Posted by jeffdrayer on 10/23/2012 1:41:00 PM (view original):
Hey -- I don't have any horse in this race, and am well aware that assists are window-dressing. And really, I'm just trying to avoid doing work here by commenting, but Billy -- your theoretical example above wherein a PG's assists go from 3.5 to 4 because a big man's LP goes up 20: I'm not sure I understand. True, the PG didn't "make the team better" even though his assists went up. But it's like saying Stockton wouldn't have gotten 12 assists per game if the Jazz didn't have Malone -- no one thinks Stockton's theoretical jump from what probably would've been 8-9 assists without Malone is due to Stockton making the team 3 assists better. And, if I were the Jazz PG instead of Stockton, I would've gotten (say) 6 assists per game even with Malone. So the real comparison is, in a world with Karl Malone, how much better did Stockton make the Jazz than jeffdrayer would've made them, and the answer would be 6 assists. So even though assists are window-dressing -- given the same HD team, but with one having a 99 spd/passing PG and the other with a 40 spd/passing PG, both set at the same distribution: the first one will have more assists. And so we learn that he was more valuable to his team. After all, assists aren't handed out randomly -- they're handed out by position (PG > C), and within that position, by a player's abilities (99 passer gets more assists than 40 passer).
But maybe I'm not understanding your argument?
its true, there is a rough correlation with assists. what im saying is its burried under so many other variables, much more dominant than "the pg impact on team fg%". what you are saying makes sense, i am saying the same thing. when stockton had malone on his team, compared to not, his assists went up. to know what the actual impact HE had was, you would someone have to abstract away the quality of the team - the lp/per, ath/spd of the rest of the players. take a PG on my team, put him on a ****** team, and his assists might double - but his intrinsic value of a player, from an objective standpoint, is the same. the amount of assists he gets or does not get really doesn't tell you much about why your fg% is where it is - the rest of the players on the team have a WAY bigger impact (in HD, that is one of the respects where HD really differs from real life - easy to believe, i think, considering it was a non factor previously).
all im saying is this - you are saying, stockton got 12 apg because he had malone. what if another pg was there? doesnt even have to be from the same time. would rondo have gotten more or less than 12 apg? who the hell knows. they have different teams and you just can't say. what im saying is this - burried under a mountain of variables, somewhere, assists might, and i would guess probably are, somewhat linked to how valuable a player is to your team in terms of improving fg%. but the other factors - what position they play - how good are the rest of your guys at improving the fg% of the team - how good are the rest of your scorers - these things are all really big factors.
and dahs - i use rebounding as an example because i think its more easy to relate to. everyone looks at rebounds, they are meaningful. but you really have to be careful - quality of other rebounders, quality of the opposing team, the fg% of your own team - these things have a major impact on how many rebounds and individual gets. plus, position, and a bunch of other factors. those factors, all summed up, are pretty damn significant - a high reb player might pull 6 on a championship team and 11 on a crap team. but you still know hes good because on that team, the next best guy pulled 5.5, and on the other best teams, those guys pulled less than 6 in general. its not a GREAT way to conclude what makes a rebounder good - try to put an exact value on ath to reb that way, its a really tough thing, 99% or more of coaches would never get there. but still, the reb figure is significant. team rebounding honestly holds more promise if you were really to track it, but nobody does, and thats ok.
assists are like rebounding, except burried under an additional mountain of stuff. the passing abilities etc of other team mates affect assist numbers even more than in rebounding. the quality of your offense has a WAY, WAY bigger impact. tempo has about an equal impact. quality of opponents might be the one thing with less of an impact. position? way bigger impact. scoring of that player? GIGANTICALLY bigger imapct. all in all, the real meaning of assists is so burried that its been a forum fact as long as ive played that assists are artificial window dressing. and sure, maybe the new engine changed that (we know for sure it was wnidow dressing 100% before). but now, maybe there is 10% of realness burried 90% of other factors. MAYBE, over a 1 year study, i could figure out that correlation. maybe. now maybe you are a lot better at figuring this **** out, but if so, i challenge you to prove it. what makes a player more valuable to a team's fg%? hell, run a simple regression, like most of them, its ****** and limited for a variety of reasons, but it gives you something. show me the stat correlation to team fg%, conclude which stats matter, run it again with those weights. then, show me assist correlation. if those graphs line up, ill give you this one (or, maybe ill argue why your approach is too flawed to interpret the results, depends how you play it). i know you are into that regression stuff, but i like to stay away from it. half because it ruins my fun, half because i think i have *just* enough insight to maybe make those regressions meaningful - and it just feels like cheating to me if you can really prove something via regression (ive yet to see anyone in HD do it). but by no means should that stop anyone else. i think in this case, you could actually glean something. regressions are always good for getting a ballpark idea, and really bad for really nailing it down, because of the interdependence of the ratings everyone cares about so much, among many other things. but take something like passing, that is pretty independently generated.
anyway, my feeling - and its a strong one - is your graphs wouldn't be close. take the original example, big man playing pg. he can have terrible pass, bh, and spd, and he will grab a number of assists. very possibly more than my elite scoring pg does on my team. of course, you cant compare assists across teams like that. and because of positions, you cant compare within a team either. so my question is, what the hell can you compare to? and what does that tell you? even if you just compare your own pg on your own teams which run the same offense and defense and build the same basic way, you still have significant variance in a number of things: 1) the passing abilities etc of teammates, 2) the scoring ability of teammates, 3) the scoring ability of the player itself, 4) the distros you set up, 5) the tempo you run. you can limit some of those but to get anywhere close to a scenario where you can isolate the variable, to me, thats unfeasible. and true, you never need to get all the way there. ive argued your side, you can deduce things like this from different stats for a long time, with most coaches against me, but i just feel assists are past that line.
10/23/2012 2:37 PM (edited)