What are these Marcel predictions? How exactly do they work? Because every single player; let me repeat that: every single player I looked up is predicted to have a worse year in most categories by Marcel. For example, Ryan Howard is expected to hit 40 homeruns. He’s not yet 30 and has hit at least 47 homeruns in each of the last 3 seasons. How by any stretch is he expected to hit only 40 unless he becomes injured? Even young players where improvement should be expected are projected to have worse years. It makes no sense to me.
Marcel has Jason Bay hitting .269 with 24 homeruns in the heart of the Red Sox lineup….I’m inclined to just completely disregard these projections because I think most of them are misleadingly conservative.
The Marcel system is, like every other projection system, based on regression analysis. It uses the player’s last three seasons, weighting the most recent more heavily, and attempts to determine their true talent level based on their prior performances. It then applies an aging curve to project future performance.
The concept of regression to the mean is an important one to grasp in understanding projections. I’d suggest looking over some articles about regression and how it applies to projections.
To read more about the Marcels, how they work, and how accurate they are (short answer – pretty much as good as even the more complicated systems), start here.
Marcel has Jason Bay hitting .269 with 24 homeruns in the heart of the Red Sox lineupâ€¦.Iâ€™m inclined to just completely disregard these projections because I think most of them are misleadingly conservative.
I’m confused as to how hitting in the heart of the Red Sox order would increase home runs or batting average. Care to share the logic on that one?
Comment by Sal Paradise — November 18, 2008 @ 10:33 pm
I just read the tangotiger article and the explanation of how guys who common sense says will hit more homeruns are projected to only hit 25 or so: Because that is what the group of similar players will average out to hit? How does that help anyone? There might as well not be any individual projections….
After reading through the link given in the article again, it seems to agree with me that individual projections are rather pointless. I mean they’re fun and all, but why not just stick with the type of format used in the explanation?
Individual forecasts are completely pointless, if you are looking for a forecast of how many HR a power hitter will hit, +/- 10% HR, 95% of the time. But, if you are looking for +/- 30% HR, 90% of the time, now we’re talking!
At the exact same time, single-game forecasts as to who will win are similarly pointless. If one team has a 52% chance of winning, then guess what: it’s not 100%. It’s 52%.
While individual player forecasts are completely pointless under the +/- 10% standard, then are very valuable at the group or team level under the same +/- 10% standard.