Regression in Fantasy Baseball

[bahamamike]

Hey all,

I haven't been as active on this site as I would like to, but I have a great reason to be able to get my activity back up in fantasy baseball this season. I am taking a regression analysis course this term, and we are supposed to use data from our dissertations as part of the class. Unfortunately for me, my data is all qualatative so I can't use it but the course is required, so my prof told me to just find data I am interested in. I know regression has been used in fantasy analsysis, and was just curious as to the types of regressions and data used, and exactly for what purposes. Don't worry I am not trying to steal trade secrets, just more interested in a theoretical debate here, to help me select my data set and developing my own regression model.

Thanks

Mike

[bahamamike]

I meant to add in here, that at the end of the quarter, I would be more than willing to post the results and formulas I used for further debate/discussion.

[Todd Zola]

Paging Gary Jennison, please pick up the white courtesy phone...

[Guest]

I assume since you're taking the class the general concepts are all in your mind - I use regression/correlation analysis all the time to try to normalize baseball peripherals into expected results instead of actuals, especially where luck/chance is involved.

So we come up with trendlines for K vs ERA, BB vs ERA, HR vs ERA, hitters K vs AVG, BB vs AVG, K vs BABIP, BB vs BABIP, etc. You name it, I've messed with it.

Correlations of age vs player growth in statistical categories, correlations of pitcher ERA vs winning percentage, team hitting and pitching peripherals vs winning percentages, team individual category stats vs winning percentages.

Players K, BB, HR, ERA or AVG in Year X vs Year X+1, X+2, etc....

Hope thats at least a start.

[bahamamike]

thanks, that is a good start, and I will probably start posting more questions, comments once i start playing with the data a little, looking at the trends etc.

[bahamamike]

Its been a while with me for regression etc, so Sorry if I ask basic questions. OUr first assignment, was to just run a simple linear regression with one dependent and one independent variable, so I am starting mine with ERA to K's per 9 inning for the 2008 season. The assignments are supposed to get more involved, but this one is just to refamiliarize our selves with the procedures.

I made ERA the dependent variable, and K per K per 9 inning the independent variable. I found that these were correlated to 99.99%. So I am trying to figure out how to relate this to fantasy. Or before I do that, am I going to have include more independent variables and time series?

Or am I totally looking at this wrong.

Thanks

[Guest]

K/9 to ERA doesn't correlate at 99.99%. If I remember correctly it's around .3 or .4 for R.

I'm not sure what the purpose of the exercise is - what I would do is calculate the trend line in excel and have that be the equation for expected Y (ERA) knowing X (K/9). But your professor may want something else.

[bahamamike]

K/9 to ERA doesn't correlate at 99.99%. If I remember correctly it's around .3 or .4 for R.

I'm not sure what the purpose of the exercise is - what I would do is calculate the trend line in excel and have that be the equation for expected Y (ERA) knowing X (K/9). But your professor may want something else.

Sorry, I wrote that wrong, it was the signifcance level. the R was .2 and a really low R squared (.048).

Sorry for screwing up the terms, haven't done stats in years, and have mostly focused on a qualatative research path, but the course is required