Any chance of making an Ascension ai?

[karazax]

I love your Dominion ai, any chance of making a similar program for Ascension Chronicles of the Godslayer?
If you're not familiar with the game, it is similar to Dominion, but with a more Magic™ draft feel, designed by Magic™ Pro Tour champions Justin Gary, Rob Dougherty, and Brian Kibler.
http://www.boardgamegeek.com/boardgame/ ... -godslayer

[joewtwins]

This would be Awesome.

[jatill]

I haven't played Ascension at all.
If anyone wants to send me a free copy of the board game, I'll give it a look

[karazax]

There is a chance I could get some players in my group to pitch in towards a copy for you if it really looks like a game you would be interested in making an AI for. It's not too terribly expensive at $26.99 on coolstuffinc.com

There is a visual spoiler card list here:
http://www.ascensiongame.com/board/topic?f=2&t=166

and the rules can be found here:
http://www.ascensiongame.com/downloads/category/rules

Check it out and let me know what you think.

[telengard]

I have been eyeballing this game too as a new project. A fixed deck might be a good candidate for a neural net AI.

~telengard

[Snacko]

Looks interesting cross-breed between Dominion and Thunderstone which again was based on Dominion.
I suppose MC based approach for AI could net good results as neural nets are very hard to get working for card games. I've ever seen only a Black Jack neural net implementation and that game is very easy to model compared to this.

[jatill]

Looks interesting cross-breed between Dominion and Thunderstone which again was based on Dominion.
I suppose MC based approach for AI could net good results as neural nets are very hard to get working for card games. I've ever seen only a Black Jack neural net implementation and that game is very easy to model compared to this.

What's an MC approach mean?

[Snacko]

Monte Carlo simulation with UTC which was used for a game go with good success. Basically you run random games, however with a card game it would be a bit cheaty AI as it would need to know the card order and other player cards, however with so much randomness I wouldn't be concerned about it as a real player could count the cards and have similar advantage.
There's a good starting article about it http://senseis.xmp.net/?UCT which included a bit of code in Delphi and Java (quite easy to understand ever if you haven't ever say any Delphi code).

Offtopic:
There's also an other Dominion like game made by an independent game designer Puzzle Strike. It uses chips instead of cards and the catching mechanic is that you don't really try to win but make others lose before you . The "card" or rather in form of chips images are available at the site, also from what I scouted they offer a pdf you can buy and print in form of cards.
http://www.sirlingames.com/pages/puzzle-strike-overview

[telengard]

Looks interesting cross-breed between Dominion and Thunderstone which again was based on Dominion.
I suppose MC based approach for AI could net good results as neural nets are very hard to get working for card games. I've ever seen only a Black Jack neural net implementation and that game is very easy to model compared to this.

Keldon's AI for Blue Moon and Race for the Galaxy (both card games) are *excellent* which is why I was thinking neural net. I have MC in my Dreamblade game and it does OK, not great and that is with a perfect information game. Minmax still rules the roost for now.

~telengard

[jatill]

I made some inquiries about making this game. The marketing director of Gary Games asked me not to move forward, so I'm going to let this one go.

[karazax]

that is unfortunate, though better to find out now than after spending a bunch of time on it I suppose.

[Fracturedorb]

I find it kind of amusing that with both this and dominion there's resistance on the developers part. I don't find the fact that there is resistance amusing, just that they are likely trying to make their own version of a AI based product but neither will see a release date that will happen while they are still relevant.

[Huggybaby]

Exactly. In fact, they will likely never be released, and the chance to capitalize on the exposure they would get from a nice free app will be squandered.

Eventually these attitudes will change, but for now most people who wish to monetize their creation are living in the past. I would NEVER turn down the chance to have a fan make a computer game from my RL game. The more the better in fact.

Nothing is being stolen. A music biz analogy: The devs don't want someone to steal the song and share it, but that isn't what jatill asked permission to do. It's more like he wants to make a video of a remix of the song. Not even the same, and nothing but free publicity which is better than any paid publicity you could buy.

I smell a lawyer somewhere in this mix.

[jatill]

The marketing director e-mailed me to let me know that they had changed their mind about this. I'm free to create the game without their interference (again, provided I do not use original art). I wonder if this thread swayed his opinion at all?

[Huggybaby]

I don't know, but that's great news. Congratulations!