Career Trajectories of Early Entry G-F, Part 1

All of the debate and discussion about the value of the young talent on the Blazers and how much the Blazers should get back in any trade of its young assets raises a very interesting set of questions: how much do players improve, on average from year to year? Do they improve quickly early in their career and then level off or is it more gradual assent? How many years does it take for a player to plateau? Or, more specifically, how good are Webster and Outlaw going to be in a couple of years? Are there many players at their position that go from their level of production during their third years in the league to being NBA All-stars? What is the chance that they regress? Certainly there are many players that take a step backwared in their fourth or fifth seniors, like Leandro Barbosa,  Luke Ridnour, Kirk Henrich, Ben Gordon, and Juwan Howard--and that's not counting players that sustain a significant injury, like Darius Miles. 

Obviously, a lot depends on the individual player. Players that are dedicated to their games improve more than those that are not (duh!). On the other hand, raw players have, by definition, more room to improve (see Ridnour). There may also be differences in player improvement by the age a player enters the league, by position, by NBA era, by skill set, and by team environment. On the other hand, not using the history of other players to inform one's projections about current players just because they are not identical seems unwise to me.

Thanks to basketballdatabase.com it is now possible to address these questions quantitatively. Data on every player that played from 1950 to 2006 is freely available. So, having a little free time, I downloaded the data and crunched the numbers. In order to focus my analysis, I selected a sample of players that I consider a reasonable comparison group to Webster and Outlaw, while leaving in enough players to have some "statistical" power. I know most BEs aren't really interested in confidence intervals or the finer points of statistics and probability, but it should be pretty obvious that if, in the last 20 years, there is only one player of Webster's height, weight, position, age of entry, and first year stats, we probably aren't going to learn much by just comparing him to that person. It turns out that there are relatively few players (36) that have entered the league at age 19 or younger and played for five seasons by 2006-07. However, I think this is an important enough issue that, in general, I went with smaller samples with players more Webster-like and Outlaw-like players rather than larger samples with many players that were quite different from Webster and Outlaw.

There are many different ways to try to answer the questions posed above, and I've used three different approaches (two shown in this post). The first is relatively simple: I selected all players that entered the league at the age of 19 or younger and were between 6 ft and 6ft 9in, started their careers after 1980, and had at least 5 years of experience. With this sample, I calculated  the average production by year of experience (1st, 2nd, 3rd, 4th, and 5th). That is, the table below shows how many points, rebounds, assists, etc that the average early entry guards and forwards produced per 36 minutes in their 1st, 2nd, 3rd, 4th, and 5th seasons. In this analysis, it is important to restrict the sample to players with 5 years of experience because if players leave the sample after their second year, then the averages calculated for the 3rd, 4th, and 5th years will be  from a different group of players than those calculated for years 1 and 2. Scaling the "stats" to "stats" per 36 minutes is also important since playing time can vary greatly depending on team depth, coach, etc. Here are the results:

Table 1

What do they say? Probably depends one what your expectations were. As much as anything, rookies look surprisingly productive, they just do not have the trust of their coaches to get minutes. Second, some parts of the game have a different learning curve than others. Points, FT% and 3P% show a pretty steady improvement. Steals, Turnovers, and FG% surprisingly, bounce around. And Rebound productivity, if anything, declines. Secondly, there is still a considerable amount of improvement to be had for players that were 21 in their third year in the NBA. On the other hand, by these numbers it would appear unlikely that a player averaging 13 points per 36 minutes in their third year, like Webster or Outlaw, will become a scoring champion by their fifth year... even though scoring is one of the areas where players improve the most. While it's true that Webster and Outlaw got slower starts than the average player in this sample, they really weren't wildly different by year 2:

http://www.basketball-reference.com/players/w/webstma02.html

http://www.basketball-reference.com/players/o/outlatr01.html

In case you are wondering, here is the career trajectory for a similar sample of players that played at least 8 seasons (though the sample only includes 12 players):

Table 2

They show the same general pattern. Note, however, that these players plateau in their 6th season from a statistical stand-point.

Now there are a lot of reasons to question the validity of any inferences using this data. A fairly obvious problem is that both samples were quite small. An alternative approach would be to include players that started later in their careers and look at productivity by age. By including players that started when they were 21 or younger, we loose the unique experience of players that entered the NBA as teenagers, but we get more players (91) and it is still interesting. So, the next table shows the progression of the same type of players (6ft to 6ft 9in, started careers by 1980, etc):

Table 3

This pattern reinforces, more or less, the observations made above. Players definitely improve, on average, over time, but expecting dramatic improvement is probably questionable, because some players get worse and some players do not improve a great deal after the age of 22 in terms of rare productivity.

Conclusion:

In my first cut of looking at the experience of other players "like" Webster and "like" Outlaw, it's clear that players tend to improve after their 3rd seasons, but that they do not tend to improve enough for either Webster or Outlaw to turn into all-star caliber players. Strictly speaking these figures do not provide a way to project Webster or Outlaws future, but if we were to do so, any where between 12 and 17 points per game would seem reasonable. Therefore, in my view, those that expect a lot of improvement from Outlaw and Webster and those that don't expect much, both have plausible arguments. One group is being optimistic, another pessimistic.    

In my next post, I will use some slightly more advanced analytical tools to make projections for Outlaw and Webster's production in future seasons. These methods (OLS regression) will provide me with a way to try to address the fact that both Webster and Outlaw were relatively "raw" players when they entered the league. The types of conclusions I have drawn thus far, assumes that the amount of improvement does not depend much on one's initial productivity, which might turn out to be false. Should be interesting.

Thoughts?

Footnote:

A statistical caveat: the amount of improvement suggested by tables 1 through 3 is biased upward to some degree. Since I dropped all those players that fell out of the league entirely and were unable to play, the players in the sample are all, in a sense, successful players-it doesn't include the washouts. On the other hand, given that I expect Martell and Outlaw to last well beyond their fifth seasons and older players are as likely, if not more likely, to have a career ending injury, I think it's reasonable to restrict the sample in this way.