Talk:Lee Sin/@comment-24496127-20140716033238/@comment-6281696-20140718221226

The point U N was making is that X might have a higher average winrate, but might only win certain matchups, while Y consistently wins 60%. So basicly the following:

X wins 100% against Z, X wins 40 against Y. X plays against Z 5 times, X plays against Y once. Average winrate: 90%, (relative) standard deviation.

Y just wins 60% of the time against any opponent. Average winrate: 60%, however, increadibly small standard deviation.

This makes Y a much better pick as he is much more reliable, whereas X is only a good pick in very specific situations and awefull in every other situation. But since the situations in which X is used are heavily biased in his favor his average performance is skewed and not a good measurement. Basicly, X is a very specific counterpick, thus making it an overal weaker pick. Outside of that specific counterpick, Y is better. If you take into account stuff like standard deviation than game theoreticly safer IS indeed better. If you just care about average performance than whichever can manage to reach the best average performance is better, however even then you'd have to take into account things that might bias it such as X being used less/more, or certain situations being more or less common.