Mike

Since the gatherer data is publically available there is nothing stopping us from importing that data as a baseline for the new system, right? Which means we can focus on...

> I would assume sending info to Firemind would also require an account though which might dissuade people from participating. While this would be the most convenient way for me...

> So maybe AlphaGo-like approach would work I've gone down this rabbit hole before :smile: I believe the main challenge here is encoding the board state in a way that...

This was most likely caused by my SSL certificate having expired on the 15th. It should be fixed now. The error message is a bit misleading because it just gives...

Thanks for getting back to me so quickly. > It could be the case that one of the other options may give a better result for this game as shown...

>I guess the equivalent in magic is to just search all the lines with the permanents currently in play with both sides attacking with everything each turn. This might enable...

You are of course correct that there are many blocks/attacks that are viable depending on the game state. I was just saying that we might get better results if we...

One more question about MCTS if you don't mind. To run the simulations the current game object is copied. This copy operation does not shuffle the libraries. Therefore the simulations...

> Not true. We don't shuffle the library, but we hide all the cards and treat them as a special UNKNOWN card during the node selection and expansion steps. Oh...

The deck is the same in both cases and is really just vanilla creatures and basic lands: https://www.firemind.ch/deck_lists/58009 I will start logging the players hand in full and also their...