You may or may not have seen that according to Wages of Wins' Wins Produced metric, LMA is a well below average player (and Przybilla is having a Hall of Fame year).  For those not familiar with the stat, .100 is an average player's production.  LMA is at .032 and Pryz is at .356 (Roy is .226 for comparison).

This intrigued me because, while no stat analysis can ever be completely comprehensive, Wins Produced (henceforth called just WP) is usually much better at separating real contributors from the Stat-Bo players.  To figure out what this meant, I first had to figure out how WP is calculated.

Calculating Wins Produced:

The actual process is very complicated.  Here's my attempt at simplifying the process.  You can skip to the analysis below if you don't care how it's calculated.

Preparation Step:  Figure out the correlation between offensive and defensive efficiency (i.e. points per possession) to the number of wins each team has.  Once you know that X points per possession is approximately worth Y wins, you can calculate the value of each individual stat that goes in to calculating points per possession.

Step 1:  Use the values from step 1 to asses production of each player.  The stats that go in to this calculation are 2 pointers made (2FGM), 3 pointers made (3FGM), free throws made (FTM), field goals missed (FGMS), free throws missed (FTMS), offensive rebounds (REBO), defensive rebounds (REBD), assists (AST), turnovers (TO), blocks (BLK), and steals (STL).  There is also an additional value that takes a players percentage of the team's total fouls and penalizes them for that percentage of the oppenents free throws made (FTMO).  The exact values change because of how they are calculated, but it is roughly equal to the following:

(2FGM + 2*3PFGM - FGMS + .55*FTM - .45*FTMS + REBO + REBD + .55*AST - TO + STL - .55*FTMO) / 3

The formula above gives total production.  This is divided by minutes played to get production per minute and then multiplied by 48 to get production per 48 minutes (i.e. per game).

This model has rebounds, turnovers, and steals equal to a 2 point field goal (i.e. 2 points) rather than the usual model that has them worth half the value of a 2 point field goal (i.e. 1 point).  3 pointers also end up being worth double what 2 pointers are worth.  Here is a chart (from Wages of Wins) with the exact values from 91/92 through 06/07.

Value of Player and Team Statistics

1991-92 to 2006-07

Step 2:  Adjust each individual's score based on how much they benefited from their teamates blocks and assists.  This is done becuase blocks create rebounds (that may otherwise have been a made shot) and assists increase other player's chances of hitting shots.  I skip the exact process and just mention that the adjustment is fairly small (usually less than .01 WP/48 minutes, rarely more than .02).

Step 3:  Adjust for team defense.  WP does not calculate individual defense.  Instead, it holds every member of the team equally responsible for their team's defensive efficency (points allowed per possession).  Again, I'll skip the exact process and mention that this adjustment is also fairly small (usually less than .015 WP/48 minutes, rarely more than .03).

Step 4:  Adjust for position played.  This is one of the main things that sets WP apart from other metrics.  WP acknowledges that the big men get more rebounds and have less turnovers, not because of their skill, but because of the nature of their position.  There are three adjustments:  one for PF/C, one for SF, and one for SG/PG.  A player that splits time between areas will ge the average of their two positions (e.g. Outlaw would get the average of the PF and SF adjustments).  The adjustments are calculated by a brute force "add up the numbers and divide by everyone in the league at each psition and compare" process.  Here is the chart for the 06/07 season:

Value of Adj. P48 Across Positions

The values on this chart are just plain ol' subtracted from the production/48 number that we've been adjusting through the process.

Step 5:  At this point we have each player's value relative to "the average player" (i.e. an average player will be equal to 0.0 at this point), but the average player is worth more than 0 wins because an "average team" wins half of their games.  An average team has .5 wins per 48 minutes, so the average player is worth .1 (.5 divided by the 5 players on the court).  The actual average wins is actually about .099 / 48 minutes because of overtime games.

To finish the process, all we do is take the value that came out of step 4 and add .099 to it.

Analyzing Wins Produced:

There are several things that jump out from this model:

- WP loves rebounds

- WP loves making the most of your minutes on the floor

- WP has a defensive adjustment, but still doens't really nail down the hard-to-quantify value of an individual's defense.  It can even end up adjusting a player up/down when they really deserve the opposite because of their team's defense.

- WP doesn't care too much about total points as much as it cares about efficent points.

WP Strengths:

- WP is completely objective.  There is no tricky "how many points is an assist worth?" guess work.  It just analyzes data to calculate that X stat has been worth Y wins.

- WP isn't impressed by large inneficient point totals.  If you need to jack up 25 shots to make 10, WP is gonna give you the smack down.  You better make your free throws too.

WP Weaknesses:

- The objectivity is good, but the lack of subjectivity can also be bad.  More on this below.

- The defensive adjustment is very crude and probably only marginally effective.  It probably gets it right more often than it gets it wrong as good defensive teams are going to have more good defensive players and bad defensive teams are going to have more bad defensive players, but this also means that when Glenn Davis was warming the bench on the Spurs, he was picking up a nice fat WP team defense bonus.  The adjustment is fairly small, but it's still something to consider when WP is way off from conventional wisdom.

Wins Produced vs LMA (and Przybilla)

So what's up with LMA?  The following are all factors:

- The Blazers have a below average defensive efficency this year (109.5 vs 16.9 league average) so he losing a little there whether he deserves it or not, but not a lot and other players are able to overcome it just fine.  LMA is usually one of our better defenders so he's probably unfairly penalized here.

- He doesn't rebound a whole lot for his position.  He is 35th among PFs in rebounds per minute (and WP loves rebounds).

- Since he is mostly a jump shooter, his field goal percentage isn't that great for a PF either (21st among PFs in field goal %, 17th in free throw %), but his position adjustment still has it's curve set by guys that primarily shoot lay-ups and dunks.

I don't think it's time to hit the panic button yet.  He is still young, he's learning more post moves, and even just this year his shot has started falling a lot more.  His defense is also improving quite a bit each year.  He also gets a lot of minutes on some nights when the offense isn't going through him, which hurts "per minute" stats like WP... but he could make up for that with more rebounds though.

Considering his age and experience level, and the prgoress he's made so far, I'm not too worried.  It would be nice to have more and I expect more down the road from a #2 pick, but all things considered I'd call his performance pretty normalwhelming.

As for Przybilla, I like that this shows how superb he's been this year... but .356 seems a little crazy.  Dwayne Wade, for example, is at .350.  I think the efficiency loving WP gives him too much credit for his 75% shooting.  To a math equation, all 75% shooting is the same, but we all know that vast majority of his shots are wide open dishes from a driving guards or putbacks.  Yes, most post players can say that to some extent and that's what the whole position adjusment is about, but Pryz takes it to the extreme.