Given that second source is me, I'd like to know how you accounted for the margins of error when you came up with those rankings, and to what confidence interval you are working. Because I don't think the data given in those charts is anywhere near enough to form the conclusions you've formed.
Dode, I appreciate your desire for statistical rigor. I do not feel that this discussion requires the effort that you suggest.
If you want to make the claim you are making you need to be rigorous. The fact is that you have no methodology because you don't have the data. That means your conclusions are suspect. Not necessarily incorrect, but unsupported.
I, also, could not find the raw data you used or your methodology. It was not made available with your outputs.
I've posted the FOL raw data in several places over the years. NAF data is available from the NAF. FUMBBL data is available from FUMBBL sources (usually Koadah's site
). OCC raw data was available from the old BBManager which SuperJC set up. If you'd asked (like I did of you) then I'd have said what the methodology was, but it's a fairly simple "margin of error of a proportion" for each race.
In the case of future discussions, would you provide me the source of your raw data and possibly the script you used? (I seem to recall you stated somewhere that you used R.)
Sources explained above. Excel was used. Margin of error of a proportion can either be calculated directly as z*SQRT(p*(1-p)/n) where z = 1.96 for 95CI or using an online calculator such as this
I do feel that there is sufficient information to suggest a conclusion, if not to confirm it.
YOU don't have enough to do so. You have no raw data and no methodology. You've not clearly stated what the hypothesis is, either. What, exactly, is "suggesting a conclusion"? You either can or cannot support a hypothesis with data.
But, if there is not enough evidence to suggest that elves are good, then there certainly has not been presented enough evidence that elves are bad.
Is anyone suggesting elves are "bad"? Aren't you the one making the claim that they are good? If so then you are the one who needs to support the claim.
The closest thing we have to an hypothesis from you is:
I think a preponderance of evidence suggests elf teams win more often than they lose compared to non-elf teams.
I am not saying I think elf teams, on average, are broken or overpowered, just better than the average of everything else.
We should try to compare apples with apples. Comparing a group of only T1 teams with a group which contains non-T1 teams is not that. To that end we should compare elven teams with non-elven T1 teams, so everything but Vamps, Slann, Pact, Ogres, Gobbos, Flings, Underworld and Khorne.
If we do that with, for example, the OCC data then we see the Elven Team win% has a 95CI range of 53.18 to 55.02 and the non-Elven Team win% has a 95CI range of 49.78 to 50.82. So yes, we can say to 95CI that Elven teams win more often than they lose compared with non-Elven teams in OCC. If we do the same with the FOL data (4 to 5 times the sample size), though, that same conclusion does not appear to hold with the Elven teams' range being 50.37 to 51.42 and the non-Elven teams' range being 50.37 to 50.85. I've not yet looked (and probably won't, tbh) at aggregating the data to see what happens overall, but it would seem that your hypothesis may be supported by the evidence in some places but not in others. There is not "a preponderance of evidence" though.