APBRmetrics

Thanks for the new link. Looks like your article generated lots of comments there.

Also: comparing career survival rates in the 1970s to the 2000s is skewed by the ABA.

untrue - the attrition rate in the nba from 76-77 to 85-86, the decade following the demise of the ABA, was 28 year olds being 52% of 23 year olds in the league, and 30 years olds being 34% of 23 year olds, almost identical to the rates for the nba in the decade of the 1970s, and nowhere near the rates of the 2000s...

That still leaves the question of why didn't some other team pick up Slick Watts... Slick Watts was, for most teams, not a very good player...

this idea that the only reason a player was out of the league at a young age (young compared to the players of today) after having been a starter being because he was not a very good player is again, simply untrue...

1st off nba teams back then employed fewer players over the span of a season than they do now. in 74-75 the average team had only 14 players play during the season, in 75-76 just 15. nowadays it's 17-19, even with guaranteed (or longer) contracts...

2nd back then not all players were on one year contracts, so it wasn't like teams could simply shuffle players in and out all season long. if they dropped a player under contract, they still had to pay him, as well as the player picked up. the league was nowhere near as profitable then as now, and teams had to watch their costs. odds are a veteran of 4-5 years would cost them more than a 23 year old, and back then the league did not subsidize minimum contracts for older veteran players like they do now...

3rd there were a number of players that were starters in the 1970s that were then out of the league just a season later. were they out too because "they were not very good"?...

ken charles was a starting guard for two teams, the buffalo braves (age 24) in 74-75 and atlanta hawks (age 25) in 75-76. those 2 seasons he played close to 4000 total minutes. but he was out of the league at age 26 (waived in 12/77 by the hawks)...

cornell warner was a starting PF/C for the milwaukee bucks (age 26) in 74-75 and los angeles lakers (age 27) in 75-76. he played over 5000 total minutes those 2 years, but he was out of the league at age 28 (released by the lakers in 76-77)...

bob kauffman was a starting PF/C for the buffalo braves from 70-71 to 72-73, averaging 39 min/g, 19 pts/g, and 11 reb/g (age 24-26). but he was out of the league by age 28 (waived by the hawks 10/75)...

e.c. coleman was a starter for new orleans in 76-77 (31 min/g) and part time starter for golden state in 77-78 (25 min/g). he was all-D 1st team for the jazz and all-D 2nd team for the warriors. the next season he was out of the league, by age 28...

were all of these players, 3 of which started for not 1 but 2 teams, out of the league only because they were "not very good"?...

Yet we have powerful evidence that teams could find no place for him...

true, but not being very good is not the only reason teams did not pick up 27-28 year old players back then...

VJL wrote:

How a player translates his college play to the NBA is controlled by his environment. Have you considered adding a measure for player development? For example, getting drafted to the Spurs probably helps you reach your potential better than, say, the Bobcats. If there's one team who could lead to their own variable, it's the Spurs.

That could potentially be a cool addition. I'm not sure what the best way to operationalize "environmental goodness" is. Team record in previous year? It would be especially difficult, because I wager most of the "environmental" importance is idiosyncratic. What is a great situation for one guy may be a terrible one for another depending on how skills mesh with needs.

Here's essentially what I meant:
1) The Spurs are really good at finding draft steals, so maybe it's beyond that and they know how to develop players. Thus, add a simple variable for landing on the Spurs to boost your draft rating by "X" amount. I'm not saying it's perfect, but it would be interesting to look into that.

2) I suppose you could go on a complicated route and try for weird fit adjustments like having quality players at the same position in the rotation. Or something else.

AcrossTheCourt wrote:

VJL wrote:

How a player translates his college play to the NBA is controlled by his environment. Have you considered adding a measure for player development? For example, getting drafted to the Spurs probably helps you reach your potential better than, say, the Bobcats. If there's one team who could lead to their own variable, it's the Spurs.

That could potentially be a cool addition. I'm not sure what the best way to operationalize "environmental goodness" is. Team record in previous year? It would be especially difficult, because I wager most of the "environmental" importance is idiosyncratic. What is a great situation for one guy may be a terrible one for another depending on how skills mesh with needs.

Here's essentially what I meant:
1) The Spurs are really good at finding draft steals, so maybe it's beyond that and they know how to develop players. Thus, add a simple variable for landing on the Spurs to boost your draft rating by "X" amount. I'm not saying it's perfect, but it would be interesting to look into that.

2) I suppose you could go on a complicated route and try for weird fit adjustments like having quality players at the same position in the rotation. Or something else.

Any of those options would rapidly approach over-fitting; in fact, I think that if you did out-of-sample validation, you would almost certainly regress "team effects" all the way to zero.

But shouldn't that be tested before we toss it out? One of the common headlines about the Spurs and the draft is that they don't just find good players late in the round; they develop them better than anyone else. This thought is thrown around a lot in NBA articles ever since the Spurs blew through Memphis. So it would be useful to say, "Hey, we have all this data and couldn't find that effect at all, so it's probably not true."

One of the common headlines about the Spurs and the draft is that they don't just find good players late in the round; they develop them better than anyone else. This thought is thrown around a lot in NBA articles ever since the Spurs blew through Memphis. So it would be useful to say, "Hey, we have all this data and couldn't find that effect at all, so it's probably not true."

It would be pretty easy to see whether players going to particular teams tend to overperform and would be a fun thing to do when I get some free time. That said, I doubt the Spurs would stand out if I ran the numbers. For all the talk about "developing" it looks to me more like they are really good at identifying underrated talent. My models really like some of their "surprise" players. Leonard is pegged as the 3rd best in his class, Green as the 9th in his, Hill was the 12th in his... even more interesting to me is that some of their less successful attempts to find a gem agree with my model so they may actually be using a similar approach. Blair was viewed as a potential stud by any purely stats approach I have seen. James Anderson was rated 14th by my model and I have seen him higher using other similar stats models. The same can be said for Denmon (though for some reason he isn't in my dataset). Gary Neal is really the only recent Spurs contributor who wasn't a great value pick out of college by the numbers, but I wouldn't be surprised if his Euro numbers tell a different story.

Just to reiterate my point of view: I don't necessarily believe in the "Spurs turn people into good players." When people take that to the extreme, I just think that's silly: the guys they find clearly have NBA skills and bodies and it wasn't from a few weeks of training camp on the Spurs. But it's good to investigate the question and have some numbers to backup a claim.

VJL wrote:
It would be pretty easy to see whether players going to particular teams tend to overperform and would be a fun thing to do when I get some free time. That said, I doubt the Spurs would stand out if I ran the numbers. For all the talk about "developing" it looks to me more like they are really good at identifying underrated talent. My models really like some of their "surprise" players. Leonard is pegged as the 3rd best in his class, Green as the 9th in his, Hill was the 12th in his... even more interesting to me is that some of their less successful attempts to find a gem agree with my model so they may actually be using a similar approach. Blair was viewed as a potential stud by any purely stats approach I have seen. James Anderson was rated 14th by my model and I have seen him higher using other similar stats models. The same can be said for Denmon (though for some reason he isn't in my dataset). Gary Neal is really the only recent Spurs contributor who wasn't a great value pick out of college by the numbers, but I wouldn't be surprised if his Euro numbers tell a different story.

The Spurs obviously take into account college production more/better than almost any other teams - so their ability to "develop" players isn't surprising to me. They often simply draft better "developed" players from the get go. Blair in the 2nd round is an obvious example.

Let's just say they won't be drafting Shabazz Muhammad or Phil Pressey - they will probably nab an Erick Green/Mike Muscala/Pierre Jackson/Nate Wolters in the 2nd round though.

VJL wrote:

What combine stats did you find have a positive correlation to success?

When I was including them alongside the boxscore statistics, no step vertical and standing reach were the only predictors worth including.

Now that I tried applying combine stats post-hoc for the newer data (including it in a regression along with the predictions posted in the parent post such that -- [Observed Wins ~ Predicted Wins + No step vert + Standing reach] I found some interesting things:

1) Standing reach is now a negative. What that tells me is that while standing reach is obviously important to playing basketball, if you can't use it to record rebounds/blocks/points... in college, you won't figure out how to use it in the pros and some of the stats you got just by being bigger than anyone else will be tougher to come by.
2) No step vert is still informative, but the effect was really weak until I included an interaction term between predicted wins and no step vert. Basically, if you don't look like a good prospect based on the box score stats it doesn't matter how high you can jump. However, as your expected production increases, being able to jump out the building becomes an increasingly important predictor of your ceiling.

Kudos on using standing reach. I just learned that Kevin Pelton didn't even use height (huh?) in his draft ratings.