Obviously, I've been neglectful of this; I don't know if it will get better.
And this isn't a puzzle, really, but something I've noticed about playing BMAIBagels -- Bagels has had a clever challenge on the buttonweavers site and I've been playing his bot a lot; it's fast and usually a pretty strong player, though I do think the algorithm it uses may not always pick the absolute best move.
one feature it gives is the bot's calculation of its probability of winning. It's usually a pretty good measure of how the round is going, though I've noticed it can be off by a few percent even in cases where there's only one move left (e.g. I have one attack left, and I'll win unless my 20-sided die rolls a 1 and it claims it has a 3% chance of winning, not a 5% chance]
But every now and then its figures are so far off, I wonder if there's not a programming error. For example: in the situatoin below, I can't lose -- if I capture both its dice, I win and if it's forced to swallow my poison I'll win -- yet it thinks it has a 92.5% chance of victory! What outcome is it ignoring, or what strategy is it misevaluating?
