MERLANDESE'S PROFILE

Very nice song, and it fits the atmosphere of the area.

Also, I think I live in the same city as your composer now. XD

author=Deltree
Woof, a division operation thousands of times per turn! I'll
E: Holy crap, I think I've discovered the method to its madness. It's aware that cards it takes can be retaken if there are an odd number of edges left over, so it's biding its time and targeting cards that can't be recaptured. I'm getting stalemates/close shave victories more often than not, now. Next, I think I'll have it play off the player's cards, just to see what happens.

I'm scared. Hold me.

Haha! Yes! This is what I meant! Sometimes your AI surprises you with new tactical concepts. I love this stuff, even though keeping your mind bent around it all can be a game in and of itself.

Glad Hedge1 mentioned the thing about the delay because it's the absolute truth. With all of my AI personalities in this board game thing I've made, they each have a sort of variable delay time to make it feel like they're thinking. It adds suspense and, more over, doesn't make it feel like you're playing against yourself. I also made it so that the easier opponents have a longer minimum amount of time (in fractions of seconds) that they wait before selecting a move so that "dumber" opponents feel like they're taking longer to think.

I'm pretty sure Triple Triad and Tetra Master showed some card selection on the opponent's side so it looked like the opponent was considering moves. I'm not sure if it actually was or if that was completely aesthetic, but definitely worth doing in Quad Pro Quo. Even though we have no idea which cards the AI has in hand, it's cool to see them shift among the cards like a human might.

author=Hasvers
For the beginning, it's unfortunately no big surprise - chess and go players have to learn openings from specialized books, and likewise you probably have to guide the AI a little. Or bake in good openings by running much more extensive MC on starting positions and selecting those that seem best.

That's also a good point. I imagine that opening moves deal a lot with protecting weak edges and using the not-as-strong cards to sort of "test the waters" of the upcoming battle. That might be a little trickier to program in, but a strong opening move could make the AI that much better.

I haven't played yours yet but, if I'm imagining scenarios well enough, I'd say that most players will not play their strongest, nor their weakest cards on the first turn. They don't want to blow their load without capturing a piece, nor do they want to be immediately open to attack on the first turn. Whether this turns out to be the best strategy or not is in the air, but if that's how most people end up playing your game it might be good to make the AI think in the same conventional terms. After all, you want the AI to feel like a person first and foremost, THEN to feel like a badass person. XD

The only issue that immediately comes to mind with that, Hasvers, is that a lot of us here--including yours truly--is only experienced in a few ways of coding or even just a few engines. I use RMXP and Construct 2 nearly exclusively to the point that even if I saw an open codebase for, let's say, the extremely promising The Tenth Line, the effort in decoding what I was seeing into terms I could comprehend would be the majority of the work.

author=Deltree
Higher level AIs build a weighted list of every possible move and then choose the best one, though I'm still tweaking those weight values for a decent challenge.

Awesome. AI is tough to do, but rewarding. You have to have them look at every possibility individually and then rank them in order of best-ness.

One way I've been messing with in a board game AI is to have them imagine placing the tile/card into every position, logging the results, and then re-imagine them placing another tile/card (not including the one that was "placed") into this new, hypothetical board where the previously placed tile/card exists. Then weigh out this list for which two-turn process results in the best opportunities.

The results are interesting because they sometimes "gamble," placing a piece that makes a little less sense so that in the course of two turns they have made two "better" moves so long as you, the opponent, didn't get in the way. Because of these inherent areas of possible fault you get an AI that is still really smart and calculating but by no means infallible.

Keep it up!