Even if it's Peter making the projections and Alex doing the pricing, I'm sure Alex's projections wouldn't be that far off from Peter's.However, this isn't the case.
The idea behind Patton's old projections was that there were going to be some outliers at the end of the year that we didn't see coming. Patton's goal was to project a few outliers in each statistical category - particularly the offensive ones - to give his projections some flavor. Even if the actual players who hit the mark in some of these categories were wrong, the idea was that there are exceptional players every year so it's worth projecting exceptions even if can't see exactly who they will be.
Rotoman's projections tend to avoid this model, and - as I've mentioned before - assume regression to the mean for almost every player.
I should point out that this isn't limited to Rotoman. Baseball Prospectus' PECOTA takes the same approach.
Let's look at RBI as one example. In 2009, Rotoman projected that 16 Major League hitters would drive in 100 or more runs. PECOTA, meanwhile, projected that 12 hitters would accomplish the feat.
As it happened, 28 hitters managed this in 2009.
Was 2009 a red-letter year for offense? No. In fact, 2008 saw 29 hitters reach the 100 RBI barrier.
I doubt that Rotoman or PECOTA is saying that offense is going to drop and that only a select few will reach 100 RBI. I do believe that both systems are saying not to assume the same level of production from most players due to a number of factors.
Getting back to Patton, his projections were far more likely to come close to mirroring what had happened the previous year. My guess is that if he were still projecting he would have had about 30 hitters with 100 RBI or more. He would have incorrectly projected some of the hitters in question, but would have come closer to predicting the actual amount of offense in the Major Leagues than most mechanical systems ever do.
The bigger question is: is this important when you're trying to put together a Rotisserie team?
It should not be. You're ranking players and establishing bid limits based on stats, so if Rotoman has fewer stats projected for 2010 than Patton would have, this won't change the pecking order in your league's projected standings.
What it will change is how "accurate" those projected standings will be in terms of guesstimaing what it will take to win each category. If you're a proponent of this method, this is where you probably miss Patton's projections the most.
1 comment:
Regressing to the mean gives more "accurate" projections, meaning that the error is reduced on the outliers.
Projecting 26 guys to drive in 100 runs is good fun, but you end up with more spectacular overreaching errors. I try to goose a few guys up and knock a few guys down each year, the one's I have strong hunches about.
When your bid prices aren't tied to your projections you can have more fun with projections. But I think when that fun gets out of hand it obliterates the useful part of projections, which is to help you profile players to target and how they fit on your team, and then manage that information during the draft itself.
Post a Comment