• HearthBuddy炉石兄弟 如何调试ai


    Sepefeets's update to botmaker's Silverfish AI

    This AI is a Custom Class for Hearthranger and Hearthbuddy which intends to simulate all possible turn actions and select the best.

    Official Threads:

    build instructions:

    • If you plan to do any significant work I recommend using linkshellextension to create junctions instead of actual copies of the "ai" + "cards" + "Penalties"

    Hearthbuddy:

    • copy the "ai" + "cards" + "Penalties" folders to the "HrtBddy" folder, then create a project out of everything in the "HrtBddy" folder
    • add references for Hearthbuddy.exe and all the DLL's that come with it
    • go to Project -> Properties -> Application and set target framework to .NET Framework 4.5
    • go to Project -> Properties -> Build -> Advanced and set language version to C# 5.0 for HB compatibility
    • build it however you want just to test that it does build but HB uses the source files not a binary
    • there might be some errors I can't fix (I don't own HB), if you know how to fix it then let me know or submit a PR
    • build Silver.exe too

    Silver.exe:

    • copy the "ai" + "cards" + "Penalties" folders to the "external process" folder, then create a project out of everything in the "external process"
    • go to Project -> Properties -> Application and set target framework to .NET Framework 4.5
    • go to Project -> Properties -> Build -> Advanced and set language version to C# 5.0 for HB compatibility
    • build it as Release x86 for performance or Debug x86 for debugging, and copy the _cardDB.txt file to the binRelease and binDebug folders

    How to simulate boards with silver.exe:

    • create a "test.txt" file in the same folder as silver.exe
    • copy your current board, like
    #######################################################################
    start calculations, current time: 00:00:00:0000 V116.27 control 5000 face 15 twoturnsim 1000 ntss 6 16 160 playaround 50 80 ets 16 ets2 160 ents 16 secret
    #######################################################################
    mana 3/10
    emana 10
    own secretsCount: 0
    enemy secretsCount: 0 ;
    player:
    1 2 0 1 0 0 0 0 0
    0 0 0 0 0 0 0 0 0 0 0
    ownhero:
    priest 27 30 0 False False 4 True 0 False 0 0
    weapon: 0 0 unknown
    ability: True CS1h_001 3
    osecrets:
    enemyhero:
    warrior 15 30 35 False False 36
    weapon: 0 0 unknown
    ability: True AT_132_WARRIOR 0
    fatigue: 9 0 7 0
    OwnMinions:
    twilightguardian AT_017 zp:1 e:212 A:4 H:10 mH:10 rdy:False natt:0 ex ptt spllpwr(1)
    EnemyMinions:
    boombot GVG_110t zp:1 e:184 A:1 H:1 mH:1 rdy:True natt:0 ex
    Own Handcards:
    pos 1 arcaneintellect 3 entity 9 CS2_023 0 0
    Enemy cards: 4
    ownDiedMinions:
    enemyDiedMinions:
    og: 169,2;1152,1;248,2;246,2;251,1;557,2;1150,2;209,2;430,2;642,1;372,1;1010,2;371,1;701,2;
    eg: 995,1;708,1;429,2;740,2;495,1;514,1;648,2;333,1;241,2;516,2;1218,1;188,2;749,2;

    it will then calculate this board and print the first 100 best boards (sorted from best to worst) and simulate the whole turn of the best one.

    if you add "test" to the first line:

    #######################################################################test
    start calculations, current time: 00:00:00:0000 V116.27 control 5000 face 15 twoturnsim 1000 ntss 6 16 160 playaround 50 80 ets 16 ets2 160 ents 16 secret
    #######################################################################
    ...

    you can input the index of the displayed boards to simulate the whole turn of this board (indexes are be displayed in the list of the first 100 best boards)

    DefaultRoutine.cs文件来自  https://github.com/ChuckForkHearthBuddy/silverfish/blob/doomsayer%2BgetPlayfieldValue/HrtBddy/DefaultRoutine_nH.cs

    How to install:

    (- make a copy of the original defaultroutine-folder (but don save it in routines folder))

    - delete the DefaultRoutine folder in Hearthbuddy/Routines

    - extract the stuff inside the DefaultRoutine_nH.zip to that Hearthbuddy/Routines folder
    (if you did it correctly you will have a Hearthbuddy/Routines/DefaultRoutine/Silverfish - folder)

    https://www.thebuddyforum.com/threads/custom-deck-silverfish-ai.170312/

    ... also known as "Ultimate Ai"

    Hi, im Botmaker (reorder the chars in my username ;P) and i present you the (ultimate)
    Silverfish-AI
    (i will use "uai" as a abbreviations in most cases)

    This ai should be able to handle most decks, with an acceptable winrate.

    How it works:
    the ai simulates (almost) every possible combination of actions, which you are able to do in your current turn.
    it gave every simulated board a value, and chooses the moves which leads to this board.

    Download:
    https://github.com/noHero123/silverfish/blob/master/HrtBddy/Silverfish/Silverfish.zip?raw=truee

    or use https://github.com/noHero123/silverfish/trunk/HrtBddy for svn checkout

    Install instructions:
    Load the rar-file and copy the Silverfish-folder into the CustomDecks-folder

    follow this guide:
    http://i.imgur.com/yfhU9kT.png

    If everything works well, you should show the two new custom decks in you dropdown list :

    2014-07-11 08_38_40-EVA (884 726 835) - TeamViewer - Licence gratuite (seulement pour l'utilisat.png

    Then you can choose one of the two behaviors (Rush for rush-decks, Control for board-control decks)
    in your Hearthbuddy.

    Control will focus on clearing the board
    Rush will focus on attacking the enemy directly

    Note : This custom deck works for Hearthbuddy versions >= .82

    Source:
    https://github.com/noHero123/silverfish (the repo after the rewrite)
    https://github.com/noHero123/HRCustomClasses (for Hearthbuddy release, look at HrtBddy folder)

    Thanks to Hankerspace, who helped me to transfer this ai to hearthbuddy (i cant use hearthbuddy :D)

    Thanks to Hankerspace, who helped me to transfer this ai to hearthbuddy (i cant use hearthbuddy :D)

    Howto Report bugs/missbehaviors:
    Got to CustomDecks/Silverfish folder, there should be a folder-named "UltimateLogs", this is the place
    Where the battle-logs are saved (each fight is stored in another logfile (and a start/stop of the bot should also create a new one)),
    Each log has the timestamp of the battle-begin and contains all the data to resimulate your turn, so im able to tweak the ai.
    Please post them whether you report bugs (or i will ignore your bugreports). (but if you are thinking, that you found a hearthbuddy-bug, post the log from your hearthbuddy-window)

    Dont know whether Tracking or Druid-choice-cards are working! please test it, and if you are thinking that the ai did the wrong move please post the log.

    Sincerely
    Botmaker

    update v113.3:
    https://www.thebuddyforum.com/heart...custom-deck-silverfish-ai-28.html#post1667060

    update v112.4
    - some tweaks with aoe-spells
    - some other bugs

    update v111.4
    -some fixed (like better mana handling, give token-minions less value etc, full graveyard in log)

    update v111.3:
    - the use of the external process is turned off atm (next turn calculations are also turned off, to compensate the speedloss)
    - silver.exe is deleted from hrtbuddy-svn-folder + zip
    - some fixes

    update v111:
    - fixed some bugs
    - added Enemy's response to your second turn simulation (default on (because of the next point))
    - added external process (is used at default) will make the calculations more than 10 times faster!
    YOU HAVE TO START IT MANUALLY (just start the Silver.exe in the CustomDecksSilverfishData -folder and DONT CLOSE IT)
    (i cant start external programms with the .cs-versions (atm))

    update alpha 15:
    - fixed bonus for killing minions that spawn other minions on death
    - raised the value of every card from 1 to 5 (fixed the wasting of cards, whether they will buff a minion on your side )
    - will not simulate the second turn, whether the enemy has lethal on that board

    update alpha14:
    - fixed board-comparison (will fix multiple use of the hero-power in one turn and other stuff)

    update alpha13:
    -tweaked multiturn simulation (new default value is 256) and the amount of calculated second turn-boards is halved every recalculation :D

    update alpha12:
    -changed two turn simulation: it will now only simulate the 2. turn of the 500(default) best boards. change the number by editing the .cs file:
    change the number in the line:
    Ai.Instance.setTwoTurnSimulation(false, 500);
    (dont set it to high!!!!, set it to 0 to turn the second turn simulation off)

    update alpha11:
    -fixed shadow word pain
    -fixed traps

    update v110alpha10:
    - rewrote a very big part of the ai

    update v97:
    - should not waste upgrade
    - tweaked soulpriest
    - lethal detection tweaks
    - should not ignore bloodmage anymore
    - tweaked mukla
    - tweaked knifejuggler
    - tweaked placing
    - fixed manacalculations

    update v94:
    - fixed hero-power usage
    - new option in advanced settings:
    set the minimum probability for using the play-around-aoe-spells-feature:
    simulation uses the enemy aoe-spells only, if the probability to have this aoe-spell in hand is higher, than the given border (a value from 0-100) (default 40)
    -fixed placement with 4 minions

    update v93:
    fixed 2 turn simulation

    update v92 (delete the old files):
    - fixed weblord + spectral knight (did this a week ago, but was overwritten due to the new database ;_;)
    - should fix ancestral spirit
    - fixed windfury
    - fixed sylvanas
    - fixed buffing spells (and target nothing)
    - fixed iceblock detection
    - fixed blessing of might
    - fixed discard penalty

    update v91:
    -should work like charme

    update v81:
    -some fixes due to last update
    -very small enemy-turn-sim-speedup
    -will draw cards whether enemy has lethal

    update v80:
    -enemy turn simulation is a lil bit faster
    -fixed stealth-issues
    -little tweak in control-board-evaluation

    update v79:
    -some tweaks for warrior

    update v77:
    - updated _mulligan rules:
    - you will not attack with an sword of justice.
    - warrior should attack first, then armor up
    - should not spawn mobs, that spawn another mobs, whether these "kids" have no place on field
    - fixed return to hand (enemy draws now a card ;D (so ai knows that destroying > return to hand))

    Edit:

    if you want to change the two turn simulation you have to change the number of this line:
    Ai.Instance.setTwoTurnSimulation(false, 256);
    if you change it to
    Ai.Instance.setTwoTurnSimulation(false 0);
    the second turn simulation is turned off.-

    to play around some aoe-spells, change:

    bool playaround = false;
    to
    bool playaround = true;
    (bot might be to carefull)
    (they are only experimantal functions)      

  • 相关阅读:
    记第一次为开源代码报漏洞
    入职第三周——总结前两周学习内容
    入职一星期之感想
    毕业季之礼
    基于mint-ui的移动应用开发案例二(项目搭建)
    基于mint-ui的移动应用开发案例一(简介)
    工作笔记一——杂项
    微信小程序实战小小应用——豆瓣电影
    React学习之坑(二)- 基础入门
    React学习笔记(一)- 环境搭建
  • 原文地址:https://www.cnblogs.com/chucklu/p/11342828.html
Copyright © 2020-2023  润新知