WNBA RAPM, etc.

[uberhikari]

Are there any sites where you can find advanced metrics like RAPM, PIPM, EPM, etc for WNBA players? Is there anyone who even calculates advanced metrics for WNBA players at all?

[Doctor MJ]

Great question uberhikari and I think I'll start posting stats in this thread that can be added to by others.

It should come as no surprise that WNBA advanced metrics have always been less publicly available than for the NBA, but I'll start off just by pointing to the WNBA's site which I think has done a great job with the basics - the same is now true for the NBA's site, but it took nba.com a long time to really be useful relative to b-r.com, whereas b-r.com only got serious about the WNBA recently.

WNBA Player Traditional Stats

If memory serves, they have raw +/- data for both regular & post season going back to 2001, and regular season only back to the beginning in 1997. Classic box score is available for the post season going back to 1997 as well.

Additionally, the play-by-play bible of a site [url]pbpstats.com[/url] has WNBA data going back to 2009.

After that it tends to get spotty just based on specific analyses people ran.

Here's a set of RAPM studies from 2006-2015 put into a database from Across the Court (who has been known to post on RealGM). My recollection is that these are 3-year models with weighting 1-2-3, meaning the last year of the trio was weighted 3 times as much as the 1st year.

Here's a stat from Positive Residual they call Estimated Contributions. Linking to the explanation - which then has a link to an actual table - because I think the explanation is necessary to be useful. It involves RAPM but tries to go beyond it.

Here's one for women's NCAA done in 2019 by Jesse Fischer.

There are also some Github accounts that come up when you google for WNBA RAPM for those looking to get into the code themselves.

Last I'll share a collection of spreadsheets I've made at one time or another here which I don't think has any personal subjectivity from me in them (I have others that are based on my opinions):

Doc's WNBA Compendium

You'll find things in there such as the W25 list and its precursors. The thing there I've spent the most time on is the Yearly PM stuff which I've gathered in a dumb, pain-staking way from wnba.com.

I'd love to see discussion about any and all of this but will leave it at that for now.

And I really do encourage people to add to the list. It's entirely possible that there's new awesome stuff I'm not aware of.

[MrDollarBills]

Great question uberhikari and I think I'll start posting stats in this thread that can be added to by others.

It should come as no surprise that WNBA advanced metrics have always been less publicly available than for the NBA, but I'll start off just by pointing to the WNBA's site which I think has done a great job with the basics - the same is now true for the NBA's site, but it took nba.com a long time to really be useful relative to b-r.com, whereas b-r.com only got serious about the WNBA recently.

WNBA Player Traditional Stats

If memory serves, they have raw +/- data for both regular & post season going back to 2001, and regular season only back to the beginning in 1997. Classic box score is available for the post season going back to 1997 as well.

Additionally, the play-by-play bible of a site [url]pbpstats.com[/url] has WNBA data going back to 2009.

After that it tends to get spotty just based on specific analyses people ran.

Here's a set of RAPM studies from 2006-2015 put into a database from Across the Court (who has been known to post on RealGM). My recollection is that these are 3-year models with weighting 1-2-3, meaning the last year of the trio was weighted 3 times as much as the 1st year.

Here's a stat from Positive Residual they call Estimated Contributions. Linking to the explanation - which then has a link to an actual table - because I think the explanation is necessary to be useful. It involves RAPM but tries to go beyond it.

Here's one for women's NCAA done in 2019 by Jesse Fischer.

There are also some Github accounts that come up when you google for WNBA RAPM for those looking to get into the code themselves.

Last I'll share a collection of spreadsheets I've made at one time or another here which I don't think has any personal subjectivity from me in them (I have others that are based on my opinions):

Doc's WNBA Compendium

You'll find things in there such as the W25 list and its precursors. The thing there I've spent the most time on is the Yearly PM stuff which I've gathered in a dumb, pain-staking way from wnba.com.

I'd love to see discussion about any and all of this but will leave it at that for now.

And I really do encourage people to add to the list. It's entirely possible that there's new awesome stuff I'm not aware of.

Thank you Doc!!!!! Wow

[Flizop]

What are your thoughts on Angel Reese's ROTY case? She surpasses CC in all impact metrics, and their teams have similar records.

[Doctor MJ]

What are your thoughts on Angel Reese's ROTY case? She surpasses CC in all impact metrics, and their teams have similar records.

It could get very interesting between Clark & Reese.

When talking about the NBA ROY I wont to point out that the history of the ROY is not one that follows "Most Valuable Rookie", but rather tends to focus on which player did the most that season to convince the world that everyone needs to pay attention to that rookie.

I'd say that Reese will have a very strong case to be the Most Valuable Rookie, but will not be the Rookie Everyone Should Pay Attention to, which I'd think generally would mean she'd lose on the ROY to Clark.

But that's where we get into the matter that any debate between these two will end up extremely polarized, and that may end up giving the nod to Reese. First, any voter who never really thought that hard before might not have a rebuttal to "Reese is the Most Valuable Rookie". Second, any voter who feels a pull not to simply give the nod to Clark looks like they'll have a clear alternative in Reese.

[style_0]

A few I haven't seen mentioned yet...

BBall Index LEBRON. The WNBA section can be accessed by tab on the right.

This box score metric made by Kevin Ferrigan, with other helpful stats like box creation and passer rating on that page as well.

And also PIPM Jacob Goldstein shared on twitter covering 2007-2018.

Read on Twitter

[Ghetto Gospel]

estimated RAPTOR and consensus player rankings for 2024

https://neilpaine.substack.com/p/2024-wnba-elo-power-rankings

[Doctor MJ]

A few I haven't seen mentioned yet...

BBall Index LEBRON. The WNBA section can be accessed by tab on the right.

This box score metric made by Kevin Ferrigan, with other helpful stats like box creation and passer rating on that page as well.

And also PIPM Jacob Goldstein shared on twitter covering 2007-2018.

Read on Twitter

Awesome, I hadn't seen any of these particular studies before! (Same for you Ghetto Gospel).

I'll say that of the group, I'm pretty impressed by what I see from the PIPM model in terms of matching surprisingly well with how I see things. Of course that may also be something of a red flag for my bias.

[cupcakesnake]

A few I haven't seen mentioned yet...

BBall Index LEBRON. The WNBA section can be accessed by tab on the right.

Unfortunate they haven't updated it for this season yet. Hopefully they will at some point?

[style_0]

Unfortunate they haven't updated it for this season yet. Hopefully they will at some point?

Yea, maybe they're waiting for the sample size to be a bit bigger or something, but one of the BBall Index guys did seem to tease this season's data on twitter the other day, so maybe it will be available soon.

[Doctor MJ]

So I just realized that b-r now has +/- data going back to 2008, scratch that, goes back all the way to 1997 for RS & PS. That's new as far as I know. While we already have data going back further, there's an ease of use on b-r for particular things that will be nice for seeing more from the players covered by that time period. Going to see what views I can come up with that I (and others) might find interesting.

[MrDollarBills]

So I just realized that b-r now has +/- data going back to 2008, scratch that, goes back all the way to 1997 for RS & PS. That's new as far as I know. While we already have data going back further, there's an ease of use on b-r for particular things that will be nice for seeing more from the players covered by that time period. Going to see what views I can come up with that I (and others) might find interesting.

Doc you bring a ton of great insight already so getting some metrics on top of that will be very much appreciated

[Doctor MJ]

Alright, so when I do my own stat sleuthing I tend to work with really simple stuff with superficial meaning everyone understands. The deeper truth about how valuable or impressive something was of course requires more context, and often statistical tools that are pretty opaque, and while they're worthwhile, I like the simple stuff as a starting point.

So here's some stuff based on the top player in the league by all-season (RS + PS) +/- for both the NBA & WNBA going back to the start of the era of availability, which now for both leagues is 1997. Not that this isn't On/Off, it's just counting up what happened when a player was on the court.

Multiple time league leaders in the NBA:

1. Steph Curry 4
2. LeBron James 3
3. Tim Duncan 2
(tie) Kevin Garnett 2
(tie) Michael Jordan 2
(tie) Shaquille O'Neal 2
(tie) Jayson Tatum 2

Note that the only Jordan Bulls years in this sample are his final 2, so he likely had other such years

Multiple time league leaders in the W:

1. Maya Moore 5
2. Cynthia Cooper 3
3. Sue Bird 2
(tie) Elena Delle Donne 2
(tie) Lisa Leslie 2

For those not aware, Moore just shows up constantly in any stat analysis I seem to do. I was never in particular a Moore fan, but her data just pops like crazy.

Oldest league leaders in the NBA:

1. Michael Jordan 34
2. David Robinson 33
3. Dirk Nowitzki 32
4. Kevin Garnett 31
5. Steph Curry 30
(tie) Tim Duncan 30

Even more clearly Jordan pops here.

Oldest league leaders in the W:

1. Sue Bird 37
2. Cynthia Cooper 36
3. Alysha Clark 32
(tie) Diana Taurasi 32
5. Tamika Catchings 29
(tie) Elena Delle Donne 29
(tie) Lisa Leslie 29
(tie) Sheryl Swoopes 29

No surprise that Cooper shows up here analogous to Jordan, but Bird being a league leader in 2018 is more interesting. With there being an 8 year gap between the first time she led the league in +/l in the second, she also has the widest range as a league leader and the biggest gap between years as league leader.

Youngest league leaders in the NBA:

1. Jayson Tatum 23
2. Tim Duncan 24
(tie) Dwight Howard 24
(tie) LeBron James 24
(tie) Dirk Nowitzki 24

I think even most NBA obsessives would be shocked to understand the scale of how impressive Tatum's been by +/- at such a young age. I say this as someone reluctant to champion myself - there are warts to his game, and I don't think he quite holds up against the toughest of defenses like you'd want him to given his primacy.

Youngest league leaders in the W:

1. Maya Moore 22
2. Lauren Jackson 23
(tie) Nicole Powell 23
(tie) Ruth Riley 23
5. Elena Delle Donne 25
(tie) A'ja Wilson 25

Yeah, there's Moore again, and we should note that she's the only rookie league leader in this category in either NBA or WNBA that we see (again, only going back to '96-97).

[ellobo]

I don't pay much attention to all-in-one type metrics, but besides wnba.com and BB-Ref, some sites I use for WNBA stats are:

https://www.pbpstats.com/

https://www.positiveresidual.com/shiny/wnba/

https://herhoopstats.com/

https://acrossthetimeline.com/

[Ice Man]

Alright, so when I do my own stat sleuthing I tend to work with really simple stuff with superficial meaning everyone understands. The deeper truth about how valuable or impressive something was of course requires more context, and often statistical tools that are pretty opaque, and while they're worthwhile, I like the simple stuff as a starting point.

So here's some stuff based on the top player in the league by all-season (RS + PS) +/- for both the NBA & WNBA going back to the start of the era of availability, which now for both leagues is 1997. Not that this isn't On/Off, it's just counting up what happened when a player was on the court.

My initial thought after glancing through your list, without yet understanding your metric, is that its view on Jayson Tatum's contributions is far different than those from other impact metrics. For example, in the season that Jayson Tatum was 23 years old (per basketball-reference), when your stat has him #1 in the league, Tatum shows up as #7 in VORP, #14 in BPM, and #8 in Win Shares.

In fact, in my memory, Tatum has never been a Top 5 player by any of those measures.

[Doctor MJ]

Alright, so when I do my own stat sleuthing I tend to work with really simple stuff with superficial meaning everyone understands. The deeper truth about how valuable or impressive something was of course requires more context, and often statistical tools that are pretty opaque, and while they're worthwhile, I like the simple stuff as a starting point.

So here's some stuff based on the top player in the league by all-season (RS + PS) +/- for both the NBA & WNBA going back to the start of the era of availability, which now for both leagues is 1997. Not that this isn't On/Off, it's just counting up what happened when a player was on the court.

My initial thought after glancing through your list, without yet understanding your metric, is that its view on Jayson Tatum's contributions is far different than those from other impact metrics. For example, in the season that Jayson Tatum was 23 years old (per basketball-reference), when your stat has him #1 in the league, Tatum shows up as #7 in VORP, #14 in BPM, and #8 in Win Shares.

In fact, in my memory, Tatum has never been a Top 5 player by any of those measures.

Ah let me declare the vocabulary I'm using to avoid confusion.

When I discuss "impact metrics" as a family, I'm talking about regression based models that are based on team correlation rather than traditional box score components, mostly these are termed with some version of "plus minus". I would term those traditional box score components and their derivatives to be "production metrics".

These are not terms I've made up by myself, rather it's a specific parlance than I'm used to using over on the PC board and elsewhere.

So then to be clear: I would say that Tatum's stature in the league is based more on his production artifacts than his impact artifacts, and while this is true for most players, and that in his case, that means I think most would be surprised at how good Tatum looks by impact metrics.

Now, I'm a bit uncomfortable calling raw +/- as an "impact stat" actually, because I think there really needs to be some On/Off aspect of things before the physical analogy to the concept of "impact" really has any basis, but I use it precisely because it is so, so simple.

I think people aren't necessarily clear what I've meant so far with raw +/-. So last year, if you take all the +/- for both the regular and post-season in the NBA, here's the leaderboard:

1. Tatum (Bos) +755
2. Jokic (Den) +695
3. White (Bos) +686
4. Shai (OKC) +670
5. Hauser (Bos) +578

Clearly I don't think anyone thinks Hauser was a top 5 player last year and that speaks to the obvious issue with using this data to the exclusion of everything else.

But still, we can all agree that if you top the list, it's a good thing, right? This in contrast to a more complex algorithm where we would presume we know a more positive number means something good, but most users don't actually know how these stats are created.

We still want to be able to use the complex algorithms obviously, but I think there's power in starting with something you really know what "1 point" means.

I'm remiss to submit my post without adding an impact measure that has Tatum high (like top 5) among league leaders. Just doing a quick Google here's one RAPM study that was done over a 5-year period from '17-18 to '21-22, which happen to be Tatum's first 5 years in the league.

1. Curry +6.5
2. Giannis +6.3
3. Paul +6.2
4. Embiid +5.8
5. Tatum +5.4'

Now again, am I saying this should be our Top 5 players for this half-decade? No, mine doesn't looking like this...but stuff like this was a surprise to me when I first realized it, because I didn't think Tatum was that impressive when I watched him. I thought he was great don't get me wrong, but he comes off here better than I saw him.

[Doctor MJ]

Okay, here's some data from what I would call the simplest kind of impact stat: On/Off.

The number represents the difference between your PM per 100 possessions when you're On the court, compared to what the team does with you Off. This is regular season data to be clear - b-r now has this for playoffs as well, but the sample is super-small.

Here's a leaderboard based the W25 and some recent players:

1. Maya Moore +20.1
2. Breanna Stewart +14.9
3. Lauren Jackson +14.7
4. Jonquel Jones +14.4
5. Elena Delle Donne +13.9
6. Cynthia Cooper +13.6
7. Lisa Leslie +12.9
8. Tamika Catchings +12.5
9. Alyssa Thomas +11.8
10. Kelsey Plum +10.4

For comparison, if I do the same thing based on the PC Board's Top 100 for the +/- era, here's what we get:

1. Nikola Jokic +11.8
2. Kevin Garnett +11.3
3. LeBron James +10.8
4. John Stockton +10.8 (doesn't include most of career)
5. Steph Curry +10.4
6. Karl Malone +10.3 (doesn't include most of career)

I think contrasting the difference is significant because I do think there's evidence of top women being more impactful than top men in their respective leagues, but as you might guess I am taking particular note of Moore way out in front. The gap between here and #2 among this group (Stewie) is better than Stewie and all others listed.

Moore obviously retired after 8 years so that probably helps her in this stat, but we can also ask:

How many +20 On/Off seasons did the player have, based on that +20.1 On/Off standard she sets. here's that leader board:

1. Maya Moore 4
2. Tamika Catchings 3
(tie) Lisa Leslie 3
(tie) Breanna Stewart 3
5. Kevin Garnett 2
(tie) Nikola Jokic 2
(tie) 7 other WNBA players 2

Again an advantage for WNBA players, but we note that Moore did this more than anyone else.

Note with all of this: I may well have missed players who belong on this leaderboard. Please do let me know if you find such an error.

[zimpy27]

Unfortunate they haven't updated it for this season yet. Hopefully they will at some point?

Yea, maybe they're waiting for the sample size to be a bit bigger or something, but one of the BBall Index guys did seem to tease this season's data on twitter the other day, so maybe it will be available soon.

If it's like NBA they usually wait for 40 games which is the midpoint of NBA season but the whole season in WNBA.