Bud Says That We Should Be Rooting For BASEBALL

[Guest Edgy DC]

All that about Johnson is true. Plus add to this

="LeiterWagnerFasterStrongr":o8pqbeb3]He was let go in Baltimore because Angelos disliked him personally, not because he was ineffective-- despite early exits, his Orioles made the playoffs in both years he managed them, and he took the AL Manager of the Year award the day he was fired.[/quote:o8pqbeb3] the fact that his Orioles were screwed over as the eventual World Series champs and their Beacon benefited from cheating.

LeiterWagnerFasterStrongr Mar 23 2009 02:44 PM

The really oversimplified answer is that PECOTA relies on fitting a given player's past stats to the performance of comparables (fetched using the Bill James concept of similarity scores); it then bases the future perf forecast on the historical performance of these comparables (like a "prediction of best fit"). [it's the current statistical crux of Baseball Prospectus' annual season previews-- damn good, oughta-be-required reading.] It tends to be a bit conservative-- wins aren't nearly as evenly distributed as its models predict-- but generally VERY good at predicting both individual and team-- via guessing at depth chart/playing time for the coming year-- performance (for instance, forecasting the Rays' rise last year, minus a handful of wins). Not only that, but the margin of error-- which was best last year among major projection systems like Marcel, CHONE and Vlad (and Vegas betting odds)-- has decreased steadily in every year between its inception in 2003 and today. In short, I heart PECOTA.

metsmarathon Mar 23 2009 03:04 PM

oh, i do too. but as far as using it to compare the WBC participants expected outcomes versus their actual outcomes leaves me wanting a control group.

[Guest LeiterWagnerFasterStrongr]

The really oversimplified answer is that PECOTA relies on fitting a given player's past stats to the performance of comparables (fetched using the Bill James concept of similarity scores); it then bases the future perf forecast on the historical performance of these comparables (like a "prediction of best fit"). [it's the current statistical crux of Baseball Prospectus' annual season previews-- damn good, oughta-be-required reading.]


It tends to be a bit conservative-- wins aren't nearly as evenly distributed as its models predict-- but generally VERY good at predicting both individual and team-- via guessing at depth chart/playing time for the coming year-- performance (for instance, forecasting the Rays' rise last year, minus a handful of wins). Not only that, but the margin of error-- which was best last year among major projection systems like Marcel, CHONE and Vlad (and Vegas betting odds)-- has decreased steadily in every year between its inception in 2003 and today. In short, I heart PECOTA.

[metsmarathon]

oh, i do too. but as far as using it to compare the WBC participants expected outcomes versus their actual outcomes leaves me wanting a control group.