patsen Posted May 30, 2009 Share Posted May 30, 2009 Last month, I started a study to try to learn the effect of ratings on stats. The technique I used for that was create a custom league with groups of player ratings set to different numbers to see the effect of a change of ratings on stats. I've since taken on a different project, so I wanted to release the work I did, in hopes that someone else can finish it. I took the original 05 rosters on the original game (the 08 mod had issues with MVPSavReader), and set all batters to straight 70 ratings, which I deemed league average. I would then sim one season using those stats. I used MVPSavReader to get the stats to HTML form, then wrote a script to leech the data, and save it in csv format, for analysis in Excel. As expected, all the players had similar numbers with expected variance. I then divided the league in 5 approximately equal groups (by skin tone to avoid biasing), and changed one stat. For example, I would give players in group A plate patience of 50, while group E had patience 90. I'd then look at the numbers and determine the effect. I'd look at the scatterplot and R^2 to determine which stats were affected and the expected amount. For contact and power, I changed vsL and vsR by the same amount, making batting and throwing arm irrelevant. I only ran contact, power and place discipline, but they had given very interesting results: - Plate discipline affects walks, but not SO. It also seems to have a slight effect on BABIP. Didn't affect power. - Contact primarily affected SO, but not walks, and had a secondary effect on Avg. - Power didn't affect SO, walks or Avg. It didn't affect 2B or 3B either. Only HR were affected, surprisingly. (2B had a slight if not insignificant decrease with higher power) I disabled injuries and all the luck-based options I could. It seems to affect games, not PA. Before changing the ratings, leadoff hitters with low durability would have less games, but more PA than cleanup hitters with high durability. I also disabled DHs and transactions to limit the number of players affected. Like I said, I've since taken on a different project, and didn't want to abandon the work, so if someone's interested in trying to complete them, contact me, and I'll set you up. I guess I can't attach the xls file I was working on, so I'll send it by e-mail if someone PMs me. Happy Wieters Day, Pat Link to comment Share on other sites More sharing options...
DylanBradbury Posted May 30, 2009 Share Posted May 30, 2009 Hey man, if you attach zip (compress) the .xls file first, you should be able to attach it. Link to comment Share on other sites More sharing options...
patsen Posted May 30, 2009 Author Share Posted May 30, 2009 mvp_study.rar Ok, the Excel file's up, I guess... but I'm looking for the rest of it. From there, people can look at the current samples I took. There's over 50,000 PA for each sample, so the results are pretty meaningful. I'll look for the rest of it, and try posting them. I can't post MVPEdit files either, so I'll compress that too when the rest is in. I just have one request, let me know if you get anything new with it, so if I try doing more tests, we aren't duplicating work. Link to comment Share on other sites More sharing options...
patsen Posted May 30, 2009 Author Share Posted May 30, 2009 Found the rest of the files... You need MVPSavReader to use it, I've been using 2.0.0... once you run a sim, save the dynasty, then run MVPSavReader on it. Run all MLB teams, and only stats pages are necessary (yes to detailed). That'll put the HTML files in the Rosters folder. Put compile.jar in that folder, and run it. It'll create stats.csv (I included the results of my other runs) Paste it into sheet3 (just realized I didn't name them), and they'll assign the groups, and it'll sum itself up on sheet1 (sheet2 is just the lookup for the groups). You should copy and paste values, so you have all the tests handy. The goal is to try to figure out the effect of ratings on stats, so that we can develop a much more powerful ratings engine. Pitchers and fielding stats will be different, but if someone picks up the project, I'll write some scripts to help out. I'm releasing this out because I took on another project, and didn't want to abandon the work, if someone else wants to finish it. I set the spreadsheet up one way, others would likely do their own, so do your own thing. Good luck, Pat ratingsstudy.rar Link to comment Share on other sites More sharing options...
MarkB Posted May 30, 2009 Share Posted May 30, 2009 Thanks for the time and effort you put into this patsen. Nice job. :) Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.