Quote Originally Posted by aethernox View Post
You can't statisically account for individual player performance. There exists no relevant metric. Additionally, win% differentials between champion matchups almost never exceed ~10%. Lolking.net's charts provide relatively small sample sizes, but here's an example. In top lane matchups, over the course of 1062 games, the Tryndamere/Shen matchup comes to 54.8%/45.2%, and that's one of the most one-sided matchups shown.

Your questions don't actually ask anything of value, in my opinion. "Would this work?" What does that even mean? Needlessly vague and difficult to answer. Yes, it would almost certainly pick teams capable of winning. That distinction, however, is effectively meaningless considering that the vast ((vast)estimation: >99%) of team compositions have a win% of >0%. Would it cause a game to be "lost or won with the first ban?" No, it wouldn't. That's not how statistics work.

The second question doesn't make very much sense; do you mean contemporary games? There's no way to answer that without access to the impossible-to-acquire data in the first place. How can we make a comparison between known data and something entirely unknown?

You've laid out an entirely unfeasible premise with absurd expectations and ask questions that are inherently impossible to answer.
If you have data on individual player skill, there is absolutely no reason why statistics cannot account for it.

Now ........ seriously. I've stated time and again that the statistics are uninteresting. Totally. And that it's a thought experiment.

For what it's worth, we can keep going round and round about things that I've specified are beside the point - it just seems somewhat pointless.

The input I'm curious about is - as previously stated - if you had a tool for picking the best statistical counter for any given situation, taking into account your own team composition, would you then:

Get better teams using this tool than not using it?
Would these teams be capable of winning (by a greater margin than otherwise)?

Neither of those questions are impossible to answer - but anyone who thinks so is welcome to refrain from trying.

I was curious because, as it appears here and elsewhere, some people put a lot of stock in statistics - like the one mentioned for Shen/Trynda. Personally, I consider the statistics to be very close to nonsensical in a practical application - not withstanding the statistical accuracy.

It's .... psyco-history, or related anyways, for anyone who gets the reference. Isaac Asimov.

Nevermind. Thanks for your insigths all - even if they kinda didn't give me quite what I was looking for =)