Drafting a Limited deck on Magic Arena, the online interface for Magic: The Gathering, is quite an impersonal task. Unlike in physical drafts, on Arena players contend with people remotely. The bright side is that while you are behind the ultimate poker face (that is to say, no face at all), you can also run simulated draft pools online outside of Magic Arena with such interfaces as Draftsim. Therein, you can draft against the website's own Draftsimbot, an artificial intelligence geared specifically towards simulating a realistic drafting experience, so as to give users an authentic example of what drafting might look like.
However, there are many different bots that Draftsim has in their arsenal, and to see which one can give players the most authentic Magic: The Gathering Draft experience, Draftsim pit five different bots against each other.
The results are interesting, to be certain, but not surprising. RandomBot, a simple simulator bot that just takes cards from packs at random, and RaredraftBot, a bot that takes the rarest card from each pack (and in the case of a tie takes whatever they have more of on-color), did much worse in Draftsim's "bot battle" than the other three. DraftsimBot, of the remaining three, did the worst, while the NNetBot, a neural network-based bot, did the best. This was all to be expected, as DraftsimBot was basically the average in this neat and complex experiment, and neural networks tend to perform better as far as artificial intelligence goes, according to Draftsim's report on the experiment.
In the end, these experiments only help get AI closer to understanding complex systems such as Magic: The Gathering. In fact, this experiment was even featured on DailyMTG, Wizards of the Coast's official news hub for the game, on Monday, August 16th.
What do you think about this experiment? If you are looking for a more in-depth explanation you can visit Draftsim's website by clicking here. In the meantime, let us know your thoughts in the comments below!