AI card prescription for decks and combos

[Triadasoul]

I don't know how relevant the idea is, but i think it better to write it here %). I think it'll make effect without complicating the AI and card implementation.

What if the user can make a list of combos used in the deck and a list of critical cards that will be in prescription block of each deck for AI to use it to cast more complex combos and to protect/counter/destroy crucial cards of his/opponent deck?
+usage of combos in certain situations such as: low life, can't attack, crucial opponent's cards on battlefield etc...

[Huggybaby]

That's an interesting idea.

[mtgrares]

Your idea is fine. Currently the AI is very crude. It has a hard time choosing targets and paying for cards and sometimes it can't even do that without generating errors (in Java errors are usually exceptions). The AI randomly plays cards from his hand. (It isn't quite random, the computer' s hand is sorted from high cost to low cost, and the computer tries to pay high costed cards first.)

The computer is so stupid it doesn't even know the result of the cards that it plays. Right now the computer is barely able to walk. Hopefully later the computer will be able to run.

[Triadasoul]

I hope it could fly, because this project is great!

Also there could be a value integer of combos/crucial cards for ai to choose. It's somehow implemented in manalink as "base card value". But in the context of a deck the values should differ for a certain card i think, so this values also could be put in prescription part.

[mtgrares]

Magma (other program that lets you play against the computer) actually lets you assign a "priority number" for specific cards in a deck, so different decks could have different cards that are important to that deck. This type of system is good because the AI could "look" for those cards by mulliganing and searching for those specific cards using Demonic Tutor.

[Rob Cashwalker]

It's a value that's only useful in context of a single deck. And even then, the size of a deck, and format may skew the usefulness of it. Think of a 100-card singleton... or 250-card Battle of Wits....

But in general, cards with longer text are probably more valuable/essential which translates to better targets for removal AI. And it's an easy decision to code.

[zerker2000]

Nice approach to killing combo pieces . Makes total sense, and will probably result in a "Oblivion Ring -> Stuffy Doll" sooner or later .

[Triadasoul]

And i think it's an easiest method to implement such cards as Zur's Weirding