I have been working on a project revolving around the NBA draft for quite a while, at this point is has been longer than a year. At some point I want to create a website, and this would be the basis for much of the web sites content, but I would have other things as well. I am not ready to do a website at this point, so I thought I would publish some of what I have done here.
My concept is pretty simple. For years I have read a lot about the players coming into the NBA draft, and one thing you see with the evaluation of incoming collegiate and Euro players is an attempt to find comparable players in the NBA to the potential draftee. There are lots of attempts to describe players relative to one or more ?comps?, and you see this much more with basketball players than with football or baseball players. The process of determining ?Comps? though has always seemed to me to be a very subjective measurement, and some players seem to have multiple ?Comps? that are pretty diverse. I decided to take on a project to make this ?Comp? determination process far more objective, and the project has grown and grown and grown in the last 15 or so months I have been working on it.
I will try and express what I am doing metaphorically.
1.) My son described it his way, ?Think of a player as a color, say blue. Now exactly what shade of blue is he? Player A is Sky Blue, Player B is Royal Blue, Player C is Navy Blue. This draftee is a little of all three of those shades of blue, but not actually any one of those 3 shades?.
2.) My son described it another way, ?There are no true synonyms, no two words mean exactly the same thing. There are a lot of words that are similar, some very similar, and you can describe something pretty effectively using a multitude of synonyms, but it will not be exact. With a player you can use ?Comps? to pretty effectively describe a player, but not exactly?.
3.) A third way to express my effort is to think of the universe of NBA players as a 3 dimensional cube. Each and every player will have some specific place inside that cube that perfectly describes them. In a 3 dimensional circle around that player will reside other players, and those players can best be described as his ?Best Comparables?.
The way I have come to determine a player?s ?Comps?, is by developing a very rigorous process of numerically describing that target player, as holistically as possible. Now I don?t have access to some data, so I used whatever data I could find about the population of NBA players and potential NBA players.
The basic components of this are;
1.) The physical attributes of the player, as measured at the NBA combine, basically height, weight, wingspan, reach etc.
2.) The NBA combine also does testing on a players ?raw athleticism?, by measuring vertical leap, speed, and agility, and some others.
3.) I studied the relationship between some of these various raw measurements, and found the relationship between them, and from that I can place a player?s physical attributes and athleticism into a context. For instance, Cody Zeller?s vertical leap was the highest of any player over 6-9 ever at the combine. That is interesting, but exactly how much of a better leaper is he than you would expect for a player of his height? I have calculated what I believe is a fair representation of that.
4.) A player?s college, or European League stats.
5.) The player?s age.
6.) The player?s competition level.
In total I have found 123 different discreet descriptive measurements of a player for his seasons prior to his NBA seasons (and in some European player?s data on seasons after they played in the NBA). I found that data (or at least as much of that data as I could find) on almost every player that has played in the NBA since 1990, as well as a number of other players who played prior to 1990. In total my database includes over 4,000 player seasons.
For each of these player seasons, I found the relationship between each player season and every other player season. These relationships are expressed as a % of similarity between one player and each and every other player. For instance you have a target player, and you compare him to 2 other players. The target player averages 10 points per game, and player A averages 12, and Player B averages 8. In both of these cases (Player A to the Target Player and Player B to the Target Player) the degree of similarity of PPG is 80% similar. The actual math is more complex, and not exactly like this, but I don?t want to get caught up in the minute?, so this is in general the way the process works. I find the relationship of a target player to all other players for all 123 measurements, and then develop a composite degree of similarity for each target player.
The composite measurement of similarity then defines a player?s ?Comps?. From this group of comps, I find the specific relationship between the target players and each of the comps, based upon their relative collegiate or Euro seasons, and use that in conjunction with the ?Comps? actual NBA stats to calculate a predicted NBA stat line for the target player for his first 3 season in the NBA.
I tested the methodology on the 2012 draft class, and some selected players from the 2011 draft class. Overall the predicted stat lines have been pretty accurate, though obviously not perfect. I would say that they have been directionally very accurate, highly accurate for some categories, good for other, all the way down to marginally/reasonably accurate on a few stat categories. Some players were extremely accurate, other players very accurate, some accurate, and a few would be described as marginally accurate. The marginally accurate were generally due to very low minutes played. I have had a number of iterations to this predicted stat line, and I have continually improved the tool, but it is certainly a work in progress, and I have a number of ideas to make it more accurate.
So as I said above, I am going to publish my list of best ?Comps? for a number of this year?s top 100 prospects, and give some other data about the various potential draft picks, and my 3 year NBA stat projection, and then eventually give my ranking of the players as I see them using this methodology.
One last thing, the data I used to develop this tool was freely available on the web, and I found and used data from a number of sites. I do want to acknowledge Draft Express and Basketball Reference, as the two main sources, and also Dean Oliver?s book Basketball on Paper. Derek, please pass along my sincere appreciation to Jonathan Givoney for all that he has done with Draft Express.
I hope you find this interesting, it has certainly been a lot of fun to do.
