Greetings. Here you can find updated (as long as I update them, though), Vision AI configs. All tests are done on EU, so expect slightly higher winrates on other regions, especially NA. If a deck is untested, you can usually tell how well it does by Casual concede games. In these games, most of the times the bot plays against other bots. If it cannot reach a 95% or higher winrate, it usually means that the deck probably can't reach Rank 5, especially on EU. Some decks, like Dragon Priest, might be an exception to this problem, but only if they are untested. Others might recover once I fix some misplays, but generally speaking, most misplays that do happen (like 2/3) are Hearthranger bugs, and if you encounter them, report them through the debug tool.
If you have bugs to report (in my configs, that is, report Vision AI bugs throught Hearthranger), please attach the replay file so I can see the match for myself - that is the painless way.
Under the spoiler, you can find out how:
First you want to find your Hearthranger folder. The folder where the Hearthranger version you used with my configs.
Then, you should see something like this:
Like the picture says, open your Review folder. Whenever you play a game, the bot records everything that happened in this file. By reviewing the bot's decisions, I can see where the config is failing and fix the logic. The bot can and usually will make stupid mistakes. For an example, recently, the bot couldn't even play Kabal Talonpriest. It blatantly ignored it. By looking at what the bot calculated at the moment, I can find whether the mistake lies in my config, or the bot. Regardless of what you used that night, Hearthranger can recalculate the situation as if it were playing the game right now, which helps me a lot.
After you open the folder, if you've played any games, you should see something like this:
Like the picture says, if you're not sure where the bot screwed up, just skim through the replays quickly. When you find a mistake, just copy the file with the same replay name somewhere.
Once you are done adding all files of the games where the bot screwed up, you can pack them as a rar, zip, 7z, whatever you like.
The last step is to upload them somewhere so I can download them. I use Mediafire usually, but things like Google Drive etc, are fine too, as long as I don't need to fill out a survey to download it.
Once you upload it, I should be able to fix it in a day or less. Until then, chill ;)
If you have suggestions on changes, or even your fixes, post them in the comments, preferably with winrates side by side, and I'll likely implement them.
If you want me to create a config for some deck, just ask, however I need you to test it because it's very likely I won't be able to; I don't nearly have the time to run the bot 24/7, nor do I have a full account - I do this for fun, to learn a bit. Also, I don't have all the cards :P
All the decks will be hosted on GitHub as that is the easiest way for me to manage. Sharing and remixing is, of course, allowed. This is not my property and you're free to improve on the configs, even claim them as your own. I don't really care :)
Lastly, I would recommend you avoid running this on weaker machines. The reason is a lot of the time a weaker machine will make crucial mistakes like miss lethal. For an example, I've run this on my i3 laptop with 8 GB of RAM. In around 200 games I've seen 10 game changing misplays that don't happen on my i5 8GB RAM desktop. Even if it is just 5%, 5% in terms of winrate is A LOT.
Post Nerf decks
DRUID
Standard Tempo Druid
Winrate (Casual Concede): 100%
Maximum Rank (Season 47): Untested
Winrate (Ranked): Untested
Note: Seems promising, testing in Ranked at the moment. Average Game Time : 6.8 minutes.
PRIEST
Standard Spiteful Priest
Winrate (Casual Concede): 86%
Maximum Rank (Season 47): Untested
Winrate (Ranked): Untested
Note: Seems to have a lot of misplays, need to fix. Average game time: 7.6 minutes.
PALADIN
Standard Aggro Paladin
Winrate (Casual Concede): 97%
Maximum Rank (Season 47): 8
Winrate (Ranked): 49%
Note: New addition, should be fairly similar in performance to Murloc Pally, if not better. Average game time: Untested.
Standard Murloc Paladin
Winrate (Casual Concede): 88%
Maximum Rank (Season 47): 6 (floating around 8)
Winrate (Ranked): 50%
Note: New deck, mostly focused on early game now. Average game time: 7.3 minutes.
HUNTER
Standard Aggro Hunter
Winrate (Casual Concede): 96%
Maximum Rank (Season 47): Untested
Winrate (Ranked): Untested
Note: Almost new deck. Focuses on Tundra Rhino combos, heavy hitting. Very fast compared to others. Average Game Time: 6.2 minutes.
MAGE
Standard Secret Mage
Winrate (Casual Concede): 88%
Maximum Rank (Season 47): 4
Winrate (Ranked): 52%
Note: Slight changes, Mulligan rework. Mostly same deck, minus Creepers, so more consistent. Average game time: Untested.
WARLOCK
Standard Cubelock
Winrate (Casual Concede): 94%
Maximum Rank (Season 47): Untested
Winrate (Ranked): Untested
Note: New addition. Some bugs to be fixed, but performs as expected. It's very slow so it'll probably never be viable to reach high ranks, only to play around on high ranks. Average Game Time: 9.5 minutes.
Standard Zoolock
Winrate (Casual Concede): 100%
Maximum Rank (Season 47): 8
Winrate (Ranked): 52%
Note: Should still be the best deck. Average Game Time: Untested.
Legacy decks (for daily wins and quests I guess?)
Standard Pirate Warrior
Winrate (Casual Concede): 85%
Maximum Rank (Season 46): 15
Winrate (Ranked): 25%
Note: This performed so poorly I had to take it out. Will return later when I have the time to see why it doesn't work that well (it's not a matter of misplays).
Standard Tempo Rogue
Winrate (Casual Concede): 93%
Maximum Rank (Season 46): 13
Winrate (Ranked): 38%
Note: Needs a lot of tweaks apparently. Sample size: 86 games. Average Game Time: 7.4 minutes.
Standard Token Shaman
Winrate (Casual Concede): 96%
Maximum Rank (Season 46): 13
Winrate (Ranked): 38%
Note: Yeah this deck sucks LOL. Will see if there are any misplays, but I watched it and it gets slaughtered by just about anything. Sample size 70 games. Average Game Time: 9 minutes.
CHANGELOG
Update v2.20180210.0 - Added post-nerf decks. Removed
Aggro Druid and added
Tempo Druid, renamed
Dragon Priest to
Spiteful Priest, added
Aggro Paladin, added
Cubelock, updated all non-legacy decks.
Update v2.20180217.0 - Fixed
Spiteful Priest.
Changelog:
Changed Mulligan rules a bit.
Fixed a misplay where Grand Archivist wouldn't get the bonus.
Fixed a misplay where the bot would sometimes play Twilight Acolyte on enemy minions with less than 2 Attack.
Reported a bug where bot would prioritize playing Northshire Cleric on turn 2 instead of a Netherspite Historian while holding a Dragon.
Reported a bug where bot would sometimes prioritize playing off-curve minion with better stats instead of on-curve if the off-curve minion can be played next turn better.
Reported a bug where Twilight Acolyte would sometimes cast on your own minions despite there being better enemy minions.
Reported a general bug where bot would ignore Arcane Artificer.
Reported a bug where bot played Netherspite Historian after playing a dragon.
Reported a bug where Charged Devilsaur wouldn't attack face even though it was from Free From Amber.
Update v2.20180217.1 - Fixed
Tempo Druid.
Changelog:
Adjusted Mulligan rules.
Added Greedy Sprite bonus when attacking other minions so it dies faster.
Reported a bug where bot would cast Anti-Magic Shell before playing all minions for that turn.
Reported a bug where the bot heavily prioritized playing Spiteful Summoner despite there not being any spells in the deck.
Reported a bug where bot didn't consider playing Ultimate Infestation.
Reported a bug where the bot didn't value playing Cobalt Scalebane at all when there were no minions for it to trigger its effect to.