Follow Us ON

© 2017 Broom on the Warpath

Advertise Here Flag

Wizards Similarity Scores: Young Guns Edition


At this point in major sports history, there are few "original" players. By that, I mean that players in today's game tend to perform in significant ways like players that preceded them. Careers often follow a familiar trajectory to similar players that came before -- which is one of the big reasons fans and analysts often talk about who a player reminds them of.

So far as I know, Bill James (the Godfather of Sports Analytics) was the first to develop "similarity scores," in which he compared players using their statistics. Others followed suit in baseball -- tweaking the approach to their own liking. Several good analysts have generated similarity scores for NBA players, including APBRmetrics' Mike Goodman, Kevin Pelton, and the good folks at Basketball Reference.

As a card carrying member of the stat goober club, I'm (for the first time) taking a stab at generating my own similarity scores. My approach has a few advantages. First, I actually understand how they're being created. And second, I don't have to bug Mike and Kevin for their results.

I won't go too far into the weeds on the method because I'd prefer you continue reading. My basic approach is...umm...similar to the others -- I measure the difference between Player A's performance in season X and every other player season with 500 or more minutes since 1977-78. 

In my similarity scores, I use per minute box score statistics that have been standardized to a pace of 100 possessions per 48 minutes. The categories used are: age, minutes, field goal attempts, three point attempts, free throw attempts, offensive rebounds, defensive rebounds, assists, steals, blocks, turnovers, personal fouls, points and PPA. 

Regular readers know that PPA stands for Player Production Average, which is an overall rating stat I developed. It's a pace-adjusted, per minute stat that incorporates defense and a "degree of difficulty" factor based on the level of competition a player faces while on the court. In PPA, 100 = average, higher is better and 45 = replacement level.

Okay, enough about that. Let's get to the scores. For today, I want to focus on the Wizards young guns -- John Wall and Bradley Beal. The team is hoping they'll form a dynamic back court that will lead the team successfully for the next decade.

Here are the players my system says are most similar to what Wall and Beal produced this season. Note that the "base" player (Wall in the first table and Beal in the second) has a score of 100. The "similars" are stacked below. The closer to 100 a player is, the more similar. Also note that I've included the player's PPA for that season as well as his peak PPA and his age when he peaked.

John Wall  PG  2012-13  22  WAS  100 139 139 22
Russell Westbrook PG  2010-11 22 OKC 87 167  177  24 
Stephon Marbury PG 1999-00  22  NJN  87  135 164  27 
Tony Parker PG 2005-06  23  SAS  86  158 187  30 
Tony Parker PG 2007-08  25  SAS  86  145 187  30 
Sam Cassell PG 1997-98  28  NJN  86  132  178  34 
Tony Parker PG 2004-05  22  SAS  85  136  187  30
Ray Williams PG 1979-80  25  NYK 84  145  145  25 
Tony Parker PG  2006-07  24  SAS  84  173 187  30 
Russell Westbrook PG  2009-10  21  OKC  84  123 177  24 
Reggie Theus PG  1982-83  25  CHI  83  122  132  28 

Kinda interesting that the similarity system produced so many PGs without including position or size as variables.

Also interesting to see that a) Wall's similars are/were high-quality players for the most part, and b) virtually all peaked at least a couple years later. Cassell's name is intriguing. At age 28, it was his first above-average season, and it came at a time when the typical NBA player is maintaining whatever he's become before beginning to decline in his early 30s. Cassell was just coming into his own, and wouldn't reach his peak until age 34. 

And, Beal:

Bradley Beal  SG  2012-13  19  WAS  100  92  92  19 
Ray Allen SG  1996-97  21  MIL  92  100  197  25 
Mike Miller SF  2001-02  21  ORL  91  97  140  22 
Jason Richardson SG  2002-03  22  GSW  90  91  141  25 
Jason Richardson SG  2001-02  21  GSW  88  84 141  25 
Dennis Scott SF  1992-93  24  ORL  88  89  141  27 
Thaddeus Young PF  2009-10  21  PHI  88  91  153  24 
Brandon Knight PG  2012-13  21  DET  87  64  64  21 
Cuttino Mobley SG  1999-00  24  HOU  87  90  114  25 
Kyle Korver SF 2006-07 25 PHI 87 79  129  31 
Metta World Peace SF 1999-00 20  CHI 87  91  143  27 

Seeing Ray Allen's rookie season show up as most similar to Beal's is highly encouraging. Allen is one of the best shooter's in league history and that peak PPA of 197 is All-NBA worthy. I'm also encouraged to a few players 5-6 years older show up in comparison. That Beal at age 19 performed at a level close players with several years of NBA experience bodes well for his capacity to improve. I also liked seeing the number of forwards showing up -- indicative that Beal plays "bigger" than his measured size and his position.

Beal's similars are a bit more hit and miss than Wall's, which is exactly what we should expect. While Beal improved radically during the season, his overall performance was a bit below average. What we know is that he has the potential to become a great player. But, if he doesn't put in the work to improve his body and his game, he won't fulfill that potential. 

The bright side in these comps is that each of these guys (except Brandon Knight, who just completed his second season) became productive NBA players for many seasons. This suggests that Beal's floor is "above average" and his ceiling is in the vicinity of "All-NBA."

Taken together, the similarity scores suggest the Wizards should have a dynamic back court that could produce legitimate All-Stars and All-NBA candidates within the next 2-3 years.