Basic Projection Formulas?

[GucciBaseballs]

I have been tinkering with my own projections and was wondering what are the basic formulas everyone uses. I would ideally like to get some formulas on here, discuss the pro's and cons, and maybe create some new formulas.

I will start off with a basic projection that is simple but effective (I believe on a scale of 0-1 it was a .68 in accuracy on Prospectus).

.5(last years stats) + .33(2 years ago stats) + .17(3 years ago stats)

obviously the method has its flaws. It discounts players like Casey who have been injury ravaged the past two years.

What formulas is everybody using for their projections?

[DaQ]

I use the ole' take what everyone else says (TSN, Rotoworld) and combine them strategy. I would suck at making a projections formula - my other one would be Get good vision + see how the player is playing + past history (if anything big happened) = results. lol

I need some help w/ Greinke: http://fantasybaseballcafe.com/forums/v ... hp?t=64277

[GucciBaseballs]

I use the ole' take what everyone else says (TSN, Rotoworld) and combine them strategy. I would suck at making a projections formula - my other one would be Get good vision + see how the player is playing + past history (if anything big happened) = results. lol

I need some help w/ Greinke: http://fantasybaseballcafe.com/forums/v ... hp?t=64277

I see a lot of people combining projections. My problem with these services is their projections are crap. They alter them to be least offensive to customers as possible, even if it disregards their statistical data.

Most services use a similar basic formula, but alter them severerly, so that only about 1/3 have a value change of over $5, while in real life it is more like 2/3 have a change greater than $5. Stat services rely too heavily on the previous season for my liking.

[matmat]

usually when I do my own projecting I look at performance per plate appearance and adjust that based on how I think the changing environment will affect a given player.

[GucciBaseballs]

usually when I do my own projecting I look at performance per plate appearance and adjust that based on how I think the changing environment will affect a given player.

what formula are you using, if any?

[matmat]

what formula are you using, if any?

well... first off the formulas that I use do not lend themselves wel to setting in plain text, secondly, a lot of the adjustments are more gut than anything.

as an example, though, I sometimes do a player ranking where I take last year's stats, sometimes also the year before...

1) divide any relevant cumulative stat (R,RBI, HR, SB etc.) by the number of plate appearances.
2) calculate how many standard deviations over/under the mean for the top (number of managers*number of hitters per roster) players in each category the player is.
3) adjust individually as to how I think players will change... say arod is +2sigma in HR over the field... I'd probably knock that down to +1.8 or so because of yankee stadium...
3) sum these values for all the relevant categories to get an "index value" for the players... get rankings from that.

it's a lot of work... but if you just use formulas, instead of doing things by feel it becomes totally mechanical.

[GucciBaseballs]

what formula are you using, if any?

well... first off the formulas that I use do not lend themselves wel to setting in plain text, secondly, a lot of the adjustments are more gut than anything.

as an example, though, I sometimes do a player ranking where I take last year's stats, sometimes also the year before...

1) divide any relevant cumulative stat (R,RBI, HR, SB etc.) by the number of plate appearances.
2) calculate how many standard deviations over/under the mean for the top (number of managers*number of hitters per roster) players in each category the player is.
3) adjust individually as to how I think players will change... say arod is +2sigma in HR over the field... I'd probably knock that down to +1.8 or so because of yankee stadium...
3) sum these values for all the relevant categories to get an "index value" for the players... get rankings from that.

it's a lot of work... but if you just use formulas, instead of doing things by feel it becomes totally mechanical.

Maybe you feel using formulas is mechanical. But using last years stats as a base projection is inherently flawed. As I have stated previously 2/3 of players have a value change of $5 or greater from the previous year. Using last years values as bases will give you very similar stats.

Additionally, how can you predict sleepers and dropoffs using last years stats?

[Zito is God]

I don't need a formula, I have a brain.

[GucciBaseballs]

I don't need a formula, I have a brain.

where did you get that brain? e-bay

Zito I'm gonna make a league like the Cafe Challenge next year and you're gonna be in it. It will be the ultimate test of man vs machine (I would be like the machine, because hopefully i will have a full set of projections from a good formula).

Just curious, do you draft players based on how you feel each draft? Do you have any sort of cheatsheet? Or are you winging it?

I have a brain too Zito, but unlike you, I put it to use, trying to solve the mysteries and myths of the Fantasy Baseball world

[Amazinz]

One thing that I think should be accounted for in any projection formula is aging patterns.

Here is a link to an old study. I've been toying with the idea of creating an updated study to see if the patterns are pretty much the same as they were in this test. Anyone aware of a more up-to-date study? I'd like to see one that was updated through 2003 numbers but I'm pretty sure the patterns wouldn't deviate much.