Page 2 of 5
Posted: Thu Jan 13, 2005 12:39 pm
Interesting stuff. What exactly was the formula you used to produce your "theory era"?

Posted: Thu Jan 13, 2005 1:19 pm
Amazinz wrote:Interesting stuff. What exactly was the formula you used to produce your "theory era"?

Thats really the biggest question. One thing that caught my eye was your statement

LBJ wrote:OK well I was playing with a bunch of numbers today seeing if I could find multipliers for K/9, BB/9, HR/9, and Team IPAvg (ratio of balls in play that a pitcher's team allows for a hit).

Why did you decide to use Team IPAvg instead of the league IPAvg? Do you use the team's IPAvg ratio to the league IPAvg?

Posted: Thu Jan 13, 2005 2:12 pm
Tavish wrote:Since when were K's any different than normal outs LBJ?
I will give this question a very brief expalnation.... K' are independant of regular out because the pitcher controls the K himself, instead of relying on fielders to help out.. A strike out is with out a doubt, an out!! (sounds llike DR Suess). A ball in play has a 70 % chance on average of becoming an out... a good K rate is the skill of the picher only, and a good independent stat to use in formulas. this is a very basic explanation, but the basis of all dips work. Focus on only the skills that a pitcher controls all by himself, without the aid(or lackthereof of others......Hope this helps Rotodog

Posted: Thu Jan 13, 2005 3:29 pm
Tavish wrote:
Amazinz wrote:Interesting stuff. What exactly was the formula you used to produce your "theory era"?

Thats really the biggest question. One thing that caught my eye was your statement

LBJ wrote:OK well I was playing with a bunch of numbers today seeing if I could find multipliers for K/9, BB/9, HR/9, and Team IPAvg (ratio of balls in play that a pitcher's team allows for a hit).

Why did you decide to use Team IPAvg instead of the league IPAvg? Do you use the team's IPAvg ratio to the league IPAvg?

Well the formula wasn't much of a formula... I just ran a regression with all the info; K, BB, HR, and IPAvg. The multiplier turned out to be: -0.179 for K/9, 0.320 for BB/9, 1.469 for HR/9, 19.475 for IPAvg (8-o) and the base # everybody starts off with, the intercept, is -2.842. I used more decimal places for the actual calculations but to save space, I limited the ones I wrote to 3. No idea how IPAvg turned out to be so important, but it doesn't vary much so despite being a huge number, doesn't sway the Theory ERA too much. Going from the best team for IPAvg to the worst team, an ERA would drop only about .65 points based on IPAvg alone. The switch from a neutral park to Coors field would change your ERA by about .75 ERA points.

I used Team IPAvg because I found that doing this was more accurate, given each team has a unique ballpark, unique defense, and therefore a unique IPAvg. At first I didn't include IPAvg at all, and if I included the league wide IPAvg, the only changes that would have been made would have been to AL relative to NL. The adjusted R square that the regression spat out was 68 with Team IPAvg and 62 without it. I realize the sample size isn't that great, only 162 games worth of data, but by using teams, I can get an idea of how switching teams would affect a pitcher's performance. If you change their HR rate based on the park factors, and then change the IPAvg from his old team to his new team, it's not so much a mystery of how well he'll perform. It's not perfect... but I think it helps. Using a 2 or 3 year average for IPAvg might be better, and considering the weight the formula gives it, it's important that the IPAvg is accurate.

I also would like to see the effect that changing leagues has on a pitcher. From a test I did, a pitcher's K/9 tends to be 0.50 higher just by being in the NL. Your BB/9 is also higher though, by .013 points. For HR, there's only a small advantage to being in the NL but it's more important to know which team they're traded to so you can account for the ballpark. I have no idea how accurate that stuff is and if anybody has more information about pitchers switching to the NL/AL I'd be happy to hear it.

And again, I need to factor in G/F ratio I think... or else have some sort of multiplier that's based on a pitcher's past comparison to the Theory ERA. I'll work on that today. Any more suggestions?

Posted: Thu Jan 13, 2005 4:02 pm
LBJackal wrote: And again, I need to factor in G/F ratio I think... or else have some sort of multiplier that's based on a pitcher's past comparison to the Theory ERA. I'll work on that today. Any more suggestions?

uggh...i thought i had found where someone just used .10 for GB and .15 for flyball pitchers, now i can't find the link.

here's FI's article though with lots of handy links
http://www.futilityinfielder.com/dips03.html

of course, it's entirely possible that '04 is just skewed for whatever reason, though the problem has come up before.

Posted: Thu Jan 13, 2005 6:53 pm
rotodog wrote:
Tavish wrote:Since when were K's any different than normal outs LBJ?
I will give this question a very brief expalnation....

It was an inside joke, sorry.

Posted: Thu Jan 13, 2005 9:54 pm
Tavish..........you, you bugger you ......well good news is that someone that never heard of the basic idea, probably will read it...........It may help someone someday

Posted: Thu Jan 13, 2005 9:58 pm
LBJ wrote:I used Team IPAvg because I found that doing this was more accurate, given each team has a unique ballpark, unique defense, and therefore a unique IPAvg.

The problem I have with using team IPAvg is that it leaves your Therory ERA as a defense based statistic. It works well as a quick and simple park adjustment, but leaving the team's defensive ability in the mix makes pinpointing the responsibility of the pitcher difficult. How much of Lowe's Theroy ERA difference is based on pitching in Fenway and how much is based on the bad defense that was behind him?

I think the value of the results definitely depends on what you are looking to find out. If you are looking for specifically how well the pitcher performed I would stick with DIPs, if you are wanting to see how well a pitcher performed within his current team he is on then you are on the right track.

Posted: Thu Jan 13, 2005 10:17 pm
Tavish wrote:I think the value of the results definitely depends on what you are looking to find out. If you are looking for specifically how well the pitcher performed I would stick with DIPs, if you are wanting to see how well a pitcher performed within his current team he is on then you are on the right track.

Yeah that's what I was looking for, how well he should have done given his team and ballpark. DIPS is good for measuring true talent of a player but in fantasy it counts just as much if you have a good defense that helps your ERA or if you're actually better but with a worse defense. So basically, that's why I decided to factor in the team IPAvg.

Any suggestions on fixing the discrepancy of players who traditionally have ERA's higher or lower than the theory ERA? Tom Glavine always has an ERA lower than what the Theory says. Same for Webb. Do I manually adjust these by looking at the trends, or is there another stat to factor inot the regression? G/F ratio would probably be good but unfortunately Lahman doesn't have it in his database.

EDIT: Just got an idea: what about the player's individual IPAvg, or the ratio of his IPAvg to his team's IPAvg?

Posted: Thu Jan 13, 2005 10:20 pm
LBJackal wrote:
Tavish wrote:I think the value of the results definitely depends on what you are looking to find out. If you are looking for specifically how well the pitcher performed I would stick with DIPs, if you are wanting to see how well a pitcher performed within his current team he is on then you are on the right track.

Yeah that's what I was looking for, how well he should have done given his team and ballpark. DIPS is good for measuring true talent of a player but in fantasy it counts just as much if you have a good defense that helps your ERA or if you're actually better but with a worse defense. So basically, that's why I decided to factor in the team IPAvg.

Any suggestions on fixing the discrepancy of players who traditionally have ERA's higher or lower than the theory ERA? Tom Glavine always has an ERA lower than what the Theory says. Same for Webb. Do I manually adjust these by looking at the trends, or is there another stat to factor inot the regression? G/F ratio would probably be good but unfortunately Lahman doesn't have it in his database.

I'm not too familiar with the Lahman database, but I know that ESPN has G/F ratios, including the actual amounts. If you could take from those, it might work.