Originally Posted by
aethernox
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.