Are there any plugins, routines or profiles that look at what cards are being played together and build up a "this guy is playing XYZ netdeck" (talking neural net)? Might be interesting to see this used to help the bot make more intelligent plays instead of going all in on the Reno mage to get him down to 4 hp instead of waiting for a OTK - see this happen a bunch with hand lock as well.
It does track cards: In every UILogg you can find some lines similar to this: og: LOE_022,1;AT_071,1;CS2_106,1;OG_312,1;CFM_637,1;GAME_005,1; eg: CS2_004,1;AT_017,1;CS2_235,1;KAR_062,1;CS1_112,1; od: OG_314,1;OG_312,1;LOE_076,1;CFM_325,2;AT_071,1;NEW1_023,1;CS2_106,1;EX1_604,2;NEW1_011,2;AT_017,2;EX1_284,1;BRM_034,1;BRM_024,2;KAR_061,1; the format is cardid, count; cardid, count etc. og = own graveyard: cards that have been played and dead minions eg = enemy graveyard: same for enemy's cards od = own deck: remaining cards in the deck Based on that you can maybe tell what kind of deck you're facing, but it won't work every time: e.g pirate warrior and dragon warrior might have the same low cost minions(pirates) but are different decks. If you play an aggro deck, you usually don't play around reno anyway, either the have it and you lose or they don't and you win. Or they have reno, but can't deal with your board the same time and go back to 10hp immediately
Well what I'm seeing a lot with PLControl (playing control shaman), is the bot going all in with weapon/spells when it can only get the enemy to 2 or 3 hp - which basically guarantees they'll play Reno vs waiting another turn and possibly doing a OTK. I've been thinking a bit about what it would take to build a full on (off-line) simulation engine that could take in a list of the top net-decks and come up with a simulated "best deck". There are a lot of variables, obviously.
Turns out there are a couple of fairly detailed HS simulation projects (one done in C#) with extremely deep and fast analysis. Might take a look at one of them.
