SilverFish
noHero123/silverfish
https://github.com/noHero123/silverfish/blob/master/HrtBddy/instructions.txt
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://github.com/noHero123/silverfish/blob/master/HrtBddy/DefaultRoutine_nH.zip
如何配置silverfish
https://hearthranger.fandom.com/wiki/Configure_Silverfish
Running
When starting the bot a second window appears with detailed infomation about the silver.dll task. Doublecheck the version at the first line of the output.
Silverfish now uses it's default settings and basic logic for mulligan and combo:
It tries to mulligan low mana cards and plays basic combos that have a high board value.
You can help Silverfish SE to understand your deck by adding custom deck settings.txt, _combo.txt, _discovery.txt and _mulligan.txt.
You'll find more infos about this later.
Configuration
Silverfish can be configured generally but it's highly recommendet to use different configuration files for each of your decks.
Therefore you have to create subfolders corresponding to the deck title within your HearthRanger\ExternBot\silverfish\Data\*CLASS* folder.
For example settings for your "Jade Aggro Shaman" should be in this folder:
HearthRanger\ExternBot\silverfish\Shaman\Jade Aggro Shaman
If it does not exists Silverfish tries basic Shaman configuration from
HearthRanger\ExternBot\silverfish\Shaman
That way you might have general settings for all of your shaman decks.
可以用职业的通用配置,以及针对特定卡组的配置
Deck Settings
The settings.txt should have at least the first two options.
Every missing option is read from settings.txt in your Silverfish directory.
Deck architype
behave = control;
Select the behavior of the A.I.:
- control - Used mostly, this focuses on boardcontroll, clearing the board as best it can before going face.
- rush - Will try to kill only the strong enemy minions on the board or when the oponent has more then 4 minions. But it will hit face more often then the previous 2
- face - This mode will just go face unless there is a taunt or a few other very specific threats.
- mana - Allmost the same as "Control" but tries to not "over"-damaging a minion to kill it (i.e. not killing a 1/1 with a 7/8) (This mode is in testing phase)
enfacehp = 15;
The amount of health the opponent needs to have before your hero attacks the enemy directly with a weapon. Examples for different deck architypes:
- Face: 30 - Straight to face no holding back
- Aggro: 29 - Allows for initial board control
- Tempo: 20 - Board control until stable
- Mid-Range: 15 - Full board control
- Control: 15 - Full board control
Play around...
playarround = false;
Should Silverfish try to play around AoE spells.
This affects the amount of your minions on the board and might also reduce your winrate. (Typo is also in the source code, don't try "playaround")
playaroundprob = 50;
Probability where the enemy plays the AoE spell but your minions are forced to survive
playaroundprob2 = 80;
Probability where the enemy plays the AoE spell and your minions can die
useSecretsPlayArround = false;
Try to play around enemy secrets. Like play around AoE this might also reduce your winrate since some plays might be to slow to win the game.
Calculations 
maxwide = 5000;
Out of all possbile boards this number boards is taken to the next level of calculations where another move may be added to them.
The more boards are calculated the higher the chance to make a really smart play.
On the other hand, this adds a lot a calculation and maybe your CPU might be overextended on complex boards with many minions.If roping happens to often, reduce the value.
On very high end computers values above 10.000 might be possible, 5.000 should work on most desktop pcs flawlessly.
twotsamount = 1000;
Number of boards where the first enemy turn is simulated. This should be around 20% of maxwide.
simEnemySecondTurn = true;
This sets whether Silverish should calculate the secend enemy turn as well.
That's a total of 4 turn calculated: your current turn, enemys turn, your possible next turn, enemy next turn.
enemyTurnMaxWide = 40;
Boards calculated in enemy first turn in first AI step. Around 10-20% of enemyTurnMaxWideSecondTime but lower or equal to enemySecondTurnMaxWide)
enemyTurnMaxWideSecondTime = 200;
Boards calculated in enemys-first-turn BUT in the second AI step (higher than enemyTurnMaxWide)
enemySecondTurnMaxWide = 40;
Number of enemy boards calculated in enemys second TURN. Around 10-20% of enemyTurnMaxWideSecondTime. (This might have been removed?)
nextTurnDeep = 6;
Maximum deep-lvls (total moves) in your second turn (recommended = dont touch this)
nextTurnMaxWide = 20;
Maximum boards calculated in one second turn deep-lvl
nextTurnTotalBoards = 200;
Maximum boards calculated in second turn simulation. Around 20% of twotsamount.
alpha = 50;
Weight in % of your 2nd turn vs your 1st (0<= alpha <= 100) eg. 30 = 30% 2nd turn result + 70% first turn result
simulatePlacement = true;
Simulate all placements for Direwolf Alpha, Flametongue Totem or Defender of Argus
Combo 
Discovery 
Mulligan 
General Settings 
useExternalProcess = true;
use silver.exe for calculations, a lot faster than turning it off (recommended = true)
concedeOnBadBoard = false;
Concede if the enemy board is too good. It's currently not recommended to use this options since it miscalculates certain situations and might give up when there is still hope.
concedeOnBoardValue = -900
concedeOnBadBoard will trigger if our best board is lower value than this (recommended = dont touch this)
speed = false;
Set this true to support sending multiple actions to bot (currently buggy so forced false)
useNetwork = false;
Use networking to communicate with silver.exe instead of a file (experimental, likely to fail after some time)
netAddress = 127.0.0.1;
network address where silver.exe is running (127.0.0.1 = this computer)
tcpPort = 14804;
tcp port to connect on
SilverFish的更多相关文章
- 用单元测试来调试SilverFish AI
[TestFixture] public class AiTest { [Test] public void Test() { Settings.Instance.LogFolderPath = @& ...
- 在HearthRanger中使用Silverfish
I'm new here and have never used this bot before. But a few days ago I tried it and I liked it :) I ...
- 用git difff 生成补丁
http://stackoverflow.com/questions/1191282/how-to-see-the-changes-between-two-commits-without-commit ...
- HearthBuddy中的class276中的地址对应
2019年09月的 intptr_0 = method_18("mono.dll"); intptr_31 = intptr_0 + 522030; intptr_28 = int ...
- HearthBuddy Ai 调试实战2 在使用海巨人的时候,少召唤了一个图腾(费用是对的)
问题 游戏面板 8是青玉之爪13是海巨人17是恐狼前锋 64是萨满 66是圣骑士63,99,46,是微型木乃伊[其中99和46都是2血3攻,63是2血1攻]57是鱼人木乃伊 微型木乃伊 "L ...
- HearthBuddy 召唤随从的问题
代码如下,在SilverFish\SilverFish\ai\Playfield.cs文件中 public void callKid(CardDB.Card c, int zonepos, bool ...
- HearthBuddy Ai调试实战1-->出牌的时候,少召唤了图腾就结束回合
期望通过ai的调试,来搞明白出牌的逻辑. 55是投火无面者63是恐狼前锋34是风怒36是自动漩涡打击装置13是空灵召唤者, "LocStringZhCn": "<b ...
- HearthBuddy炉石兄弟 格雷迈恩
getDecks(); 设置 private void getDecks() { Dictionary<string, int> tmpDeck = new Dictionary<s ...
- 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 Hearth ...
随机推荐
- Oracle学习笔记:rank、dense_rank、row_number、ntile等排序算法
在 oracle 中有很多函数可以实现排序的功能,但是不尽相同.下面一一解说. row_number函数 功能:可实现分组排序,为数据行添加序号,多用于分页查询. 语法:row_number() ov ...
- JS笛卡尔积算法与多重数组笛卡尔积实现方法示例
js 笛卡尔积算法的实现代码,据对象或者数组生成笛卡尔积,并介绍了一个javascript多重数组笛卡尔积的例子,以及java实现笛卡尔积的算法与实例代码. 一.javascript笛卡尔积算法代码 ...
- css优先级及其对应的权重
1.选择器的优先级 !important>内联选择器(style)>id选择器>类选择器 | 属性选择器 | 伪类选择器 > 元素选择器>通配符(*) 2.选择器的权重( ...
- Flask框架入门
Flask-基本入门 简介 flask被称为微型框架,只提供了一个强健的核心,其他功能全部通过扩展库来实现:也就是说可以根据项目需要量身打造.他适合入门学习以及高手研究. 组成:WSGI.模板引擎(J ...
- django 文件上传样例以及遇到的一些问题
使用django上传文件 主流有两种方法 from表单以及ajax,为了自由度高一点,选择了ajax来实现文件的上传 前端部分代码如下: 主要关注 一 有一个文件上传(type='file')的按钮, ...
- PCQQ - 发送自定义的XML卡片消息
效果: 原理: qq分享产生的xml卡片消息存储在qq内存中,可以在qq运行内存中搜索找到其xml源码,记录源码相应的内存地址,通过内存地址修改掉内存数据,再次转发这条分享的消息就会发现内容的变化. ...
- 【Java并发】线程通信
一.概述 1.1 什么是多线程之间通讯? 1.2 案例 代码实现 解决线程安全问题 二.等待通知机制 2.1 示例 2.2 wait与sleep区别 三.Lock锁 3.1 概述 3.2 等待/通知机 ...
- python实现数据的压缩与归档
楔子 尽管现代计算机系统的存储能力日益增长,但生成数据的增长是永无休止的. 无损(lossless)压缩算法以压缩或解压缩数据花费的时间来换取存储数据所需要的空间,以弥补存储能力的不足. Python ...
- Linux Centos虚拟机扩容
Linux Centos虚拟机扩容(/dev/mapper/centos-root) 1:.首先查看我们的根分区大小是多少 df -h 文件系统 类型 容量 已用 可用 已用% 挂载点``/dev/m ...
- TensorFlow指定GPU/CPU进行训练和输出devices信息
TensorFlow指定GPU/CPU进行训练和输出devices信息 1.在tensorflow代码中指定GPU/CPU进行训练 with tf.device('/gpu:0'): .... wit ...