 2.85. the difference in their errors is an order of magnitude smaller than the error. that's a difference (from EPM) of about 0.6 wins for BPM and 1.0 wins for WS48, presumably against a background guess of 41 wins for the average team. and comparing to RPM basically eliminates any difference with BPM and makes the difference with WS48 about 0.7 wins. these are small differences. EPM isn't God talking to us on a mountain, except instead of the 10 commandments, he's giving us his top 10. and BPM/WS48 isn't a shirtless drunk guy yelling at the end of the bar. even PER manages to trip and fall, without even trying or being the same thing, into only about a 2 win difference with EPM and 1.6 win difference with RPM. the overlap of the predicted value for individual players on these things is going to be pretty substantial and leave us well within the zone of interpretation of what matters more for each player. and of course, the RAPM/PIPM stuff is inbetween the BPM and WS48 errors.

[SpreeS]

Of all the advanced metrics winshares seems to be the best at indicating the value of a player. Contribution to winning is the aim of any player and the team. An owner of a franchise wants to win the most possible as that is success in basketball.

On/off stats are often mentioned as a good indication of the value of a player. It seems logical that the higher the number in this metric the higher the value of a player and thus the higher the contribution to winning.

That is not the case with some players. This is the case with Garnett who has among the highest on/off numbers and relatively low winshares and winshares per 48 minutes, particularly compared to other great players.

What is the view in relation to these statistics?

Really? Wiseman has 3rd WS/48 on Warriors? It must tell you everything about this stat what need to know…

[LukaTheGOAT]

Uhhh, all time? I don’t really see how one could argue Paul over Curry given the past few years lol

Yeah, the past few years Curry has been trying to catchup with Paul. They weren't in lockstep in terms of career value before that.

I think if you were to just go by like a summation of career value I kind of get that, but I also Think it just is a tad bit absurd thinking what’s gone on over the past 7-8 years, even if it’s cheesy or narrative driven I find it hard to compare someone thats probably had the most success in the nba since he hit his stride vs a guy who gets memed for the past few years lol

I think cp3 has been a victim of unfortunate circumstance to an extent for sure though

I mean you basically just explained the argument. And if CP3 had a teammate that surpassed him in playoff plus-minus measures over a 5 year stretch, his championship odds would increase dramatically as well.

[MyUniBroDavis]

Yeah, the past few years Curry has been trying to catchup with Paul. They weren't in lockstep in terms of career value before that.

I think if you were to just go by like a summation of career value I kind of get that, but I also Think it just is a tad bit absurd thinking what’s gone on over the past 7-8 years, even if it’s cheesy or narrative driven I find it hard to compare someone thats probably had the most success in the nba since he hit his stride vs a guy who gets memed for the past few years lol

I think cp3 has been a victim of unfortunate circumstance to an extent for sure though

I mean you basically just explained the argument. And if CP3 had a teammate that surpassed him in playoff plus-minus measures over a 5 year stretch, his championship odds would increase dramatically as well.

I understand the argument of course but it’s just a bit silly I think from a real world perspective to be honest

[MyUniBroDavis]

so before i go any further, am i reading this table correctly?




they put in the metric of choice, use it to predict, and the RMSE represents the average error on the team net rating?

so like they use EPM, guess +2.48 net rating and then on average, since the RMSE is 2.48 (let's just assume RMSE and MAE are close enough for government work here), the team ends up being either a 0.0 (41 wins) or a +4.96 (54 wins)? so if they plug in WS48 and guess +2.48, it could be -0.37 (40 wins) or 5.35 (55 wins)?

Seems about right, yeah. Though maybe someone whose good at this stuff like unibro can check we're interpreting this properly

so this seems to get at the point i was talking about in our other back and forth about all of these metrics:

...but is still going to be a struggle where i think the error bars (i.e. who actually gets the credit from the lineups) probably exceed the differences we are talking about...

setting aside PER for a second, which isn't even trying to predict anything, the error on EPM is 2.48 and the error on BPM is 2.7 and for WS48 is 2.85. the difference in their errors is an order of magnitude smaller than the error. that's a difference (from EPM) of about 0.6 wins for BPM and 1.0 wins for WS48, presumably against a background guess of 41 wins for the average team. and comparing to RPM basically eliminates any difference with BPM and makes the difference with WS48 about 0.7 wins. these are small differences. EPM isn't God talking to us on a mountain, except instead of the 10 commandments, he's giving us his top 10. and BPM/WS48 isn't a shirtless drunk guy yelling at the end of the bar. even PER manages to trip and fall, without even trying or being the same thing, into only about a 2 win difference with EPM and 1.6 win difference with RPM. the overlap of the predicted value for individual players on these things is going to be pretty substantial and leave us well within the zone of interpretation of what matters more for each player. and of course, the RAPM/PIPM stuff is inbetween the BPM and WS48 errors.

This isn’t win difference I don’t think, it’s estimating off and def rtg and just the split of it or something like that

Predictive value wise I remember every time it’s been tested there’s a huge gap between EPM/LEBRON and other things.

BPM are pretty useless in comparison, winshares are completely useless period

[migya]

I think if you were to just go by like a summation of career value I kind of get that, but I also Think it just is a tad bit absurd thinking what’s gone on over the past 7-8 years, even if it’s cheesy or narrative driven I find it hard to compare someone thats probably had the m9ost success in the nba since he hit his stride vs a guy who gets memed for the past few years lol

I think cp3 has been a victim of unfortunate circumstance to an extent for sure though

I mean you basically just explained the argument. And if CP3 had a teammate that surpassed him in playoff plus-minus measures over a 5 year stretch, his championship odds would increase dramatically as well.

I understand the argument of course but it’s just a bit silly I think from a real world perspective to be honest

That's why there's stats and eye witness observation. The world's perception doesn't really matter

[f4p]

Seems about right, yeah. Though maybe someone whose good at this stuff like unibro can check we're interpreting this properly

so this seems to get at the point i was talking about in our other back and forth about all of these metrics:

...but is still going to be a struggle where i think the error bars (i.e. who actually gets the credit from the lineups) probably exceed the differences we are talking about...

setting aside PER for a second, which isn't even trying to predict anything, the error on EPM is 2.48 and the error on BPM is 2.7 and for WS48 is 2.85. the difference in their errors is an order of magnitude smaller than the error. that's a difference (from EPM) of about 0.6 wins for BPM and 1.0 wins for WS48, presumably against a background guess of 41 wins for the average team. and comparing to RPM basically eliminates any difference with BPM and makes the difference with WS48 about 0.7 wins. these are small differences. EPM isn't God talking to us on a mountain, except instead of the 10 commandments, he's giving us his top 10. and BPM/WS48 isn't a shirtless drunk guy yelling at the end of the bar. even PER manages to trip and fall, without even trying or being the same thing, into only about a 2 win difference with EPM and 1.6 win difference with RPM. the overlap of the predicted value for individual players on these things is going to be pretty substantial and leave us well within the zone of interpretation of what matters more for each player. and of course, the RAPM/PIPM stuff is inbetween the BPM and WS48 errors.

This isn’t win difference I don’t think, it’s estimating off and def rtg and just the split of it or something like that

i was translating net rating into wins using 2.7 wins per 1 point. that's what it used to be, perhaps it is slightly different. a quick test on BBRef with predicted win/loss record for this season seems to give me 2.4-2.8 win/NetRtg point depending on which team i pick.

Predictive value wise I remember every time it’s been tested there’s a huge gap between EPM/LEBRON and other things.

BPM are pretty useless in comparison, winshares are completely useless period

that chart in my post is from OhayoKD's article that shows them testing these very things. the gap from EPM to something like BPM is 0.22 compared to overall errors of 2.48 and 2.7. so not even 10% of the actual error.

[ceoofkobefans]

That’s because the other thing that is heavily overvalued by WS is raw efficiency which KG isn’t great in (usually around +2 to +3 rTS in his prime)

The playoff numbers don't show that for Garnett. For 03, 2010 and 12, his ts% was 53.9, 53 and 54.1 and his ws/48 was .141, .148, .149. In comparison Robinson only had comparatively low ws/48 in 94. In 93, 95, 98, 00 and 01 his ts% was 52.9, 53.6, 49.6, 46.3, 53.8 and his ws/48 was .172, .176, .183, .220 and .207. That's much higher than Garnett. Even Garnett's best PO runs in 04 and 08 were 51.3ts% and .163ws/48, 54.2ts%, .199ws/48. That's also low comparable to Robinson's low PO performances. Robinson's 96 and 99 PO runs are much better than Garnett's.

Interesting that Webber had two POs with low ts% and higher ws/48, with high bpm himself. 00 and 02 he had 48.8ts%, .153ws/48 and 9.5bpm, 53.7ts%, .157ws/48 and 5.6bpm.

Wait do you genuinely believe win shares are the best advanced stat or are you trolling, I assumed this was bait because I didn’t think anyone would seriously think that

Yea pretty much any testing we have for impact metrics agrees that WS is a lower tier impact metric

[Lost92Bricks]

I understand the argument of course but it’s just a bit silly I think from a real world perspective to be honest

Fine, but don't just use the "real world" when it comes to CP3, be consistent. You guys rate KG higher than Kobe when you would get laughed at outside of this site for it.

[MyUniBroDavis]

I understand the argument of course but it’s just a bit silly I think from a real world perspective to be honest

Fine, but don't just use the "real world" when it comes to CP3, be consistent. You guys rate KG higher than Kobe when you would get laughed at outside of this site for it.

It his isn’t a take I share so idk why ur talking to me about it

[MyUniBroDavis]

so this seems to get at the point i was talking about in our other back and forth about all of these metrics:



setting aside PER for a second, which isn't even trying to predict anything, the error on EPM is 2.48 and the error on BPM is 2.7 and for WS48 is 2.85. the difference in their errors is an order of magnitude smaller than the error. that's a difference (from EPM) of about 0.6 wins for BPM and 1.0 wins for WS48, presumably against a background guess of 41 wins for the average team. and comparing to RPM basically eliminates any difference with BPM and makes the difference with WS48 about 0.7 wins. these are small differences. EPM isn't God talking to us on a mountain, except instead of the 10 commandments, he's giving us his top 10. and BPM/WS48 isn't a shirtless drunk guy yelling at the end of the bar. even PER manages to trip and fall, without even trying or being the same thing, into only about a 2 win difference with EPM and 1.6 win difference with RPM. the overlap of the predicted value for individual players on these things is going to be pretty substantial and leave us well within the zone of interpretation of what matters more for each player. and of course, the RAPM/PIPM stuff is inbetween the BPM and WS48 errors.

This isn’t win difference I don’t think, it’s estimating off and def rtg and just the split of it or something like that

i was translating net rating into wins using 2.7 wins per 1 point. that's what it used to be, perhaps it is slightly different. a quick test on BBRef with predicted win/loss record for this season seems to give me 2.4-2.8 win/NetRtg point depending on which team i pick.

Predictive value wise I remember every time it’s been tested there’s a huge gap between EPM/LEBRON and other things.

BPM are pretty useless in comparison, winshares are completely useless period

that chart in my post is from OhayoKD's article that shows them testing these very things. the gap from EPM to something like BPM is 0.22 compared to overall errors of 2.48 and 2.7. so not even 10% of the actual error.

That wasnt testing it’s predictive value, so I don’t know why ur referencing the article that you didn’t read

I don’t know of the testing by EPM but the last test I remember on win projections was something like a 10-20 win difference in prediction or something between the good ones and the bad ones

[MyUniBroDavis]

I mean you basically just explained the argument. And if CP3 had a teammate that surpassed him in playoff plus-minus measures over a 5 year stretch, his championship odds would increase dramatically as well.

I understand the argument of course but it’s just a bit silly I think from a real world perspective to be honest

That's why there's stats and eye witness observation. The world's perception doesn't really matter

But your eye witness test thinks Curry is just a shooter and you don’t know how to pick good stats lmao

[OhayoKD]

This isn’t win difference I don’t think, it’s estimating off and def rtg and just the split of it or something like that

i was translating net rating into wins using 2.7 wins per 1 point. that's what it used to be, perhaps it is slightly different. a quick test on BBRef with predicted win/loss record for this season seems to give me 2.4-2.8 win/NetRtg point depending on which team i pick.

Predictive value wise I remember every time it’s been tested there’s a huge gap between EPM/LEBRON and other things.

BPM are pretty useless in comparison, winshares are completely useless period

that chart in my post is from OhayoKD's article that shows them testing these very things. the gap from EPM to something like BPM is 0.22 compared to overall errors of 2.48 and 2.7. so not even 10% of the actual error.

That wasnt testing it’s predictive value, so I don’t know why ur referencing the article that you didn’t read

I don’t know of the testing by EPM but the last test I remember on win projections was something like a 10-20 win difference in prediction or something between the good ones and the bad ones

Tbf I read the article twice and I still don't know what y'all talking about

[OhayoKD]

so before i go any further, am i reading this table correctly?




they put in the metric of choice, use it to predict, and the RMSE represents the average error on the team net rating?

so like they use EPM, guess +2.48 net rating and then on average, since the RMSE is 2.48 (let's just assume RMSE and MAE are close enough for government work here), the team ends up being either a 0.0 (41 wins) or a +4.96 (54 wins)? so if they plug in WS48 and guess +2.48, it could be -0.37 (40 wins) or 5.35 (55 wins)?

Seems about right, yeah. Though maybe someone whose good at this stuff like unibro can check we're interpreting this properly

so this seems to get at the point i was talking about in our other back and forth about all of these metrics:

...but is still going to be a struggle where i think the error bars (i.e. who actually gets the credit from the lineups) probably exceed the differences we are talking about...

setting aside PER for a second, which isn't even trying to predict anything, the error on EPM is 2.48 and the error on BPM is 2.7 and for WS48 is 2.85. the difference in their errors is an order of magnitude smaller than the error. that's a difference (from EPM) of about 0.6 wins for BPM and 1.0 wins for WS48, presumably against a background guess of 41 wins for the average team. and comparing to RPM basically eliminates any difference with BPM and makes the difference with WS48 about 0.7 wins. these are small differences. EPM isn't God talking to us on a mountain, except instead of the 10 commandments, he's giving us his top 10. and BPM/WS48 isn't a shirtless drunk guy yelling at the end of the bar. even PER manages to trip and fall, without even trying or being the same thing, into only about a 2 win difference with EPM and 1.6 win difference with RPM. the overlap of the predicted value for individual players on these things is going to be pretty substantial and leave us well within the zone of interpretation of what matters more for each player. and of course, the RAPM/PIPM stuff is inbetween the BPM and WS48 errors.

I'm really not smart enough to interpret the 20% thing you threw out, so I'd rather let you and unibro duke it out(10 wins would be a pretty big gap, 1.2 wins i guess not though maybe that snowballs into bigger gaps in players).

I think I addressed "its not meant to predict winning" in the other thread though and I think some of that stuff applies even if the stats were equal or edging to the box-stuff. If you wanna continue that convo we should probably do it there

[migya]

I understand the argument of course but it’s just a bit silly I think from a real world perspective to be honest

That's why there's stats and eye witness observation. The world's perception doesn't really matter

But your eye witness test thinks Curry is just a shooter and you don’t know how to pick good stats lmao

That's not what I said, he has nothing but shooting and scoring better than CP. The stats I used are legit and frequently used.

[SinceGatlingWasARookie]

Winshare per 48 had 2017 JaVale Mcgee as something like a top 3 player of all tome in the playoffs amoung players that played more than 10 minutes per game. McGee reakly was good in that playoff season but not that good. Winshres liked that McGee was on a winning team, and had a high points per minute extremely efficient scoring playoff season. But McGee was getting open for lobs because Curry, Durant and Klay had to be tightly covered and Draymond was a great passer, a underrated passer who could make the lob pass for McGee. Birdman Chris anderson playing with LeBron and Wade and Bosh also was way high on that list playing a similar game o what McGee played.

[MyUniBroDavis]

That's why there's stats and eye witness observation. The world's perception doesn't really matter

But your eye witness test thinks Curry is just a shooter and you don’t know how to pick good stats lmao

That's not what I said, he has nothing but shooting and scoring better than CP. The stats I used are legit and frequently used.

You’re using winshares. This isn’t legit. I don’t care if you think they’re a holy grail because they’re objectively bad

I’m sure others have corrected you and demonstrated why using winshares as your end all be all statistical argument is dumb so I don’t know why you’re going to start a topic and then play an impression of a brick wall

[migya]

But your eye witness test thinks Curry is just a shooter and you don’t know how to pick good stats lmao

That's not what I said, he has nothing but shooting and scoring better than CP. The stats I used are legit and frequently used.

You’re using winshares. This isn’t legit. I don’t care if you think they’re a holy grail because they’re objectively bad

I’m sure others have corrected you and demonstrated why using winshares as your end all be all statistical argument is dumb so I don’t know why you’re going to start a topic and then play an impression of a brick wall

Cut out the slight of insult.

If you don't agree with my view that's your opinion, I'm not hindering you from having it but you won't hinder mine.

Winshares is consistent for great players. Bpm is relevant with CP.

[f4p]

This isn’t win difference I don’t think, it’s estimating off and def rtg and just the split of it or something like that

i was translating net rating into wins using 2.7 wins per 1 point. that's what it used to be, perhaps it is slightly different. a quick test on BBRef with predicted win/loss record for this season seems to give me 2.4-2.8 win/NetRtg point depending on which team i pick.

Predictive value wise I remember every time it’s been tested there’s a huge gap between EPM/LEBRON and other things.

BPM are pretty useless in comparison, winshares are completely useless period

that chart in my post is from OhayoKD's article that shows them testing these very things. the gap from EPM to something like BPM is 0.22 compared to overall errors of 2.48 and 2.7. so not even 10% of the actual error.

That wasnt testing it’s predictive value, so I don’t know why ur referencing the article that you didn’t read

I don’t know of the testing by EPM but the last test I remember on win projections was something like a 10-20 win difference in prediction or something between the good ones and the bad ones

huh? from the article:

Team Ratings and Continuity Calculations

Team Adjusted Net Ratings were used in the analysis which were adjusted for strength of schedule (home-court advantage and strength of opponents faced). These ratings were predicted in the analysis based on player metric values from the previous season and prediction errors were assessed.

A measure of roster continuity was used in further analysis to compare metrics in the context of changing rosters for a more accurate assessment. Continuity was calculated using the method outlined by Ken Pomeroy which is the sum of the minimum percent of minutes played out of both seasons for each team. This controls for a differing number of minutes played for each team-season given overtime periods and truncated seasons (lockout and postponed seasons).



OVERALL RESULTS

The first analysis was run using the entire dataset and with rookies given replacement-level values. The table below shows the overall prediction error for each metric (in terms of Root Mean Squared Error, or RMSE). The lower the error, the better the metric predicted the following season’s team ratings.


the guys who wrote the article certainly seem to think they were predicting something.