f4p wrote:DraymondGold wrote:Hi f4p! I'd like to push back against the idea that "the preponderance of the evidence is in kawhi's favor [over Curry's] before he was taken out" and "i think we have plenty of evidence kawhi is a better playoff performer than steph". I answered these in my last post (sorry to repeat myself!), but these blanket statements that 2017 postseason Kawhi clearly trumps Curry by the numbers just aren't true. There are plenty of stats that take Curry overall over Kawhi, and even 2017 playoff-only stats.
2017 Postseason only: Postseason APM, Postseason RAPM, postseason AuPM, postseason Backpicks BPM, DARKO, and plenty more take 2017 Playoff Curry over 2017 Kawhi.
we might never see eye to eye on these stats with curry. i put too much into 7 playoff losses in 2016 and then 1 in 2017 after adding KD to think curry's impact is really as high as those metrics would lead one to believe. i put too much into the fact that when klay last played in 2019, the warriors looked like they might win game 6 of the finals, then he missed 2 years and they didn't even make the playoffs (the start of 2020 did not give much hope even if curry played), then he came back and they won the title to think the impact metrics are really seeing everything with curry. the warriors are obviously dominant, but the number simply don't appear to be able to get away from giving all the credit to curry, no matter the circumstances. look no further than curry being better in 2021 and missing the playoffs and worse in 2022 and winning the title to see that there must be plenty of impact to go around.
Thanks for the thoughtful reply! And it's true we might not see eye to eye on these
Like you say, there's definitely quantitative arguments against Curry (you mention PER and WS/48) and qualitative arguments against Curry (2016 loss with Curry playing, 2019 finals loss without Klay/KD, 2020 league-worst performance without Klay, 2021 loss to just miss playoffs without Klay) that go against Curry.
I was just replying against the idea that "the preponderance of the evidence" favors playoff Kawhi over Curry, which definitely isn't true. There's just as many quantitative arguments in favor of Curry (like the stats I mentioned) and qualitative arguments in favor of Curry (2016 had Curry and other Warriors injured, 2019 played well in the playoffs with Curry and no KD, 2020 was missing Curry too, 2021 Warriors were 8th seed which is a playoff-
level team with Curry and no Klay / poor bench). Though if you favor the anti-Curry arguments more, I could understand why
f4p wrote:But the problem is metrics like PER and WS/48 just aren't as accurate as the plus minus metrics.
PER is just about the worst all-in-one metric we have. It has a great feature which is that it goes back so many decades. But we're dealing with players that do have more recent one-number metrics. Actual NBA analysts who work for NBA organizations consistently rate PER as the single worst all-in-on metric at measuring value (
https://hoopshype.com/lists/advanced-stats-nba-real-plus-minus-rapm-win-shares-analytics/), and it's actually the least effective all-in-one metric at predicting future success, at times even worse than points per game haha
(
https://fansided.com/2019/01/08/nylon-calculus-best-advanced-stat/). WS/48 is a little better than PER, but it similarly drops behind all the plus-minus based metrics we have.
i hear this often but i wonder how much it applies to the players we are talking about in GOAT and peak projects. here are the highest prime (age 22-35) playoff PER's of all time:
1. Jordan
2. Lebron
3. Shaq
4. Hakeem
5. Duncan
other than bill russell, who is essentially a walking enigma when it comes to stats and conversations like these, i wouldn't have a problem saying those are my 5 best playoff performers ever. PER seems to be quite good for high volume offensive players, and that's who we are talking about. i have no doubt it is a terrible stat if you are a front office trying to dig into the mid-tier free agency pool and divine impact for your future contract offers. it cannot handle the shane battier's and bruce bowen's of the world.
or ben wallace and bill russell. those are its limitations but they are understood. i think it summarizes the box score for high volume offensive guys quite well. while scaling to league average. and whether the box score is great or not, the best players throughout history have tended to put up great numbers.
That's a really thoughtful argument!
You're right that the top 5 track to all-time players (I believe this is by design -- it was designed to make sure the standard top 3-5 names pop up near the top 3-5). But like you say, it's also not without its questionable ratings (like many other stats). Bill Russell falls under players like LaMarcus Aldridge and Pau Gasol. Garnett falls below Dwight Howard and Amar'e Stoudemire. Larry Bird falls below Russell Westbrook and Tracy McGrady. (For what it's worth, Curry's above Jerry West, Kareem, Magic, Wilt, Kobe, Wade, Bird, Russell)
I think it's worth asking: Why does it underrate Bill Russell? Why does it underrate Larry Bird or Kevin Garnett? I see a bias for on-ball players against players who provide a lot of value in the margins, e.g. off ball, with defensive leadership or communication, in how they warp the defenses, etc. Instead, it over-values big men who are high in rebounding, low in turnovers, and high in efficiency (like Amare Stoudemire).
Here's a second question: Why does it have these biases? To me, this comes back to the fact that PER was designed to just correlate with players that public-opinion and the creator's opinion perceive as valuable. It's purely a (somewhat-arbitrary) combination box score stats (formula here:
https://thezscore.com/2016/02/17/the-definitive-per-criticism/). That's why I favor plus-minus based stats, since these stats aren't just correlated with value, they're actually causally connected to the value a player adds to their team (though to be clear, these aren't perfect either -- see below).
The fact that Curry performs better in plus-minus based stats than in PER (despite still being all-time in PER) makes me confident that some of his offensive value (particularly in the playoffs) comes from his ability to demand so much defensive attention. In the 2017 Finals, per my film analysis, he received doubles or at least had 2 defenders focusing on him in over 60% of possessions he was involved in. In 2018 Finals, he received two-thousand percent more doubles than KD. That's a crazy outlier... I'm not sure if anyone else in history generates that many open looks for teammates in this way, even with other all-stars on the court, without getting basic box-score credit for it (e.g. he's often not credited with an assist in these scenarios where he creates an open dunk or 3 for Kevin Durant of all people
). This kind of value would be missed by PER> I see similar problem with PER in its evaluation of Bird and KG (who are also off-ball players) and Russell (who's more of a "little things" guy on offense, and of course gets value on defense too).
f4p wrote: and PER seems to also move up and down within a player's own career in ways that seem to align with their peaks and valleys. and very importantly, as you mentioned, we have the numbers back to the beginning. we're doing well to get full data on the other numbers on the 90's guys, much less the first half of league history. it is difficult to find high PER playoff performances that don't seem impressive (especially multi-series numbers).
the plus/minus numbers don't go back forever, fluctuate more than i think a player's actual ability fluctuates, feel like team context can matter too much, and seem to pick favorites in ways that don't always seem to line up with the results (individual or team) on the court. the magnitude in difference also seems to be too much for some players, even if you could convince me of the general result of who was better was correct.
maybe i just don't like them because they seem to underrate a lot of guys i pick in these things, but i often pick based on who seems to win the most relative to their team's talent and their opponent's talent. when the plus/minus starts reward too much or too little of the credit to certain people, i tend to tune it out more and more.
Plus minus numbers definitely fluctuate like you say, which is why I'm so big on getting a larger sample size for those numbers. The true plus-minus numbers also don't go back incredibly far, like you say.
If you like having stable metrics in small samples, that's why I tend to favor WS/48 (and better yet Basketball Reference's Box Plus Minus, or even better yet Backpicks Box Plus Minus) over PER. Those are better at capturing value according to NBA organization members (who I imagine know their stuff haha) and perform better at using value to predict future success. The box plus minus stats in particular are better at capturing current value and predicting future success. I also like what they do philosophically more. Unlike PER (which I believe incorporated public opinion of who's best into its design), Box Plus Minus are designed to try to imitate adjusted plus minus stats (which correlate to actual value added). To me, this seems a bit less like circular logic, and feels less subject to the biases of public opinion. And like you mention, these stats also go back for older players!
f4p wrote:How much does Magic's lack of playoff difficulty concern you?
I remember in a previous thread, you mentioned that playoff difficulty was one of your biggest concerns with Curry.
it is definitely concerning. in fact, it's part of how i talked myself into kawhi 2017 over him. i believe 1987 was the year the lakers made the finals playing 3 teams with a combined sub 0.500 record. he almost certainly had an easier road to the finals over his career than lebron in the east, especially when factoring in kareem and worthy. and even lost to undertalented rockets teams twice, showing why lebron making 8 straight is still crazy, no matter the opponent. as for comparisons with non-kawhi people, magic probably benefits in having played just before i was old enough to start watching the league, so he lives on only as a legendary player who won a lot and who i couldn't see fail and who i couldn't see probably cruise through some series with sub-par play because it was so easy that it didn't matter if he played well or not.
but i also have to admit, even when trying to figure out things like putting hakeem over magic all time, that magic really seemed to be pretty amazing. i think he was better than bird and ruled over a decade, his offenses were amazing, he has like a whole highlight reel of crazy buzzer beater jumpers for a guy not known for jump-shooting, and he was still playing at an extremely high level when he retired the first time. it's hard to ignore making the 1991 finals after kareem or the fact kareem struggled to win in the weakest decade ever and then won 5 win magic, though obviously magic was extremely fortunate to be paired with kareem (and then a whole extra #1 overall pick in worthy). so i guess 1987 still feels like an extraordinary peak.
Yep, I believe you're right about 87 Magic playing 3 combined sub 0.500 teams.
And it's definitely true that looking at the end of the 80s, Magic seemed to outperform Bird. Like you say, he wasn't too far back from his his offensive peak in 1991 (though he majorly declined in his defense). It's such a shame we couldn't see either Magic or Bird stay healthy into the 90s. I would have loved to watch post-peak Magic and Bird play against peak MJ! To me, I tend to credit Magic has having a longer prime or better longevity than Bird, rather than a better 1-year peak, but I can also see how his end-career accomplishments might make you feel favorable to Magic at his peak.
I think there's a similar kind of analysis of late-prime success to re-contextualize true peak success applied to other players. For example, Russell's unique leadership in 1968 and 1969 might make us higher on peak ~62-65 Russell's leadership, Wilt's success next to perimeter stars in 72 might make us higher on his peak in 64/67, Kareem's success with the 80s Lakers might make is higher on Kareem's ceiling-raising ability in 77, Hakeem's resilience in 94/95 might make us higher on his resilience in 93, LeBron's playoff success in 2016 might make us higher on his resilience in 2012-2014, and Curry's resilience without KD in 2022 might make us higher on him in 2017-2019. For Bird, I think the fact that he was injured but still played occasionally (rather than just dropping out of the league like Magic) makes it harder to do that analysis... he didn't retire when he was "on top", unlike Magic (ignoring 96) or MJ (ignoring the Wizards years), which might bias us against Bird since our most recent memory was when he wasn't as good. Regardless, it's very valid to use a larger sample size to re-contextualize a 1 year peak, so long as we're careful to account for context and avoid bias.
As a side note: I really enjoy this debate. We definitely don't agree a lot haha, so it's fun to try to push against each other's arguments and occasionally find common ground
