ill get started on it

prob Christmas Paladin IMO

I don't think it'll ever happen. The game is insanely complex and there just aren't enough people working on it.

Even in the game Go, top human players are better than top engines.



I meant for BRM.

But yes possibly for TGT.

What it comes down to is: There is no way for a computer to read the player.

At the top level reading your opponents line of play is just SO important. Like oh he's got a threatening board, but won't over extend. He's trying to bait my brawl, maybe i'll hold it for another turn and see if he over extends.

A computer just see's 3+ minions and decides that it is brawl worthy.

that's not how hearthstone algorithms work... bots have been legend rank before lol

e: nvm im thinking of the kripp 10-11 arena win bots video, but still

This post was edited by Airwaves on Sep 9 2015 07:43pm

I disagree, I think reading is almost negligible if you have a computer than can accurately calculate the exact probability of any event or any card being in your opponent's hand or drawn. I think calculating exact probabilities is likely to be more accurate than a human's flawed probability calculations combined with reading, especially since it's an online game so there are no tells or anything like that.



Bots have gotten legend before, but those bots are extremely weak players compared to good human players. An actually near optimal bot would take millions of dollars and years of work to make (and would get Rank 1 Legend every season).

This post was edited by Voyaging on Sep 9 2015 10:06pm

No. The computer would have to have a running knowledge/history of all previous meta changes and trends (down to the hour), since things can change as quickly and simply as a streamer playing a deck. You do not need perfect probability to make good decisions, you need a much broader range of knowledge that cannot be programmed into a bot to operate the way a human would. There are tells in the game, although they are not perfect, neither are the tells in any other game, such as poker. For example, players mousing over cards, players showing that they have a targetable spell and then not using it, the order of the cards in hand and which ones are being held, and many of the things dr lat previously stated such as holding out for more value on a board clear

This post was edited by x]FoRsaKeN[x on Sep 9 2015 10:27pm

Programming some initial metagame into the bot and then having it adjust with each game it plays would be decent, though yes that is one area a human would likely be superior.

Unless the bot was really sophisticated and pulled data from the Web on decks to try to construct a metagame.

And yes you are right there are a few tells. Some of them can be programmed into the bot (say, if somebody targets a spell but doesn't cast it, the bot then knows that there must be a targetable spell in hand).

Still, the power of sheer perfect probability distribution calculations and perfect expected value play would trump all of these many times over. A bot that sophisticated though would be exceedingly difficult to make.

This post was edited by Voyaging on Sep 10 2015 12:10am

I'd think that it's basically impossible to make. Not difficult.



Yes it is. That is EXACTLY how the current hs bots work. I used to keep track of them hoping they'd become safe enough for me to use...

They didn't.

You can modify the bots "rules" and get much more complex with a list of conditions that would need to be met before you cast any spell, but the default rule list is basically as i said. "Once opps board reaches 3+ minions flamestrike/brawl/etc"

And bots only got legend with super easy to pilot decks. Shaman at one point was the most popular botting deck and probably the last botting deck to make legend consistently.

This post was edited by DrLatBC on Sep 10 2015 01:56am

I'm pretty sure the sea Giant shaman bots kept me out of legend many seasons ago. They were everywhere rank 1-3