by beholder » 31 Jul 2011, 11:24
A question though: Will Curse of Chains replace Narcolepsy ? It's functionally the same, right?
by madmarcel » 31 Jul 2011, 11:40
I'm just having a great time implementing and testing new cards.
Just played mirror match of this versus AI:
20 Argentum Armor
AI defeated me with a flying 24/26 that destroyed 4 permanents per attack
Just noticed something else I have to test:
"You may target any permanent with the triggered ability, not just one controlled by the defending player. If no other player controls a permanent, you must target one of your own."
Edit: Nope, AI doesn't do that, declares that there are no legal targets. Hrmpf.
- Posts: 18
- Joined: 17 Jul 2011, 03:18
- Location: New Zealand
- Has thanked: 0 time
- Been thanked: 0 time
by melvin » 31 Jul 2011, 12:01
@madmarcel: The list of cards looks great Don't forget to "push" your changes to the main repo.
by beholder » 31 Jul 2011, 12:12
Oh, I agree but I personally don't see the use of having cards that are strictly worse. There are already a few of those in the game now, like:melvin wrote:I think it will be nice to have other formats besides cube in Magarena. In any case, new cards are always good to have so that the program stays updated with new developments in the game.
Eager Cadet vs Akrasan Squire
Crazed Goblin vs Goblin Gaveleer
..and possibly quite a few others.
by melvin » 31 Jul 2011, 12:27
For example, the default ubeefx cube has only Akrasan Squire but not Eager Cadet.
Coming back to your question, when Curse of Chains is in the game, it should replace Narcolepsy in the default cube, but I don't think we need to necessarily remove Narcolepsy from the game.
by ubeefx » 31 Jul 2011, 12:37
Target hint should be none probably, like for instance for Flametongue Kavu or Skinrender.Edit: Nope, AI doesn't do that, declares that there are no legal targets. Hrmpf.
Well ideally Magarena would support all cards ever printed and people can create any cube they like.Does he want to stick with the 'cube' idea or make as many cards as possible available in MagArena?
Or just create custom decks instead. So I see no reason to restrict the card pool, but it will add complexity.
That is why I started the topic about scripting whatever is possible, with a slow but steady transition.
Token generators are one the main flaws in the AI.Kemba and Luminous Angel can both quickly bring about game states where the AI takes a long time to think about what to do next BTW.
The combat code and to lesser extent targeting code number of combinations explode.
For combat, Melvin might consider a divide and conquer approach.
Break the attackers and blockers in smaller groups and calculate best options for each group.
Then combine the results and of course this will not always give the optimal choice.
But Magarena already uses other techniques which will not always give the optimal choice, so that is no difference.
Curse of Chains was in Magarena before and removed due to Narcolepsy, which is the cleaner version imho.
by melvin » 31 Jul 2011, 12:52
Indeed, blocking is one huge combinatorial explosion. I thought about a few approaches and decided to just settle for something simple for now. The AI checks a random sample of blocks and keeps the top 12 in cases where it is not feasible to check all possible blocking options.ubeefx wrote:The combat code and to lesser extent targeting code number of combinations explode.
by beholder » 01 Oct 2011, 12:31
There are less rules now for what cards can or will be added so I'll sum them up here:
- limit modeling and gameplay complexity : e.g. no Planeswalker cards
- no cards that require multiple targets
- only lands that help produce mana of multiple colors
- only cards that are played from the hand
Please do not request more than 10 cards at a time. It's daunting to have to go through a wishlist a mile long. When I have added your requested cards feel free to ask 10 more. If for some reason a card is not added, I'll try my best to explain why.
Keep in mind that our goal is not to add every card in existence but to add cards that the AI can play reasonably well.
by beholder » 02 Oct 2011, 05:14
You want to see how the AI handles these, right?
I saw the AI use Immolation as removal and as a way to pump it's own creatures. It's never perfect of course.
by jeffwadsworth » 02 Oct 2011, 14:11
by beholder » 04 Oct 2011, 05:32
If all goes as normal, you should be able to play with these at the end of the month when 1.19 comes out. Or download the source and compile
by beholder » 05 Oct 2011, 06:14
Zuran Orb - I'd normally skip this one because it looks like a card that the AI will have trouble with, but I thought I'd try and see what would happen. And indeed, the AI would need more hints,conditions and whatnot to play this card decently. May never happen, but who knows.
Psychatog - Discarding is hard for the AI but we have cards that use it. The problem here is the exiling of two cards. Which isn't impossible but something we avoid for now, hence the "no cards that require multiple targets" rule.
As much as I'd like to add every interesting card out there, I prefer having a challenging match without them. There are still lots of great cards that could be added though so keep them requests coming
Who is online
Users browsing this forum: Google Feedfetcher and 2 guests