Simulator buys a bunch of oracles; up to 3. Buys one envoy early, and possibly a second one late.
Default bots give a 50-40-10 advantage to envoy. Oracle play is imperfect, so reality is probably closer; if better oracle play makes a difference in, say, 5% of games, that already pushes the percentages to 45-45-10.
Fun fact - the advantage of envoy over oracle in the simulators is less than the first-player advantage. Oracle-BM as first player has the advantage over second-player Envoy-BM.
I'm not quite sure how to go 'beyond simulators' in this other than just playing a bunch of games.
I'm not sure envoy is any higher variance than Oracle. Because while with Envoy there's a lot of luck in where in the shuffle you draw your single envoy, with Oracle there's a lot more luck every time you play it - if the top of your deck is [estate, estate, silver, silver] that's way better than [silver, silver, estate, estate] if you have an oracle in hand, but it's equivalent if you have an envoy instead.