Magarena 1.49

[Huggybaby]

My point was that if you're worried about Hasbro or anyone else, Forge has more cards, has more people involved, and is still doing fine.

NTM Waguc, which has an even larger forum, and sells an app on the app store.

And I could go on, but my point is there's no point in that line of discussion.

[hong yie]

My point was that if you're worried about Hasbro or anyone else, Forge has more cards, has more people involved, and is still doing fine.

NTM Waguc, which has an even larger forum, and sells an app on the app store.

And I could go on, but my point is there's no point in that line of discussion.

agreed

[Huggybaby]

Spelling correction=Wagic, not Waguc, arghh.

[Greendogo]

Are there any additional features on the horizon for Magarena, such as games involving more than two players (AI), or more than one human player (via network play)?

[Lodici]

Are there any additional features on the horizon for Magarena, such as games involving more than two players (AI), or more than one human player (via network play)?

I don't think either of these are in the thoughts of the current development team. I would say the primary concern is implementing a strong and challenging AI, increasing the pool of supported cards (without compromising AI) and improving the UI (my particular interest). But who knows, someone might come along tomorrow and start working on these features. That's the beauty of open source.

[ShawnieBoy]

Are there any additional features on the horizon for Magarena, such as games involving more than two players (AI), or more than one human player (via network play)?

I don't think either of these are in the thoughts of the current development team. I would say the primary concern is implementing a strong and challenging AI, increasing the pool of supported cards (without compromising AI) and improving the UI (my particular interest). But who knows, someone might come along tomorrow and start working on these features. That's the beauty of open source.

I agree with Lodici - Although having more than one AI opponent would be far more likely than a human opponent. But that isn't even on our radar right now.

The card pool (my particular interest) is steadily growing, practically daily, and with every card viable for the three AI's. The 50% mark of all tournament legal cards printed is in sight.

Magarena's all about the AI

[kannrain89]

well, I wonder if Magarena's AI could be implemented into Forge. If then, it will be the ultimate program.

[Lodici]

The ultimate program would be to create a common java MTG library/framework containing all the best bits from the various java implementations onto which you could easily bolt different UIs.

[melvin]

well, I wonder if Magarena's AI could be implemented into Forge. If then, it will be the ultimate program.

Magarena's AI implemented in Forge would probably not be as good as Magarena's AI in Magarena, it may even be worse than the current Forge AI. Magarena is built around the goal of creating a challenging AI, we achieve this by altering game rules and having a restricted card pool. The AI algorithms is just one of the ingredients for a challenging computer opponent.

[ShawnieBoy]

we don't have data at 6s, but based on results we published at https://code.google.com/p/magarena/wiki/AIComparison

At 8s, the strongest is Monte Carlo Tree Search.

This doesn't seem to be the case at the moment - I find it very strange that a 4s MCTS is performing better than the 8s...

[melvin]

This doesn't seem to be the case at the moment - I find it very strange that a 4s MCTS is performing better than the 8s...

The page was updated recently for 1.49, the older data I cited is further down on the page. It does seem odd that 4s does better than 8s. The actual difference is very small, so it could be due to not having enough games to reduce the variance in the scores. Another thing to note is that scores from the previous version are not comparable with scores in the new version.