SubscribeNot too long ago I posted a link to an evaluation of how various projection systems did in 2009. My takeaway from that article was that the ideal fantasy projections would combine the free, computer-generated forecasts with a by-hand adjustment for playing time.
How might that be done? My first choice for predicting playing time would be Tom Tango’s Community Forecasts, where he recruits large numbers of fans to predict the playing time for players on their favorite team. Unfortunately, when that survey is opened up in late March, it will be too late for lots of fantasy drafters.
So I’ve attempted a different approach in the meantime. I started with the average playing time that resulted when SG simulated 100 seasons using the CAIRO projections. I then made some manual adjustments based on my own guesses for teams’ depth charts (e.g. putting Alcides Escobar as the Brewers’ primary SS over Craig Counsell). I added estimated PA and IP for free agents. Finally, I knocked 10% off of pitchers’ IP and 5% off of hitters’ PA, which scaled them pretty closely to match the existing projections.
If that sounds like I’m just making stuff up as I go along…well, I sort of am. But I think it’s possible to improve on the wisdom of the computers with some subjective (but commonsensical) knowledge. For those who may be concerned with my meddling, I plan on keeping the unadjusted numbers around, as well.
You can see what the composite projections look like with playing time adjustments by selecting 2010 Composite (Adj.) on the Price Guide. I would love to hear input on where these can be improved. My initial concerns will probably be dropping PA for catchers a little bit and seeing about inflating the top tier a little bit.
Related posts:
Mays,
Great! Have you updated the CSV files on the upload site to include the adjusted playing time?
Mays, what about the fangraphs.com community projections? Would those be a good measure to use? Are those available for you to use?
Also, another question. My auction will be very close to the start of the season. Is there any way to run the price guide and exclude the players who have been cut during spring training?
I’ve been using a blended forecast of CHONE and CAIRO for rate production (ie: HR/AB, RBI/AB, etc). For the playing time I went with the FANS database on Fangraphs. For any player who didn’t have a FANS projection I just reverted back to an average of CHONE and CAIRO. There was a discussion of this at BBTF that I’ve linked to with my user name.
@John: The upload templates had not been updated, but they are now.
Hi Mays,
You may just be testing something out now, but the non-adjusted has disappeared for me (actually, the one that is left doesn’t say adjusted, but it’s giving the values that the previously “adj.” one did). It just occurred in the matter of the last hour or so, in fact, I was looking at the non-adjusted version and then went back to the main Price Guide page, when I noticed the change.
If this is intended then feel free to ignore this comment.
@Josh: I took off the unadjusted numbers when I added ZiPS, because I didn’t want to have to update both versions with the new projections.
If people are still interested in using the non-adjusted numbers, I can add them back. (But it might be a while before I get ZiPS added in with it…)