Baseball Stat Optimization

What contributes to a winning team? How much is a hit worth, versus a walk? To address these questions, we consider the team stats and optimize the weights of each statistic to wins using the simulated annealing method of parameter optimization.
Individual stats (such as hits, strikeouts, stolen bases) and not collective stats (runs, runs allowed, ERA) are considered, although the code can be trivially modified to consider any stats.
From these stats, we predict a number of wins, and optimize parameters until the differences are minimized according to the popular "chi-squared" measure.
From this we can estimate wins due to individual players. This does not require a theoretical "replacement player". We can compare any two players.