Just wondering but how well does the bot interact with tracking?

Will this line interact as I intend it?

discard;hunter;rogue;LOE_046:1:AT_058+LOE_046

I am trying to get it to discard 1 LOE_046 if the hand is LOE_046+LOE_046+AT_058

Would discard;hunter;rogue;LOE_046:1:AT_058+LOE_046 or discard;hunter;rogue;LOE_046:1:AT_058+LOE_046+LOE_046 be correct?

Plays Muster for Battle while I still had a better weapon available. Should have attacked with weapon first.

end of log

Was able too get 66% with Midrange Hunter and on V121.4 Last night in Ranked



Running Face Shaman the past few days on versions prior too V121.4 I was getting similar sub 50% Results like the others, but Midrange Hunter worked well last night, not sure if was the deck or V121.4 that was the difference, will do more testing once season Restarts

@parakeet - Sorry I was extremely tired and misunderstood. I'll double check and add those too.

@Svenek - From my own logs (dark peddler picks it a lot) it shows that the zerus card stays zerus in your hand every turn and I'm not sure if the data about his changes is even available in the data SF has. It will take some time to investigate and since it's generally a bad card it will be a while... Improving _discovery.txt is a high priority though and will let you avoid picking it if that's how you're getting it.

@sadpwner - Tracking is just another discover card as far as SF cares.

I can't really spend time figuring out how to fix your mulligans but maybe ask in the rank 5 thread or the mulligan faq for others to help. One of my plans is to do a lot of work on the mulligan code to make it accept both forms of mulligans (botmaker's and HB's), actually check the rules to make sure they are valid and give errors when they aren't, and provide a way to test mulligans similar to testing boards with test.txt.

@Dre - I don't know, there's plenty of other people that are having good results with a variety of decks. I don't fix win rates, I fix misplays.

I know but not (anymore with zoolock and faceshaman) but its ok, just wanted to know if u perhaps found something, gonna test 121.4 tonight with faceshaman if it's still worse then 121.0 i Will leave face shaman and focus on another deck for next season

It's a hassle, but you can check for increased roping in the logs by looking to see if all of the planned actions for the turn were carried out.

V121.1SE changes included:

some additional database lookups. These could marginally increase calculation time but realistically it would not increase roping. I expect these improved winrate.

+ && !(hasJuggler && (card.type == CardDB.cardtype.MOB || this.summonMinionSpellsDatabase.ContainsKey(card.name)) && p.enemyMinions.Count > 0)
+ && !(hasCouncilman && (card.type == CardDB.cardtype.MOB || this.summonMinionSpellsDatabase.ContainsKey(card.name)))


Tunnel Trogg was made a priority target. On par with flamewaker but less than manatide and flametongue. This represents a minor play-style change and should not have impacted the winrate drastically. If anything I would expect an improvement. You can check the impact of this by seeing if the winrate versus shaman decks went way down.

+ priorityTargets.Add(CardDB.cardName.tunneltrogg, 5);


Large penalties were added for playing Tunnel Trogg and Overload improperly or for playing knife juggler/darkshire and minions improperly. Potentially, if the simulations only included the 'incorrect' play order, and it now is assigned a large penalty, the bot might make sub par moves. For example, without the penalty, playing tunnel trogg and overload in the wrong order might still be better than any of the simulated alternatives - but with the penalty, it is worse. If the bot did not simulate the correct play order, because the number of simulations in settings.txt is not high enough, then the bot may not find the correct play order. Sepefeet might be able to comment on how realistic this scenario is. If this code is the cause of the terrible winrates, only decks with tunneltrogg, knife juggler, or councilman should be impacted.

if (card.type == CardDB.cardtype.MOB || this.summonMinionSpellsDatabase.ContainsKey(card.name)) pen += 20;
+ }
+
+ if (a.own.name == CardDB.cardName.tunneltrogg)
+ {
+ if (card.overload > 0 && card.name != CardDB.cardName.elementaldestruction) pen += 20 * card.overload;
}
}
+ if (a.actionType == actionEnum.playcard)
+ {
+ if (a.card.card.overload > 0 && a.card.card.name != CardDB.cardName.elementaldestruction && card.name == CardDB.cardName.tunneltrogg)
+ {
+ pen += 25 * a.card.card.overload;
+ }

+ if (card.name == CardDB.cardName.knifejuggler && (a.card.card.type == CardDB.cardtype.MOB || this.summonMinionSpellsDatabase.ContainsKey(a.card.card.name))) //prioritize jugglers 1st
+ {
+ if (a.card.card.name != CardDB.cardName.knifejuggler) pen += 20;
+ }
+
+ if (card.name == CardDB.cardName.darkshirecouncilman && (a.card.card.type == CardDB.cardtype.MOB || this.summonMinionSpellsDatabase.ContainsKey(a.card.card.name))) //and councilman 2nd
+ {
+ if (a.card.card.name != CardDB.cardName.knifejuggler && a.card.card.name != CardDB.cardName.darkshirecouncilman) pen += 20;
+ }
+ }
}

However, Midrange Shaman is back at the Rank 4-5 border using 121.4SE and 5.9.

Thanks parakreet,

The face deck has trogg where mid-shaman has mostly not - but i didnt saw much misplays with trogg in the logs. I will look at face shaman again later - got rank 5 with it this season on 121.0.

This new season i will first focus on perfecting the mid-range hunter because i think this is one of the most easiest decks to play for the bot (no overload / life tap or other big thinking decisions) just hit it hard.

After that any requested deck i will take a look at.

I tested the discard rule by making a huge list of cards to discard (pretty much my entire deck). The list was filled with only discard rules so there were no hold rules. It kept every card I told it to discard.

A PM I just sent:

I'll take any more suggestions on what you want to be able to do that you can't currently but don't worry about the specifics of how to format it.

Can you write a tutorial on how to fix misplays? I think development would go much faster if multiple people were working on it. It would also free up time for you to work on the mulligan and discovery.

Damn, apparently,
hold;all;all;CS2_236/BRM_017 isn't supported

I think a mix of discard and hold are supported. You just can't place conditions on the discard making useful to adjust all commands only.


I think the utter most important feature for mulligans is priorities. The use of discard rules can be completely avoided with priorities. Alternatively, discard rules can be used to create priorities if conditions were applicable to the discard rules.

The "if not" feature Dre mentioned above can be done with discard rules, so I guess what we really need is the ability to place conditions on discard.

One example for Dre would be discard;all;all;Knife jungler:1:2 mana card A/2 mana card B.... 2 mana card N

This is essentially the "if not" rule.

___

Could you make it so the bot saves mortal coil for exactly soul priest if it is able to clear off the current minions without coil?

_

Also, could you add a keyword for discovery for Reliquary Seeker so I can give it different values based on whether it can be activated or not?
__

In the future, can we see some way to specify exact plays for the bot based on the information it records rather than using general rules?

__

Finally, what are you working on for discovery? Discovery seems mostly fine as it is currently.