Hnefatafl: The Viking Board Game (download on page 7)

[jsj795]

oh yeah, the factory. lol I totally forgot about it I hope to finish this within a week, and start working on TLM again.
I think I did pretty well on the exams. I don't have anything major now, except AP tests given on May.

[DJ Omnimaga]

aaaah ok ^^, glad to hear ^^

As for Factory I was worried

[Zera]

Went ahead and updated the first post so everyone knows the new download is on page 7.

Was the king's throne issue addressed, or is that going to be in another release? The other capturing mechanics seem accurate, though.

[DJ Omnimaga]

Btw is the game supposed to work when both players are set to A.I? Will the two AI's try to beat each others or will it error?

[jsj795]

The game is supposed to work when both players are set to AI, and I want them to play differently. And the king's throne will be in another release (I already fixed it, but I want the defender's AI to be working)

[DJ Omnimaga]

Aaah ok I'll give it a try then

[jsj795]

Wait... if you set the AI as defender, it won't do anything. I haven't coded for defender yet. Right now, I'm figuring out the way to optimize the attacker's AI.

[DJ Omnimaga]

Oooh ok, thanks for the info

[jsj795]

YES! I finally found a way to make the AI faster!! I made a small look-up table that gets created every time AI goes. The position and its value that the calculator already found is stored into two lists, one for position and one for value, and if the calculator later tries to search for the value of the position that it already found, then it just gets the value from the list and move on. Using this, the AI speed got cut in half, resulting in ~10 seconds for the AI movement. I have to run tests for any bugs or things that might be present, and exactly how faster it runs, but I think it's good enough. And the amazing thing is that it only added ~50 bytes to the program (although it requires two ~30 dimension lists during the run-time)

[Zera]

That's awesome. Sounds like A.I. time will be significantly cut short.

Some rounds could go on for a pretty long while, depending on how skilled the A.I. is. It might have been tedious to wait some 20 seconds for it to execute its move each time.

[DJ Omnimaga]

Wow nice optimization. I agree with Zera. Sometimes I played chess with my bro and we would think for 20 seconds each sometimes, causing a game to last a few hours

[jsj795]

:'( Sorry I lied. It did make the AI think shorter, but only on certain times when there are overlapping position thing. I was excited that it took 10 seconds to take the 2nd and 3rd move, but as more possible moves opened up, the longer AI took to make a move... AI spends 8~70 seconds depending on available moves possible. And note that I'm running the program with TI-84+SE, so it would take twice as long on TI-83+...
The AI itself is not too stupid, sometimes it makes obvious mistakes, but it also surprises me with the move that I would normally miss if I didn't look carefully. But it would be easy to destroy AI if you are intermediate player playing white.

I tried to make another version of AI, where the list contains all the possible moves and their values, and the positions and values fluctuate, but it took too big of a memory (two 1000 bytes lists), speed gain wasn't too big, and the AI got more stupid, so I got rid of that idea.

What I am going to implement is the alpha-beta pruning (although idk if that's going to lower the speed significantly) and some more minimax algorithm. Also I might try to get rid of the For( group (it will increase the size of the program a lot, but I think for AI the speed is the utmost importance). Hopefully, I can get something to work.

[Zera]

That might not be a problem, if the longer moves are the ones that occur less frequently. If those types of moves aren't generally important to the A.I.'s skill level, then maybe they could be removed, or supplemented with another strategy.

[jsj795]

Well, it seems like that the AI takes longer and longer as the game goes on, because there's more available moves possible when the pieces are more spread out and the board gets more complex. So the average speed seems to be around 30~40 seconds

[DJ Omnimaga]

ouch sorry to hear, I wish you good luck

Hopefully a solution can be found