I did more math guys!
I wanted to see if I could predict the odds that the Mavs' young players get better going forward. My inclination was to try and find a pattern (a regression) in all the young player's efficiency ratings so far, and then see what that would tell me about the future. This did not work; to the extent that there was a pattern, I wasn't going to get anything concrete from it. So I tried something else, and it worked.
It occurred to me that I could use basketball reference to find all the players who had PERs similar to the Mavs' young talent at the same times in their careers, then see how all of the non-Mavs players progressed throughout the rest of their careers. Once I had the career PERs of all the similar players, I made a normal distribution of the data that would tell me the odds of the Mavs' players getting better, worse, or staying the same.
The hope in finding players who had similar PERs is not necessarily to find similar players, but to find players who were in similar career situations and see what happened to them as their career went on. For example, Mayo and Shane Battier had the same PER for both their 3rd and 5th seasons in the NBA, but Mayo and Battier are not really similar players. They did, however, both have similar value, minutes, and roles (to a certain extent) for their team for those first few years of their career. In constructing the distribution, I'm really looking to see how players moved on from the same points in their careers. So, what's the likelihood of Mayo or Collison moving on in any particular way?
The mean of Mayo's distribution was 13.7748, with a standard deviation of about 3; meaning we can most likely expect Mayo's PER to average out to be somewhere around 13.8 for the rest of his career, while there's a 66.7% chance of him averaging a PER of anywhere between 10.8-16.8. Here's what that distribution looks like:
Probabilities based off of this distribution:
Likelihood that Mayo gets better (has a PER greater than 14.6): 37.83%
Likelihood that Mayo gets worse (has a PER below 14.6): 60.64%
Likelihood that Mayo stays about the same (PER between 14 and 15): 15.51%
Likelihood that Mayo gets significantly better (PER greater than 17): 15.62%
Likelihood that Mayo gets significantly worse (PER below 13): 39.74%
-This does not add up to 100% and is not supposed to; note the overlap between different scenarios.
Remember that this is raw "here's what happened to other guys" result data. So it's worth noting that we watch Mayo, and we know that he has way more talent than his current 14.6 PER indicates, but that he just isn't tapping into. A lot of the players he got compared to really didn't have that talent at all (erhrm, Lionel Hollins). So there is an extent to which his odds of improving are probably better than this distribution would have you believe. But how much better? Probably not as much as we'd really like to hope.
There is some encouraging data there though: The odds that he gets way better are about the same as the odds that he doesn't change at all, for example. I have to admit, I see OJ as a guy who can improve, but I don't know that I see him as an 18-PER-with-significant-minutes kind of guy; but a 15% chance is really nothing to shake a stick at. Add a couple percentage points to all the good probabilities and take a few away from the bad ones, for faith's sake, and the picture is...
Well, still not that good. But it's not awful either. 40% chance of improvement? Less than 50-50, but it could be worse.
Last fun note: going through all the similar PERs, the player who stood out as the most similar to Mayo over the first 5 years of their career was Michael Finley. I don't think there's too much value in semi-arbitrary player comparisons, but it's hard not to see a few similarities between the two. I think the most important similarity, though, is that both players learned how to shine on bad teams. Finley carried the Mavs when there was no one else to do it, and the same with Mayo pre-Dirk-return. The difference is that Finley learned how to play with a good roster. So the question ends up being the same as it was when we started this, "can Mayo learn too?"
I was immediately surprised by how high Collison's PER for this season is: 16.4. I mean, that's only slightly above average, but he doesn't really strike me as an above average asset for this season. PER does a really poor job of taking defense into account though, and that's his EXTREME weak point, so maybe that's it. At any rate, the mean for his distribution was surprisingly low given his PERs up to this point: only 14.6068, with a standard deviation of 3.5. So, again, this means that he's most likely to average a PER of 14.6 over his career, with a 66.7% chance of averaging between 11.1-18.1. Here's what his looks like:
Likelihood that DC gets better (has a PER greater than 16.4): 29.46%
Likelihood that DC gets worse (has a PER lower than 16.4): 68.79%
Likelihood that DC stays about the same (has a PER between 16-17): 11.1%
Likelihood that DC gets better than how I feel like he's been playing (PER greater than 14.7): 48.8%
Likelihood that DC gets worse than how I feel like he's been playing (PER below 14.5): 48.8%
Likelihood that that DC plays up to about how I feel he's been playing (PER between 14-15): 1%
Likelihood that DC gets significantly better (has a PER above 17.5): 20.33%
Likelihood that DC gets significantly worse (has a PER below 13.5): 37.45%
The odds that DC averages a PER between 14 and 15 is only 1%?! Crazy.
I had to wonder why the mean PER was so low, given that Collison has had a PER of 16+ in every season of his career so far other than a dip last season. The answer was a part of the trends I was seeing with players like Collison: most of them were pretty good in the first seven or so years of their career then tanked early, averaging PERs of 12 or lower for most of their career after their early success. Unlike Mayo, I didn't find any one player who really stood out as having similar PERs over every year, but I did see a LOT of people who had this trend of early decency then terrible early decline. Again, because the players I'm comparing Collison to are not so much similar as they had similar early careers, there's really no way to know if Collison will follow this trend, but it does affect the curve, and it certainly isn't super encouraging for the long term. On the other hand, Steve Nash actually appeared in the comparison twice, so you never know (and yes, I know the two aren't similar at all, it's just supposed to be a little bit of hope, that's all).
I think it's fascinating that DC has about a 50-50 split between being an above average versus below average player for the rest of his career. There's not really too much to be said about that other than "wow." I mean, more than anyone on our squad, the decision to develop or ditch Collison is almost purely a roll of the dice. Though, his defense is horrendous enough that maybe that's the tipping point; maybe that's what knocks him into "not worth it" territory.
I couldn't run a distribution for Brandan Wright: 90% of the players who came up in comparisons played for less than 5 seasons or are too young to give much data. But some of the names that came up are incredibly encouraging: Carlos Boozer, Chris Webber, Antonio McDyess. The Boozer comparison is especially nice; the two had the same PER (21.6) in their 4th seasons, and I think the two of them have a similar enough game to feel really good about that particular comp. Other than that, not much to be said about B Wright.
I also couldn't run one for Jae Crowder, simply because he hasn't had enough time in the league for me to get any semblance of a reasonable sample.
So what do we take from this? The odds are interesting, but they're just that: odds. So there's about a 60:40 split of Mayo getting worse:better; so we can expect him to get worse, but just barely. And like I said, we know that he has more talent than he's showing, so what is that worth? How does that change the numbers? What we're looking at is interesting, but all it is is what happened before now. Anything can happen in the future.
So the question is: Do you play the odds?