CR stats on this are meaningless since people tend to play people at a similar level.
I'll bet if Marin or Obi or WW or Rabid or anyb of those in the top 30 played level 15s their win rate would be 80%+. In Colony games it would be 90%.
Ed
Good point, and your hypothetical involves high 40's playing 15s (48 - 15 = 33). What do you suppose (a guess is fine) the win rate would be when 15s play 1s or zeroes? Smaller gap (15 - 0 = 15) between levels, but the gameskills for a 15 are much less developed than for a 48, I should think. Any thoughts?
Well we can use the math behind trueskill to estimate the winning chance. I looked and found that the beta parameter is defined as the skill difference at which the stronger player wins 80% of the time. Generalizing that, the chance to win would be 4^(skilldiff/beta)/(4^skilldiff/beta+1). This is a little overly simplistic, I think, because it ignores the draw chance and uncertainty parameter. But whatever. Beta for isotropic is set to 25.
So if a 48 plays a 15, you'd expect a win rate of 86.2%. For a 15 vs a 0, you'd expect a win rate of 69.7%
EDIT: I found a better resource for trueskill here
http://jmlr.csail.mit.edu/papers/volume12/weng11a/weng11a.pdfThey say that the cumulative distribution function follows a logistic distribution of the funtion (skilldiff/(sqrt(1/2)*beta), assuming beta is the same for both players, which it is here. The CDF is then 1/(1+exp(skilldiff*sqrt(2)/beta)) which looks like this:
http://www.wolframalpha.com/input/?i=plot+1%2F%281%2Bexp%28-x%2F%2825%2F1.41%29%29%29+from+-50+to+50So for 48 vs 15, you get a 86.54% and 15 vs 0 is 69.97%, which is pretty close to my previous estimation.