I've collected a "considerable" amount of data to try and determine some rudimentary park factors for Hardball from the limited amount of data that WIS gives us.

Because WIS doesn't give us runs scored when we look at stats for home and away, these won't be exact to what you might find other park factor systems are based on. The stats that we are given for batting home/away and pitching home/away that we can do some simple work with are H, HR, AVG, OBP, SLG, and OPS(if we add OBP and SLG for pitching). 

While not exactly a perfect measure of offensive output, I based these park factors on OPS, but can provide ratings for the other categories listed. 

To get to the final number, this is the formula I used: (home batting OPS + home pitching ops)/(away batting OPS + away pitching OPS). I didn't do any of the different adjustments that other systems may use, because the data from WIS is lacking. 

I grabbed every completed season from 12 worlds, encompassing more than 821 seasons: American Pastime, Big Sky Alumni, Cooperstown, Fingers, Mantle, Major Leagues, Moonlight Graham, Public Works, Second City, Summer of '49, Vin Scully, and Warning Track to come up with these. Some cities are lacking data, mainly Omaha, Tucson, Montgomery, and Cheyenne just to name a few.

The second column is the number of seasons of data I've collected for each city. The third column is the average park factor for OPS for the seasons I've gathered; 1.000 is exactly average. The last column is the margin of error based on a 95 percent confidence level. For cities with more than 70 seasons worth of data, it typically holds to within .006 +/- the average, so we can be fairly certain it's pretty close. To get a MOE of less than .004 would take more than 300 seasons worth of data for each city, depending on its standard deviation.

One thing that should jump out at people is the fact that on the high end, Colorado and Santa Fe both have PF's large than those on the low end by a considerable amount. 

Another thing is despite being at all -4, it appears that Tacoma is no more depressing offense than San Diego, Seattle and Burlington. Maybe that isn't actually the case if I were to include 40 more seasons worth of data for Tacoma, but I found it interesting. 


Thoughts?

I'd be interested in see it, but I can't resize the examples.  They are about postage size and are illegible.  Sounds like a heck of a project though.

I've discussed some of these with lc over the last week or so.   Based on personal experience with parks I've used, it's pretty accurate.

Posted by real_toddb on 2/28/2013 8:22:00 AM (view original):
I'd be interested in see it, but I can't resize the examples.  They are about postage size and are illegible.  Sounds like a heck of a project though.

You can either right click on the picture and expand view the image in it's full size, or there is a setting you can change for viewing them through the forum, but I'm not entirely sure where to find that. I kind of stumbled upon it a long time ago.

1)  Thanks, this is terrific work.

2)  While I understand the methodological limitations based on available stats, OBP is significantly more important to run-scoring than SLG is, so these numbers will relatively overrate park effects in either direction due to HR, and underrate effects due to 1B.  The old dogma was that (1.4*OBP) + SLG was about the correct ratio for a single stat, but I've read that that multiplier (the 1.4) may even correctly be as high as 2.
     If it's just tweaking a couple of Excel formulas, would you want to run the same analysis using (1.4*OBP) + SLG instead of unadjusted OPS?  Parks like OKC, SFO, and SJU will move towards the middle, I'd imagine, if you do that.

3) That said, the use of some sort of OPS-like number instead of runs should be an excellent surrogate for runs scored in HBD, since there are no differences across parks in errors, stolen bases, etc. in HBD.

Posted by dedelman on 2/28/2013 9:53:00 AM (view original):
1)  Thanks, this is terrific work.

2)  While I understand the methodological limitations based on available stats, OBP is significantly more important to run-scoring than SLG is, so these numbers will relatively overrate park effects in either direction due to HR, and underrate effects due to 1B.  The old dogma was that (1.4*OBP) + SLG was about the correct ratio for a single stat, but I've read that that multiplier (the 1.4) may even correctly be as high as 2.
     If it's just tweaking a couple of Excel formulas, would you want to run the same analysis using (1.4*OBP) + SLG instead of unadjusted OPS?  Parks like OKC, SFO, and SJU will move towards the middle, I'd imagine, if you do that.

3) That said, the use of some sort of OPS-like number instead of runs should be an excellent surrogate for runs scored in HBD, since there are no differences across parks in errors, stolen bases, etc. in HBD.

I can definitely do some adjustments like what you mention. I plan to post a spreadsheet on google drive sometime in the next few days, probably this weekend, with the PFs for H(minus HR), HR, AVG, OBP, SLG, and OPS so people can get a general feel for how much a ballpark affects performance.

You could even take the crude form of runs created, being OBP * SLG * AB, to see how many runs above/below average an average team would be expected to score in an 81 game season in a certain park.

There is a potential workaround to the runs issue. If you go to home/away splits for player batting statistics, total runs scored for the team is calculated at the bottom. The problem is there isn't a corresponding stat to that in home/away splits for player pitchers statistics. However, there is earned runs. If you take a teams total runs allowed, subtract earned runs, and then just assign half of the difference to both the home and away earned runs, then you would have those stats. But being that convoluted, it just wasn't feasible enough to be able to grab enough data in a reasonable time period.

Here is a google drive spreadsheet with all the values that I've been able to grab. Check it out here:

https://docs.google.com/spreadsheet/ccc?key=0AmhuTkFm0UgYdHBOY3J0bG9vbWFTbC1KUjA2R21ONEE&usp=sharing

In terms of the difference between OBP*1.8+SLG and normal OPS, there doesn't seem to be too much difference in terms of overall rankings. The main difference is a couple points change in each parks coefficient.