Swepicduels.com, Swepicduels.com

Hey everyone,

A friend of mine and I made a version of star wars epic duels in silverlight for fun. You can play it at: http://swepicduels.com/default.aspx. It's only 1:1 - against the computer, and you can only play 4 characters - Vader, Luke, Boba, and Han.

There are probably a couple of bugs here and there, but it should run for the most part.

I'm thinking about adding more characters in the future, but I figured I may as well roll it and see what you guys think.

Also, the AI is kinda weak. A better AI should be my next update, but let me know what you guys think.

Thanks,

I'm impressed. I don't even know what Silverlight is, but I installed it, and gave it a whirl. My Boba Fett easily wiped the AI's Vader off the map.

The Good:
1. This is the first online version of SWED I've ever played with a computer AI. Very cool.

2. The background music was a nice touch and added a cool "Star-Wars-y" element to the game.

3. Pretty intuitive. I just kind of guessed on what to click and how to operate the thing, and my guesses were right in most instances.

4. The little quotes after a card is played was very entertaining at first.

The Bad:
1. I would like to see the actual maps used. If not the images themselves, then the actual layouts of the Hasbro maps

2. The card images looked like they were made in Deck Designer? I would have preferred actual scans of the card images.

3. The AI is very weak. To make it really usable, this would need to be improved. The thrill of playing a computer AI is never going to be as good as playing a real opponent, but there is definitely room for improvement here. The movement choices in particular were pretty bad.

At this point I'm most interested in knowing what the heck is Silverlight, and how did you go about making this? Did it require any advanced knowledge of computer coding?

This is impressive. Being a developer myself, I realize how difficult it is to create something like this. There are so many possible plays in Epic Duels that creating a good AI can be tricky. Are you using any sort of Min-Max algorithm?

I started a similar project myself, but didn't get very far. Kudos to you for having something this complete!

Thanks for the quick replies.

Darth Trumpetus - Silverlight is sold as Microsoft's version of flash. It's not coded the same way, but you can play games/videos/etc via client side. It has yet to boom considering flash is installed on 99% of all browsers, but MS is sticking with it.

Silverlight's front end is coded using XAML - pretty much coding in XML. The backend - I'm using C#. There was some extensive coding needed for this program - mainly the AI.

Thanks for the constructive criticism. The AI's movement is quite poor, and it'll never be as fun as playing a real opponent. Hopefully one day, I might be able to make this game multiplayer, but that's far down my list right now. AI, the other characters are probably next on my list. Adding the maps is a good point too. Green tiles just aren't too appealing to look at.

CoreyRIT - I am using a MinMax algorithm. The problem with it is that - in a given turn the average number of of actions the AI can do is about 40. Compared that with chess - when on a given turn the average amount of moves is 10.

So, the main issue with the AI is that - it can't look too far in the future. It can only see 1 move ahead. To see further into the future, I'll have to increase the amount of time in between turns to about 10 min - which is not a viable solution.

I've been trying to think of a good solution to limit the amount of moves/actions so that the MinMax algorithm will only have a select amount of moves to use, but haven't come up with anything yet.

Thanks again guys for the replies.

I agree. The other problem with MinMax for something like Epic Duels is the blind moves. For example, in chess, you know your opponent's possible moves on their turns after your own. In Epic Duels, you have no idea what cards your opponent is holding (true, the CPU could cheat), so it is hard to evaluate how good a move is several steps into the future.

Have you considered using any sort of learning (maybe Q-Learning?). That is, being that this game is online, log every single move that is played by your users and evaluate it's results and reward your AI when it does something good. That way when a similar situation happens in the future, it will know whether or not it should do the same thing.

I think there's a bug with wrist cable when it's played twice in a row. After playing two of them on a single turn, I ended up have 2 turns of just movement (no actions) before I got my full turn.

*EDIT* Experimented with it again to see if I could replicate the results. This time it had me roll the die 5 times before I could play out my turn with actions. Also, the game froze when it ended rather than declaring a victor and restarting.