Rhoids Sports Analysis
AL hitters ranked by game state wins
By (roughly) subtracting the probability of a win before a player's PA from the probability of a win after a player's PA, these guys calculate the win probability change attributable to each PA and thus derives a "probabalistic win" model that they call game state wins. There are problems with this sort of metric, of course, but some people are stumbling blindly around the concept of leverage so I figured I'd point it out.
Among the caveats to include in discussing this statistic:
1) Does not include defense
2) It doesn't appear to be using park factors with its 2005 numbers, so the numbers have the potential to change once the season is over and park factors are finalized.
3) The system assumes that the win probability after after the PA is the average win probability for that event, as far as I can tell. This approach beings us to a line-drawing contest where we're looking at statistical averages for some things in evaluating players (game win probability matrix states) while ignoring statistical averages on other things (the win value of a double, for example).
4) The site fails to address the question of whether a player deserves "extra credit" for good timing. It merely measures that "extra credit" using game wins probability matrix states.
Edited by dnramo, 21 September 2005 - 12:05 AM.