I'm a computer scientist who's dabbled in using AI (neural-networks, specifically) to play a few games and I might be able to provide some insight into this.īots are generally really great when you have one of two things (ideally both).Ī huuuuuuge corpus of history of very high-level play, which it can mine either for statistical trends, or mine using machine-learning in order to build a model of what 'very high-level' looks like, or an environment where bots can just play each other, and learn what good looks like by starting out badly, and slowly improving against each other, learning from their past games. There's what, maybe 3 of a particular common in a given pool? In MTGa drafting you could well face all 3 Secretkeepers and Didnt Say Pleases, while in MTGO you are only facing less than 0.5 per deck. Human drafts eliminate this kind of exploits, 100% of the drafted cards end up in the matchmaking pool, as opposed to the 12.5% on bot. In person draft had a very balanced field, whenever I had a league draft on MTGO I never had a single run where I faced the same archetype twice, while on MTGA its rare that at least half the matchups are not the "meta" deck. ELD MTGA draft had basically 2 metas, monoblue mill and then Naya Aggro. I could paint a good picture of whats happening around me on MTGO and have gambled (and won) wheeling some of the hybrids. Signals are much clearer with humans, this was abundantly clear in ELD no matter which iteration of the bot. I will use ELd draft as my example because my gaming lappie broke through most of the THB season.
That doesn't matter much - it's not like it's an amazing card - but it adds to the variety of the format to get to play with more cards. If you weren't rare-drafting, you didn't get to play with Mizzium Tank in WAR. In THB on MTGO, drafting multicolour decks was somewhat popular, but that wasn't really a thing on MTGA.Īnother difference is that you'll see the lower tiers of rares in play more often. I checked, and I have one in my collection after 60ish THB drafts.īecause the bots take practically every rare they see, you get fewer interesting decisions when you're passed a strong rare which isn't in your colours, and fewer chances to draft powerful multicolour mess decks. Can anyone remember seeing a Captivating Unicorn in play in a THB draft? It's a perfectly playable card, but the bots seem to like it more than that. This isn't obvious unless they get something drastically wrong, like Merfolk Secretkeeper in ELD or the gate payoffs in RNA, but it does reduce variety in the draft. The bot pick orders seem to be fairly static, so you don't see many of the commons they like, and you see a lot of the commons they don't.
One big difference is that there's more variety with human drafts.