lessthanjake wrote:Doctor MJ wrote:.
I think there were some consistent themes in your responses, so instead of quoting one by one, I think I’ll just consolidate my replies into one list:
1. You mention several times that you think someone is doing the same thing when they make and miss shots. At a fundamental level, I don’t actually agree with that. They may be doing the same thing at a macro-level—as in, they’re doing the same things to try to get their shot off and whatnot (basically the “process” stuff you refer to). But they aren’t actually doing the same thing at a micro-level (i.e. exact shooting mechanics), because if they were then the result would definitionally be the same. And if you do the micro-level stuff differently such that your shot goes in, then I think that means you did better. Basically, actually making the shot definitely matters, and someone pretty much definitionally did not actually shoot the same in their makes and misses, since if they did then the result would be the same. The actual difference in what they did may be really tiny, but there’s definitely a difference and one is absolutely better than the other.
You mention winning bias, and I will say I definitely don’t see that as winning bias. Winning bias is basically saying someone/something was better because they won due to factors out of their control. Saying someone did better because they were successful due to things that were actually in their control (i.e. making a shot) isn’t winning bias. It’s saying that it’s better for someone to have done the things in their control in a better way. Which has to be right. If you won because you actually were better, then it’s not biased to conclude that the person who won was better. It’s only biased if it wasn’t in their control. Of course, someone might’ve done the things in their control better because they had a better “process” or because they just executed things better at a micro-level. In both cases, though, they’re doing the thing better, just for different reasons.
2. You mention multiple times that I’m focused on a drop in TS%, and I don’t think that’s actually a fair portrayal of our exchange. The way this exchange started is that I mentioned to someone else that I don’t think Garnett played as well in the playoffs as he’d played in the regular season. I used data to back up this claim, but it wasn’t really TS% beyond a brief mention of it. Rather, I pointed to BPM, EPM, and WS/48, which all showed Garnett dropping significantly. You then responded with some box data that looked similar in regular season and playoffs and asked if the only difference was therefore “efficiency.” I pointed out that Garnett’s volume went down basically across the board on a per-possession basis, and noted that you were right that his efficiency also went down because his TS% went down and his turnovers went up. Your responses since then have focused a lot on the TS% piece, so that is what my replies to you have focused on. But I don’t think it’s fair to say that I’m actually particularly focused on TS%. For one thing, it’s not really what I led with. Rather, my argument on this initially focused on BPM, EPM, and WS/48. Your response specifically keyed in on “efficiency” but even then I talked about volume decreases as well as another form of inefficiency (i.e. higher turnovers). So, to some degree, I feel like you’ve focused the discussion on TS% and then criticized me for being too focused on TS%.
3. You mention that “if there were infinite universes, each differing based on effective luck, I'd expect that jake in those other universes would be arguing very different things if he used this reasoning.” I think that’s generally true, and it’s something I’m comfortable with. Like, as I said, if we ran the 2004 and 2006 playoffs over again, I’m definitely not certain that Wade would outplay Garnett as much as he did in reality. In fact, my guess is that he probably wouldn’t. But I’m most concerned with what actually happened, not what might’ve happened in a hypothetical world where we had a larger sample. I think it’s perfectly reasonable to take a different approach than I do, and to try to make a holistic assessment of a player’s goodness and go with that. However, that’s not the approach I take to an assessment of “greatness.” I care most about what happened. If what happened might’ve been a result of “luck,” then that’s just the way the cookie crumbled in reality.
One significant caveat I’d add here is that I’m not sure I’d call a player’s own performance “luck,” whereas I would say the performance of other players is more about luck since it’s not under the player’s control. I think the way I’d conceptualize the concept you’re talking about is as randomness, not luck. It’s perhaps just a semantic difference, but I figured I’d mention that.
4. I think you may not be getting the distinction I’m making between variance affecting greatness even if it doesn’t affect goodness. You said “You connect small changes to TS% over small samples to changes in how good a player was playing.” And yeah, I think if a player makes shots more often, then all else being equal, they are actually playing better (see above for some further thoughts on that). However, the fact that they’re playing better in a small sample doesn’t mean I would extrapolate that to a conclusion that they really *were* a better player. So, for instance, let’s say Player A and Player B play equally well across a large sample of games. Then they go into a playoffs, and Player A’s shot is much more on than Player B, but otherwise they both play exactly the same. To me, Player A played better in the playoffs than Player B. And, since it’s the playoffs, that probably would lead to a conclusion that Player A was higher up in a ranking of “greatness.” But it wouldn’t necessarily tell me that Player A is actually a better player than Player B in general. After all, in this hypothetical, we have a large sample of them being equally good, so I have good reason to believe that Player A doing better in a small sample was just random. But I think randomly playing better in the most important games can make someone’s year “greater” than the other guy’s.
You push back on this and say you “reject” the idea that “goodness of play is a different concept from goodness in general.” I think the above probably covers this, but let me just try to clarify more what I mean here. I think goodness of play *in small samples* is different from goodness in general. If someone plays better than someone else in large samples, then I’ll just conclude that they’re better. If someone plays better in a small sample, then I won’t necessarily conclude that that means they’re actually a better player, rather than that they happened to have played better in a small sample of games.
As it applies to 2006 Wade and 2004 Garnett, I look at large samples and they indicate to me that Garnett was probably the better player. But I look at the playoffs and I see Wade having played better in those games. One explanation for that is that the playoffs are different than the regular season and 2006 Wade was simply a better playoff player than 2004 Garnett. That’s possible, but another explanation is that 2006 Wade having played better in the playoffs than 2004 Garnett did is just a result of variance and doesn’t tell us he would’ve been a better playoff player if we had a larger playoff sample. For purposes of the “greatness” of a year, I don’t really care all that much about which explanation is correct, because I’m focusing on what actually happened in that small sample (because what actually happened in that small sample is extremely important to the “greatness” of a player’s year), and I think what actually happened is that Wade played significantly better in the playoffs than Garnett. For purposes of how “good” these players were in those years, though, which explanation is correct does matter. If 2006 Wade would be a better playoff player in a larger playoff sample, then perhaps he was simply the better player, despite being less impactful in regular season samples. But if Wade happened to have randomly played better in those playoffs but Garnett would’ve been better in a larger playoff sample, then Garnett would definitely be the better player. I definitely think it’s possible that the latter is correct. Indeed, I might even think it’s more likely than not the correct explanation. But that doesn’t change the fact that I think Wade played substantially better in the playoffs in reality and that that holds huge independent weight for me in an assessment of the greatness of their years.
5. I think it’s quite likely that rTS% is generally higher in the playoffs than in the regular season. To illustrate this, here’s the difference between the league’s regular season and playoff TS% in the last 25 years (with a positive number meaning that RS TS% was higher):
2025: +1.0%
2024: +1.4%
2023: +1.5%
2022: -0.1%
2021: +0.1%
2020: -0.9%
2019: +0.9%
2018: +0.1%
2017: -1.1%
2016: +0.7%
2015: +0.8%
2014: -0.7%
2013: +0.6%
2012: +0.7%
2011: +1.2%
2010: -0.0%
2009: +0.0%
2008: +0.8%
2007: +1.1%
2006: -1.1%
2005: -0.5%
2004: +1.6%
2003: -0.6%
2002: +0.6%
2001: +0.7%
So, on average in the last 25 years, we have the league’s regular season TS% averaging being only 0.35% higher than playoff TS%. Which is a tiny difference. This means that the average playoff rTS% is higher than the average regular season rTS% as long as the average playoff team had an opponent TS% in the regular season that was at least 0.35% lower than the league’s regular season average. Which seems extremely likely to be the case, given that playoff teams generally had good regular season defenses and 0.35% is a small amount.
6. On the stuff about Duncan/Garnett and 3PAr, you provide a lot of interesting data, which I’ll have to delve into further. One initial reaction is that the teammate 3PAr info is interesting, but I feel like it has to be missing something. After all, if the Timberwolves 3PAr goes down more with Garnett on than the Spurs 3PAr does with Duncan on (which is what the PBPstats data I provided shows), and both Garnett and Duncan shoot a similar volume of shots and neither one shoots threes, then the explanation for that difference basically has to be something about what’s going on with their teammates. But you provide info suggesting that teammates’ 3PAr went up more with Garnett on the floor than with Duncan on. There basically has to be some explanation that squares these pieces of info into a coherent picture. One potential explanation is that the Timberwolves were less likely to play three-point shooters with Garnett than the Spurs were with Duncan. If the Timberwolves put Garnett in lineups with non-shooters more than the Spurs did, then we might see the Timberwolves 3PAr with Garnett go down more, even if the guys who were actually shooters tended to shoot plenty of threes with Garnett on the floor. Another explanation might be that the teammate-3PAr data looks different for the players you didn’t look up. There may be other explanations, but that’s what I can think of. I am losing some steam here and don’t have time to actually delve into data to try to figure out which explanation is right. I tend to think it’s probably not the latter, since you went through a bunch of teammates. So my guess is it’s the former. If that’s the case, then we’d have to ask why different lineup decisions were made with these players. And the explanation may go back to what I was saying—which is that Garnett’s offense is less conducive to the team producing lots of threes. If you have a star whose offense isn’t very conducive to producing a lot of threes, you’ll probably stagger minutes such that your shooters are often on the court when he’s off the court. Which would result in the team 3PAr being lower with that star on the floor, even if individual teammates that are shooters still tended to shoot threes when they took shots with the star on the floor. Again, though, I’m kind of just talking off the cuff here, without taking time to figure out if what I’m saying is borne out in the data. And I know you must’ve spent a long time pulling all the data you provided there, so apologies for not really giving as much effort in my response to that.
