Why don't the big projection guys put out midseason projections? You think there would be a market for them. I would like to know what Shandler, ZIPS and PECOTA have to say about some of the upstarts and slumpers. I suppose they don't want to undermine their preseason projections because they'd be admitting some wrongs.
I wouldn't expect them to be all that different than the preseason ones except for a few alterations for development/regressions. The biggest change would be for expected playing time and that is something most people could likely do themselves.
But take a guy like JJ Hardy. He's already 2 homeruns over PECOTA projections. PECOTA does adjust on their website for playing time but it seems they don't change anything else, and JJ Hardy is not going to hit -2 homeruns for the rest of the year nor is he going to hit 0. With slow starters it's a little easier for someone to figure out since you can take the original projection and fix it to account for the slump.
Projections have to be different because another half a season has been played. There is more data to account for especially for guys who haven't played much. Projections change from year to year so why wouldn't they change from month to month? There's so much data analysis that goes on before the season starts and then once the season begins it seems to get thrown out the window for gut feeling and total trust in that preliminary data. Why don't we continue to calculate during the season instead of throwing the books in the garbage and go back to them in February? Like I said before, I think it's because the people compiling the data might have to much pride at stake to change their preseason projections and admit wrong.
I'm sure the totals would change, but the rates likely wouldn't. Most of the midseason projections would be nothing more than taking the preseason projections, dividing them in half and adding that to the current total. I doubt anyone is going to come out and project Hardy to 50 HRs. They would take their preseason projection of 20 (or whatever) and figure him to hit 10 more in the second half.
Most midseason projections might be that way, but I don't see why a full season needs to be played to analyze a player again. I understand it's a better sample but drastic changes can be made in a players projections in one year which is in itself a small sample size when it comes to a baseball career.
Let's take PECOTA for example: I've bought Baseball Prospectus for years and know that drastic changes are made season to season in players projections. That's a given. PECOTA uses comparables to draw up conclusions on how a player will perform. 80 more games should change a players comparables if PECOTA was way off on the players projection.
Most other projection sites use 3 year splits so why not drop the first half of the 2004 year and add the first half of the 2007 year to get a more accurate picture of how the player will perform in the second half.
Again, I just don't get how all of this statistical analysis is done preseason and then it just comes to a halt and treated as the gospel truth until the next preseason when new projections come out. I just think it would be beneficial to keep that analysis going throughout the season or at the least revisit it at all-star break.
GotowarMissAgnes wrote:Besides, PECOTA projects a range of performance. Hardy may be over his median PECOTA, but his 90% projection?
He's one HR away from hitting his 90 percentile mark. The average is a little better up at that mark though.
That's not really the point though. It's not that projections are wrong and should be changed. Projections are never going to be 100 percent accurate. The point I'm trying to make is that projections should be updated more than once a year just for the sake of 80 more games being played.
Thats why I love my hockey prjection (Dobber you rock ) he has his own website with articles everyday, a message board where you can ask him questions, daily ramblings and of course a mid-season guide