NBA Player Performance Value Ranking Systems

TRU ratings is our basic attempt at rating NBA player performance. We wont claim
it to be the formula to define performance. Frankly we are just reinventing the
wheel. The most commonly used value right now is the rather complex PER by John Hollinger and the controversial Wages of Wins (WOW) by Dave Berri. TRU was developed about eight years ago before I had even heard of these rating systems. The major criticism of PER revolves around the fact that scoring seems to be overvalued. The main issue with WOW is the high value of rebounding and how anything less than 50% shooting seems to be detrimental. In defense of WOW Berri argues that his formula correlates strongly with team wins, So he doesn't seem to be arguing that David Lee is clearly better than Kobe Bryant just that what Lee does in the court seems to have a stronger affect on winning. PER argues that it's OK for a player to shoot often because it creates more offense for teams. There is a boat load of debates about player usage vs player efficiency on the APBR forums that discusses the issue. The debate seems to revolve around when does too much usage (shot attempts primarily) become a detriment in relation to efficiency (efficient shooting).

Leaving that alone for now, I'll say that TRU seems to split the difference. This wasn't the original intention but that is where we ended up. We value scoring less than PER but many of the other stats are similarily valued. Without getting into sweaty details yet about how we came up with our formulas, I'll give a sample look at how the  results differ. Our sytem works on the functional level. It passes what many would call the eye test in terms of the perception of what many consider the NBA's best players. It's pretty basic and not without flaw but I kinda like that its simple without being stupid. Things like game pace and fouls aren't factored but going through the rigors of adding this seems to just spin the wheel some without changing the prize winners much.

I'll use Kobe Bryant's recent NBA season high 53 point game and I'll use the
"simple" versions of PER and WOW to present the formulas

The Scoring column reflects only the scoring categories. You can see that WOW
relies a great deal on the other stats in its method. Almost 47% of the total
comes from the non-scoring categories. TRU comes out surprisingly more similar
to WOW than PER. Surprising to me even because this is the first time I've
compared these. The primary differences our how scoring is weighted when
compared to the non-scoring stats  PER relies heavily on his scoring (bout
86%) and TRU is in the middle again (67%). I'll have to do this with all league
averages eventually instead of just one game.

The numbers above don't mean much because they only become truly important when comparing them to other games or players in the same ranking system. Just
because Bryant scores 38.6 with PER and 20.15 with TRU doesn't mean that the PER
system thinks that Kobe had a great game and TRU doesn't. In fact all three
thinks Kobe had a great game. It's just that they differ in the importance his
scoring had in the game when compared to his other stats. Kobe was 19-37, 9-17
3pts, 6-7 fts, 10 rebs, 1 ast, 3 steals, 1 turnover, 0 blocks, and 3 fouls. The
Lakers lost!

 If Kobe had shot 10 for 10 with 2 threes and no free throws his WOW score would
lower a half a point to 23. The PER score would drop to 24.3 versus 38.6. The
TRU score would show up as 18.10 instead of 20.15. So if Kobe Bryant had started
out 10 for 10 with 2 threes, The rest of his game would be 9-27 fg, 7-15 3fg,
6-7 ft's for 31 points on about 31 possessions. Can that portion of his game be
considered a good game? Probably. PER rewards it nicely. The argument is as long
as you are shooting and are creating points for your team it should be rewarded.
If the rate is a point per possession than it should be OK. Someone has to score
right?  TRU and WOW reward it slightly and mostly because of the threes and
FT's. No conclusions here in fact it just gives a simplistic overview of the
headache of the Usage vs Efficiency debates. In regards to the credits gained
WOW and TRU are more similar. However like PER we punish missed shots less
because of the offensive rebound chances that missed shots provide. It's clearly
better to miss a shot than turn it over.

 My model for TRU started with the approach of not relying solely on the value of
one possession but on the value that one stat category has when compared to
another. When you make a 2 pointer you get 1 point because you are getting 2
points when a typical NBA game possession is worth about a point. Every other
category is then compared to the 1 we set for 2 pointers. It follows naturally
that a turnover would be -1 credit and a missed shot would have to be worth
more. How much more? I followed the same logic as Hollinger. Because missed
shots lead to another possession for the shooting team nearly 30 percent of the
time, we can calculate a missed shot as -.7 as opposed to -1.  Every 10
shots lead to 3 offensive rebounds which lead to about 3 points or 7 points less
than the average amount of 10 points per 10 possessions. Every 10 made is 20
points or +10. Every 10 turnovers means 0 points or about 10 less than average.
When we put it together we get turnover -1 and missed shot at .7

I'll go over this and explain TRU in better and maybe even clearer detail in an
upcoming post up. Especially how we use it to determine point differential and
wins created. I just wanted to start with an overview of the logic, probably
flawed with TRU and a brief introduction to other widely used linear weights NBA
player value systems.  Post ups welcome!