Originally Posted by blandman
He wasn't bashing the system, he was explaining away the singular long-time inconsistency in the model.
But the model works well with other teams. Generally, large discrepancies are easily explained, usually by players not on the roster having an impact. There is no need to account for future trades or unexpected callups in the model. They are not rating the farm system, bank account of the team, or the general manager. They are rating the players on the roster and nothing more. As a system, it works really really well.
The bottom line is the PECOTA team projections don't outperform "experts" predictions, which usually don't outperform random message board predictions, as far as team W-L results. We all know who has the most talented teams, that's really no secret. In fact, if someone where to tell you what Mike Trout would have done last season before the season started, what kind of odds do you think you could have received if you bet Oakland would finish 5 games ahead of the Angels? There will always be a few surprises and a few dissappointments. Their system is fine, and every team in baseball reads their work, but there is a reason they still play the games.