Finding memory values?

[Maddin1803]

    public class Log
    {
        public static void debug(string msg)
        {
            log(msg);
        }
        public static void log(string msg)
        {
            Console.WriteLine(msg);
        }
        public static void say(string msg)
        {
            UIStatus.Get().AddInfo( msg );
            log(msg);
        }
    }

So it wraps Console.Writeline
Which drops it output into
Hearthstone/HearthstoneData/output_log.txt.

[overture2112]

Yea, it's pretty bare-bones. Also, note that UIStatus.AddInfo is a bit buggy and doesn't handle queuing messages well (might be solvable by having it wait a bit before resuming execution).

[TheEnglishGuy]

Is it possible to stop the 'stop gaining experience from automated loosing.' If yes, this would make leveling really really easy, as you can already concede at the moment you found an opponent.

[overture2112]

Is it possible to stop the 'stop gaining experience from automated loosing.' If yes, this would make leveling really really easy, as you can already concede at the moment you found an opponent.

Unlikely, but you can probably code a bot smart enough to beat really bad players once it tanks it's MMR (and maybe you give it a good deck).

[Maddin1803]

Back when i first played around, it was enough to just play cards, didnt even need to attack to circumvent the anti xp code.

[TheEnglishGuy]

Back when i first played around, it was enough to just play cards, didnt even need to attack to circumvent the anti xp code.

So you saying that if you play a game, play cards, but don't attack so you lose very fast, you will get max/normal xp anyways? Sounds unlikely to me, but still, it would take quiet long before the game is even ended. Of course less long than if you play it normal but still. Now, I am trying to find a way to make use of the quick/small DC = lost game problem in the game. Does anyone know if the info of the enemy you playing against is saved in your files? (or anything that has to do with a game, so every time you play a new game new info about that specific game will be stored?, as it was in League of Legends, It might also be in hearthstone)

[Maddin1803]

Dont see the Point in that atm
The understanding of the client is getting better every day, building a basic ai which is capable of winning against the blizzard ai shouldnt be terrible complicated.

[TheEnglishGuy]

Dont see the Point in that atm
The understanding of the client is getting better every day, building a basic ai which is capable of winning against the blizzard ai shouldnt be terrible complicated.

True, very true indeed.

Do you know what the percentage of the XP gain difference is of winning a game in AL mode vs Real Opponents?

[phyte]

Is anyone interested in forming a HS controls API? I get the feeling that a lot of people in this thread are going off and implementing the same stuff at the same time.

[Maddin1803]

@phyte sadly i simply dont have enough time to do much aside of giving suggestions :O hopefully this will change in the near future.

but i would suggest, fork overtures repo and push example plugins, a nicley wraped api is convinient but not realy needed at this stage. The HS Code is good structure itself allready. When we are able to do any legit move without big bugs, wrapping everything into an easy to use api should be quiet trivial.

@theEnglishGuy

Even a technicly stupid ai can do some crazy results, its easy to get the BEST possible move for one round and the reaction of the enemy as long as you dont try to calculated possible future turns. Even if this isnt anything advance, it still should yield quiet good results.
But as you have plenty of ressources, even by far more complicated ai's are possible. (hint hint , i would suggest anyone to play around with this, designing ai algos is fun ! And sometimes its hilarious to watch when it does realy clever or realy dumb things :P )

[overture2112]

but i would suggest, fork overtures repo and push example plugins...When we are able to do any legit move without big bugs, wrapping everything into an easy to use api should be quiet trivial.

Exactly. With the framework I posted it should be trivial for anyone to start messing around and work on improving the functions to drop and play all types of cards. Once we can do that without crashing, we can make a library that exposes that functionality and people can start working on plugins to make intelligent bots.

@theEnglishGuy

Even a technicly stupid ai can do some crazy results

Yep. For kicks, I made a "dumb" deck (no spells, no battlecries, just plain minions). Running the TestPlugin in the framework I posted (play all minions in arbitrary order if it can afford the mana, and each minion in the field attacks the first valid target), it was beating the beginner mage blizzard AI until it crashed around turn 8 or 9. That's probably entirely due to the deck's mana curve. A bit smarter target selection and it could probably beat the beginner AI very reliably.

In theory, if you just want xp, you just need to win every once in awhile in order to not trigger the "massive loss" detection that blizzard uses to stop giving xp rewards. So even a 10% winrate is probably good enough for grinding xp.

[overture2112]

Do you know what the percentage of the XP gain difference is of winning a game in AL mode vs Real Opponents?

Wiki says:

Players can gain XP in Practice mode, up to max level. Above level 10 extra XP is gained from played games against human opponents; however, given the relative speed of most Practice mode games, this may still be considered a quicker way of levelling up heroes.

Practice mode - Hearthstone: Heroes of Warcraft Wiki

[overture2112]

I toyed around with making the code for dropping minions and attacking a bit closer to what the client does and sprinkled logging and delays around so I could tail the log more easily in order to figure out what's going on while watching it play instead of just postmortem. Available here: https://github.com/sigmundnetski/Sig...ugin/Plugin.cs

This reduces the number of failed DoNetworkResponse calls to zero. It also reduces the crash rate greatly; I've watched the bot complete games and get near the end of a 2nd before crashing now, with the average usually being 8 turns.


I believe there's only one more cause of crashing and it occurs in DoNetworkResponse when picking up a minion to attack (ie. inside the first DoNetworkReponse in DoAttack).

There's also occasionally this warning: "PhysicsSDKL NpBoxShape::setGlobalPose: WriteLock is still acquired. Procedure call skipped to avoid a deadlock!"


My guess is when you pick up a minion and it goes to highlight potential targets, a race occurs with the main thread (as it might be rendering stuff too at that moment).

It sounds like we need to either get the plugins running on the main thread or figure out a way to add a lock on anything using the unity3d engine.

[youknowjack]

... It sounds like we need to either get the plugins running on the main thread or figure out a way to add a lock on anything using the unity3d engine.

Just hook an update method that runs every frame to call your code (e.g. InputManager.Update)

[overture2112]

Just hook an update method that runs every frame to call your code (e.g. InputManager.Update)

Yea, or just create our own MonoBehaviour.