Thanks for the reply! Yeah, this was a fun project and took a decent bit of work.
A few responses:
The sample is much bigger than that. The level of observation isn't really my account, it's each battle (using my account's deck). So it's more like 350+ samples per bot. Which is a really nice sample size for this kind of work.
Yeah, I have more value in Rebellion than most accounts, but not more cards. I only have about 30 Rebellion cards. The value is there because they are GF cards for conflicts. And a large part of doing this experiment now was to see how quickly the bots adapted to new cards being available. I did slightly handicap XBOT by having it first and giving it less time to learn. Restarting XBOT yesterday I'm already seeing more Rebellion usage.
Yup, that's exactly the advice I gave - do this for your own account (because you're right that even though the sample size is large, it's only large for my account). So hopefully people get that takeaway.
RE: The Epic conclusions: XBOT vs. Archmage in the era of rebellion