What is the true best strategy for duelyst? What is the best strategy for the meta? What are the best cards for each meta and in what novel ways can they be used?
All these are interesting questions, and so I am now introducing the Duelyst Card Evaluator and AI Trainer.
The first steps will be to make a working duelyst engine, and then the stadium fights begin! At first the fights will actually suck horribly, but hopefully, we’ll get something good in the first few million iterations!
That’s really cool!
How do we help, or is this just an advertisement?
Hmmm, I’m exploring the source code and writing the client right now. I’m not quite sure if there’s a way to help 
Guess it’s better than asking if there is anything we can do to hinder you…
Actually, since the duelyst source code isn’t obfuscated, the variable names are still there, and it makes sense. . . It might be easier just to mod the duelyst client to run the simulations rather than port everything to c#
I’m confused about what this is.
Is this a system that’s going to have bots duel each other to find the best deck possible (in the AI’s eyes)?
That’s the basic gist of it. I’ll probably start figuring out the deck compositions of di-s ranks to simulate ladder composition.
Sounds interesting. Do you have already an idea of how are you going to do that? Or you want to figure it out “on the fly”? Could it be ML, since you want to try a lot of different card combinations?
This is interesting then
Might wanna figure out the RNG cards tho
because if they get the best result from say, a paddo, or a chrysalis burst, or a flash - kujata’d kyhmera (double mechazor) they would think it’s the best thing ever
I have a few ideas already, but I need to make a fully functional duelyst client for C# first.
If you want to use the duelyst’s, already written code, you can try the embedded chromium browser for c#, and invoke js functions directly from c# (and viceversa). If you don’t have another idea already, of course 
Ayyy that’ll save TONs of time 
thanks man
But that is the best thing ever!
Monte Carlo evaluations will fix that
Are you still working on this? Any Major progress so far?
I’d love to see this once it’s done.
I went through some of the source and realized it might be faster to just rebuild the client from scratch. The source is like 300 thousand lines and absolutely huge.
It’s a pain in the ass because I’m no Javascript expert and I’m working through what looks like one of hover’s custom engines 
Or maybe Hilmi did this. Whatever it is, I can’t make heads or tails of it
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.
