Well, that statement was made to be a bizarre bastard love child between the binomial distribution and the one-proportion hypothesis test, except much less reliable than either because of the large number of mitigating factors. You'd have a tough time convincing me a person would have a good statistical reason to be unhappy their vote has been for the winner more than 25% of the time, though.
Actually, there is a good statistical reason the average voter should win WAY more than 25% of the time.
Lets use the following matches as examples:
Match 8
Chen 82
Kotohime 16
Reimu Hakurei 44
Ghost (PC-98) 8
Total Votes: 150
Match 16
Koakuma 81
Yumeko 20
Su-san 10
Unzan 18
Total Votes: 129
In match 8, 82 voters won out of 150. Thus, the average voter had a probability of 82/150 of winning.
In match 16, 81 voters won out of 129. The average voter had a probability of 81/129 of winning.
Probability of winning for a given match, using this as a sample: (82+81)/(150+129) = 163/279 = .58
Course, this sample isn't significant, so lets use a larger sample - all completed matches.
Total votes cast across first 16 matches = 2207
Total votes cast for a winning character = 1086
Average voter wins with a probability of .492, or 49.2% of the time.
Using
this applet calculate the binomial, only 4.4% of voters should have lost 12 matches or more.
And with 120 to 180 vote matches, it means there are about 6 or 7 people who have every right to be annoyed, as they fall into the bottom 5%.
(Edit: sorry for some of the edits, I got a couple numbers backwards when I typed this.)
