Analysis of Default XML with Fictional Players v2.0
Posted: Sat Sep 10, 2005 12:37 am
Okay...
Took longer than i had anticipated, but I finally found some time to pull all the data together. It's in the slides that you will find by following this link:
http://home.comcast.net/~zbryanknight/T ... ML_2.0.ppt
Here are the quick hits:
1. Stats are pretty consistent from year to year despite the XML multipliers. The most notable exception is 3b, which is *dead* on
2. The distribution of stat outputs for individual players is more skewed towards higher values and less diverse than in real life. (i.e. there's a big clump of "good" players in puresim instead of having some "good" players, a smaller bunch who are "really good", and a few stars)
This analysis is a little less rock solid than the stuff I do at my job... mostly because the time investment would be rather killer if I got this stuff totally buttoned up. So I'd wind up doing hardcore analysis for like a week straight between home and office... and that's kinda draining.
So if you think I jumped to a conclusion, or if anyone has any questions about my assertions -- specifically what's on the "Recomendations" slide, feel free to ask, and I'll elaborate on my thinking.
Shaun -
I can do more of the analysis that's put forth in section 2 -- the distribution piece -- but it's rather time intensive. The same basic points are put forth in section 3, just in a less concise fashion. Given that I have a stack of grad school applications to get through and a huge project at work, it'd be helpful to know exactly what you'd like to see for anything to be actionable before I went ahead and did it.
~Bryan
Took longer than i had anticipated, but I finally found some time to pull all the data together. It's in the slides that you will find by following this link:
http://home.comcast.net/~zbryanknight/T ... ML_2.0.ppt
Here are the quick hits:
1. Stats are pretty consistent from year to year despite the XML multipliers. The most notable exception is 3b, which is *dead* on
2. The distribution of stat outputs for individual players is more skewed towards higher values and less diverse than in real life. (i.e. there's a big clump of "good" players in puresim instead of having some "good" players, a smaller bunch who are "really good", and a few stars)
This analysis is a little less rock solid than the stuff I do at my job... mostly because the time investment would be rather killer if I got this stuff totally buttoned up. So I'd wind up doing hardcore analysis for like a week straight between home and office... and that's kinda draining.
So if you think I jumped to a conclusion, or if anyone has any questions about my assertions -- specifically what's on the "Recomendations" slide, feel free to ask, and I'll elaborate on my thinking.
Shaun -
I can do more of the analysis that's put forth in section 2 -- the distribution piece -- but it's rather time intensive. The same basic points are put forth in section 3, just in a less concise fashion. Given that I have a stack of grad school applications to get through and a huge project at work, it'd be helpful to know exactly what you'd like to see for anything to be actionable before I went ahead and did it.
~Bryan
