Esta en ingles, no lo voy a traducir ahora, pero básicamente explica como funcionan los bots y los fallos que cometen, ademas explica que hay otra versión de bot que lo que hace es asesorar al jugador sobre cual es la mejor jugada después de evaluar cualquier jugada posible ese turno. Parece que aprenden con el tiempo y que cada vez serán mejores a medida que tengan bases de datos y porcentajes de victorias etc.
Así que prepararos, si esta season hay mil y juegan bastante bien, la season que viene se clasifican para la blizzcon.
Trozo de texto sacado de hearthpwn:
they basically work like chess engines. they look at the board position and assign values to the different creatures on the board and to the life totals and to the number of cards in hand.
when deciding what to do, the bot will try every possible move (often thats 30 options or more), generate all the board positions that can be reached with any of the possible moves, and then they assign values to all these positions. the move that leads to the board position that has the highest value is then executed. basically a bot can play a virtual game of hearthstone against himself, and uses this in a brute force approach to find the strategy that leads to the best board position, and then makes the moves that lead to that ideal position. instead of calculating just one move, they may calculate seval moves into the future, and then the number of possible positions that can be reached grows exponentially. there are limitations because you dont know what the opponent can do, you can only make assumptions based on the class you play against. so the bot might calculate its own move with precision, then in the opponents turn it is assumed that the opponent has a chance to do AOE, a chance to play a minion without taunt, a chance to play a minion with taunt, a chance to do spot removal, and a chance to do nothing, and the second turn of the bot can then again be calculated with precision based on these assumptions.
the strength of a bot is determined by how accurate the internal version of hearthstone is, the one that the bot plays against himself internally, and by the accuracy of the evaluation of the board positions. for example, a low-quality bot may treat Nat Pagle as a 0/4, ignoring the chance to generate card draw for the opponent. and a high quality bot will assign a value for the opponent whenever Nat Pagle survives unsilenced into the opponents next turn, a value equal to the value of 50% of a card draw. now that bot will try to kill or silence Nat Pagle if there is nothing better to do, and will ignore nat pagle if he can go for lethal or can remove an even bigger threat you played.
knowing the bots might be helpful, if you know them with such precision that you can abuse certain mistakes they always make. for example, when you play Starving Buzzard and Snake Trap, good players will play around it and will deny you these 3 extra cards. if they attack you and play a small minion and the trap doesnt trigger, it must be snake trap, which would be trggered by attacking one of your minions. so they will use a card to remove or silence the buzzard, and then attack your minions when that synergy isnt there. a bot might always assume Explosive Trap, always attack the buzzard, and always give you the 3 extra cards.
but just like in chess, there may be a bot or engine that has no obvious flaws that can easily be abused. in chess, some engines have certain flaws, like for example they might incorrectly evaluate a certain pawn structure, which causes the bot to allow you to go after a certain advantageous pawn structure in certain situations where it would be better to deny you that pawn structure. but then there are better engines that are borderline genious when it comes to correctly evaluating pawn structure, and then it doesnt work anymore. in chess, there are tournaments for engines, and chess engines are unbeatable for humans when normal time control is used. they can only be beaten by other engines, or by grandmaster-level humans with unlimited thinking time, or by humans with bot-assistance.
i think just like in chess, the most powerful possible player will be the hydra. thats an engine-assisted human player. basically you use a bot, and the bot suggests the best moves he can find, and then you add your own knowledge, consider things that a bot might miss, and you make the move. so you are in charge, but the bot prevents you from making any obvious mistakes, and the bot may point out promising options that you may have missed, greatly improving your play. in chess, in the cases where this is not allowed, its called "electronic doping".
