Drafting can be massive when deciding the outcome of games even at the lower skill levels. Opening up the entire hero pool just means you're largely evaluating the ability to draft in the current patch more than actual playing ability.
It would not surprise me if the OpenAI 5 would have lost against OG had the entire heropool been available, had the series been long enough or had the prizepool been big enough for OG to take the game seriously enough to warrant picking a splitpushing strategy (which is considered cowardly in some circles).
Tinker is an example of a hero that took (and still takes) a lot of gold to useful. If you're drafting tinker to stop early game push strats you're going to have a bad time. If anything Tinker is actually an example of an hero that the AI would probably excel at, the same goes for any micro dependent hero.
In general including heroes that excel with great mechanics is a poor way to evaluate how well the AI actually plays the game. No one doubts an AI can send inputs faster than humans.
the game is balanced around the entire roster. in a pool of 17 heroes, the answers to the locally optimal strategy likely don't even exist within said subset. drafting proper compositions, adapting to the opponent's heroes, and dynamically changing your gameplan during the match are all part of actual playing ability.
No, the rebuttal is how can someone who has never played the game before claim x,y,z is important or not?