I'm responding to a few of the comments on my last post about the Iverson-Miller trade.  Normally, I'd do this in the comments section, but this blog site only gives me a limited amount of space for each comment, so I'm doing it as a new post.

1. EdHardiman wrote: I like the post I just don't find arbitrarily cherry picking stats valid in micro-analyzing. It has some macro-validity in quantifying a team but no economic model however well padded with math has explained the movement of an individual item in any market, be it financial or sports. There are too many unquantifiable variables that exist from luck to injury to a butterfly flapping its wing in Mongolia that are and can't be accounted for. In other words no metric will ever predict the flip of a coin. Even if the metric is right 2300 times in a row, it is still simply an illusion that satisfies our need to impose order on a chaotic system. 

My Response: 

Regarding the cherry-picking of stats, I wanted to use the last four years of both Miller and Iverson in order to give me a sufficiently large sample size.  However, I ended up using the last five years of Iverson to be generous to Iverson fans, as Iverson had one of his better years five years ago.  If I had only used the last four years of his stats, his Win Score would have been even lower. 

I think that you are glossing over the value of statistical significance.  When the correlation between two variables is found to be statistically significant, there can be little argument that these variables move together, regardless of how minute the movement.  This is not to say that one variable "causes" the other one to move, but you cannot deny that a relationship exists.  And when you have a sufficiently large sample size (over 1000 observations), we know that the finding of statisical significance in the relationship between variables isn't just random luck of the draw.  On page 103 of Wages of Wins, the authors have a table that shows the impact on wins of a one-unit increase in any of 16 NBA statistics.  For example, every missed field goal decreases a team's total wins by .034.  Put another way, if you miss 100 field goals, you'll find that your team's wins will decrease by 3.4.  The calculations were made based on data from 1000s of NBA games, and the statistics were regressed against the dependent variable of winning.  These relationships will continue to exist as long as NBA basketball is played under its current rules.  I just don't see how such statistically significant figures can be compared to the "flip of a coin."

Also, while the metric doesn't predict who's going to win a game on any given night, it is very good at predicting how a team will perform over the course of a season.  Page 110 of Wages of Wins  shows the authors' predictions of team wins based on player production, compared to actual team wins from the 2003-04 season.  The average error was just 1.67 wins, and the largest amount that the authors' model was off is 4.41 wins.  I find this to be excellent empirical evidence of the validity of the models.  Moreover, the authors' model predicted that the Lakers should have won 51.7 games that year, and the Pistons should have won 56.7.  Actually, the Lakers won 56 games, while the Pistons won 54 games, and the sportsbooks had the Lakers as heavy favorites going into the Finals.  We all know who won that series, and the authors' calculations showed that it really shouldn't have been a surprise, as the Pistons players were more productive than the Lakers' players.  The unquantifiable factors of luck probably contributed to the Lakers overperforming during the regular season and the Pistons underperforming.  And when I say "luck," I'm talking about those factors that contribute to winning, yet are not quantified by any measurable statistic.

2. TrainontheBall wrote: I know all about win shares (consider ourselves in the minority), etc. I don't agree with it 100 percent (after all, it's a questionable formula that's being used), but I do think it makes a great gauge - if not fun one to check out.  Thus, bear in mind that the difference between Miller and Iverson is .054 points. Win Shares doesn't take into account variables like momentum, clutch performances, leadership, etc. But I understand what you're getting at, and yes, I would agree that Andre Miller is more efficient than Iverson.

My response:

The measurement that I used in my post actually isn't a Win Share.  Instead, it's a relative measure of player performance that doesn't actually have a meaning in terms of wins produced.  Win Score is simply a way of comparing the efficiency of players who play the same position.  However, it is possible to calculate each player's contribution to win production.  The authors use a measure called Wins Produced Per 48 Minutes (WP48), and this actually has a concrete meaning.  For example, Iverson's career WP48 is .086, meaning that he produces .086 wins for his team per 48 minutes that he plays.  Now, think about what an average player's WP48 would be.  If we assume that an average team wins 41 games, then the five "average" players on this average team would have a WP48 of .1 wins.  Since each contributes .1 wins per 48 minutes played, the summation of their efforts over 48 minutes is .5 wins.  82 games multiplied by .5 wins is 41 wins for the average team. 

Back to Iverson: his career WP48 of .086 is .014 wins below average.  That's not good if you're a Nugget fan thinking that he's the "answer" to the team's playoff woes.  Some of this is softened if you expect Iverson to perform closer to his performance in the last two years.  He had a .152 WP48 in 2004-05, and a .127 WP48 in 2005-06.  Still, that's a far cry from Andre Miller's .217 WP48 this year.  Perhaps, Miller would have eventually declined to his performance from last year of .170 WP48, but that's still better than Iverson has ever produced in a single year.  As a simple comparison, Steve Nash had a WP48 of .319 wins last year, so that's the elite level.  While Miller obviously hasn't approached that level this year, anything above .2 WP48 is great for an NBA point guard.  My main problem with the trade is that the Nuggets took a guy who had become a major contributor to their team in terms of win production, and they traded him for a commodity that's never been able to sniff a WP48 of .2 or higher.  Perhaps, George Karl will be able to harness Iverson's talent and turn him into a winner.  But Iverson's history suggests that won't happen.