eminence wrote:I liked the criteria thread that was made last time around, so here goes for me, stream of thought on criteria.
1. I focus on cumulative career value, with the primary unit of measurement being the season. Games in a season, possessions in a game, games/minutes played by a top player have all changed significantly over the years. But it's always one season (well, minus those split league years where there were kinda two seasons at once). A season in the 1950's having the same baseline value for me as one in the 2020's (even seasons right next to one another may not be identical, but I'm not starting from a point of 'seasons in the 50's are worth 60% as much as ones in the 10s').
2. To determine seasonal value I have what I would call a 'Corp-adjacent' technique. The thought of value assigned per season is similar, but not strictly about determining # of championships, as I believe that requires league analysis (league structure, playoff structure, etc) beyond evaluating the talent/value of the players. Informally I'd assign rough value tiers along these lines:
-Replacement Level Player (no actual career value assigned to these seasons/players)
-Bench Player
-Starter
-All-Star
-All-NBA
-MVP
-ATG
-All the above are relative to the other players playing at the time or near to (comparing seasons a few apart is fair game).
-Tiers are more about degrees of separation from the pack than ordinal ranking from the top, eg there were a few years in between Mikan and Russell where no MVP or higher level players graced the game, despite there always being a literal 'Most Valuable Player'.
-I do not assign negative value seasons to any player ('16 Kobe comes to mind).
3. I do offer a slight upwards curve on longevity to players who retired prior to or near ABA creation (significant healthcare gains, more lucrative careers, etc).
4. Mikan/Russell uniquely get a bit of a 'no more ghosts to chase' bonus. I go back and forth on giving Magic a boost due to the unique circumstances of his first retirement. At a minimum I use it as a very strong tiebreaker.
5. I generally feel modern APM style stats do a pretty good job of capturing the value I talked about in #2, but they are certainly not perfect. WOWY type measures prior to that have some lesser use, but are of limited availability. If even WOWY type things are unavailable minutes or games played and team quality isn't worthless (lots of minutes on a good team, you're probably at least alright). Box-score stats are about showing the shape of a players impact or to point towards likely increases/decreases in impact.
If you have any thoughts let me know, questions, things I missed, etc.
The bulk of this matches closely with my approach (little deviation with #2 but otherwise right in line).
If I were to add on for my own standards…
A.) My process features a mild element of weighing “accomplishments”, e.g. I struggle to place Karl Malone above Julius Erving even though I see the case for Malone having an overall more “valuable” career and think he could have conceivably won just as much as Erving did if in some sense they switched places.
B.) On that note, hypotheticals can be relevant, but unprovable abstractions like “Garnett could have potentially won five titles in Duncan’s place” are not a substitute for real production.
C.) When I say “accomplishments”, I am referring to winning, winning more than expected, or being a top player. A scoring/rebounding/assisting title is a literal accomplishment but not one which matters much to me.
D.) I generally do not care too much about title counting, but there is a ceiling to not winning more. Everyone in my top fifteen has a title. And it would take some extraordinary circumstances to make it in there without winning, and similarly extraordinary circumstances to make it into my top eight with only one title. A true outlier like Russell does secure him from falling far though, and if we were only looking at the “best” thirteen years, I may never be able to move him out of the #1 spot.
E.) I am more interested in succeeding with different variables than I am on repeated success with an established core. I also do not care much about timing; 1999-03 Shaq is not more valuable to me than 2003-07 Duncan just because one was LWWWL and the other was WLWLW, and if anything, I am inclined to the larger roster variation we see with Duncan.
F.) Era context matters to me. If I feel you were specifically benefitting from or being limited by certain era standards, I will not just be taking your in-era status at immediate face value.
G.) Size matters. Being a good defensive guard is not as valuable as being a good defensive forward, and being a good defensive forward is not worth as much as being a good defensive big… but as we have seen with Jokic, it is much more possible for the best offensive forwards and bigs to replicate the offensive production of guards.
H.) Efficiency needs to be contextualised to team role and scoring load. This tends to be a mistake with people who mostly are looking at raw boxscore averages, but I see it with people watching games too. Yes, there are players who “chuck”, and yes, that can be frustrating, and yes, it would be better if those players could score more with less, but it is not feasible to funnel shots to roleplayers. There is a burden to shotmaking.
Terry Pluto wrote:One of my first conversations with [Lloyd/World B.] Free led to him asking, “Do you know how hard it is to get 20?”
“Twenty points?” I asked.
“No, 20 shots a game,” he said.
I laughed.
Free was serious. He explained how defenses were set up to stop players like him, scorers on bad teams. The goal was to keep the ball out of his hands. And when he did have the ball, he often faced two defenders. It took strength, energy and ingenuity to get off 20 decent shots a game.
“It can wear you down knowing you have to carry the offense for your team,” he said. “But I did it, year after year.”
Not everything is about “ceiling raising”.
H.) Relatedly, while I generally support APM use as a guideline, role and team context is an essential consideration. Stockton was not a secret superstar anymore than Kyle Lowry has been.
I.) Perhaps tying back to the idea of “accomplishments”: there is a value cap to players who never exceed all-NBA/all-star status. NBA history strongly indicates that having a top tier superstar correlates best with sustained success. Stockton did not drive winning teams more than Patrick Ewing did, and it has bothered me to see that so readily abandoned just because the Riley Knicks maintained fair frontcourt depth behind Ewing. Perhaps we can see this as an encapsulation of a lot of these points: excessively penalising “inefficiency”, irrationally inflating the value of guard defence, improperly failing to properly appreciate big man defence, etc.
J.) Another spot where I think APM has often been misused: seasonal variance and age curves do not result in these massive impact fluctuations or drops the way APM metrics may suggest. Garnett did not go from one of the most valuable regular seasons ever to having more neutral impact. Although I think box score comparisons are reductive and too often skewed to specific archetypes,
internally, this is where they are useful for stabilising what “impact” could suggest. If a player’s production holds firm, that should clue you in to there being an alternative explanation.
K.) On the subject of aging curves: the best players tend to hold onto more their value than what “impact” and production dips may otherwise indicate. There inevitably is a time where that stops being true, but people seem weirdly eager to get ahead of themselves on those declines. Again often a case of vaguely gesturing at minor declines and overselling their real effect on a team.
More may come to mind as we go deeper into the project, but that should be good place to start.