Why an AI for Outwitters is hard to do

[Kamikaze28]

The ability to predict enemy behavior is of course imperfect for both humans and AI. However, experience can increase your success rate.

Once again: the AI doesn't need a set of rules like the one you described ("if enemy is defensive, then strike his weakest point as heavy as possibly", "if enemy attacks early, neutralize his offensive and counterattack", ...). I described above the concept of game state exploration. The AI has the current state of the game and simulates all the actions it can do and all the actions its opponent can do and all the actions it could do in response to that and so on - to a certain extent (or to a certain number of rounds in advance). All these strategies you are describing are the intuitive result of this process - to look ahead and choose the most promising course of action. Defining these strategies "by hand" would be a herculean task at best and if you go to the extreme, you'll end up with a reflex-based agent Harti described above which has numerous flaws I pointed out.

[Solan]

ok i've got your point, realize what you meant now. Making an A.I.... would be a helluva job. especially in any sort of strategy game.

[tactfulgamer]

This was a good little read. I understand where you are coming from.

Maybe you can create a offline "practice zone"
So I can work on my combos and character synergies.

This game really has surprised me. I would love to invest more into its strategy.
Offline practice zone please. No AI.

[Harti]

@tactfulgamer:
Try Pass'n'Play mode!

[tactfulgamer]

This was a good little read. I understand for one where you are coming from.

Maybe you can create a offline "practice zone"
So I can work on my combos and character synergies.

This game really has surprised me. I would love to invest more into its strategy.
Offline practice zone please. No AI.

[revlisoft]

How is ths different from Starcraft and Warcraft though where ai exists. Sure SC time slices are different but doesn't the same concept apply? If we apply a grid to SC map and then instead of everything happening in real time (every CPU cycle) we slow down time so that it appears one CPU cycle is 1 minute that would be similar, no?

[projectantares]

Total Annihilation's nigh infinite community-based expansion packs certainly didn't require millions of dollars in funding, and my oh my some of them were more advanced and ass-kicking than any human I used to play against.

[metalsquid]

Games like Warcraft and Starcraft have development budgets that run into the millions of dollars and spend years in development. OML is a 2-man show. And as you can see, to implement a competent enough AI to provide a satisfactory challenge to a player takes a significant amount of development time and money.

---

(07-13-2012 02:41 PM)projectantares Wrote:  Total Annihilation's nigh infinite community-based expansion packs certainly didn't require millions of dollars in funding, and my oh my some of them were more advanced and ass-kicking than any human I used to play against.

They had the backing of a parent company like GT Interactive. They bought Humongous Entertainment for $76m in 1996. Cavedog Entertainment, a division of Humongous, put out TA in 1997. You sure they didn't have millions in funding? Even if accurate, hundreds of thousands of dollars is still nothing to sniff at for an indie developer.

[awpertunity]

Actually in SC (I would assume for WC too but not sure there), the computer does not have a fog of war, which greatly helps them write AI for it. I suppose you could start with the same idea for an Outwitters AI, to completely remove uncertainty from the programming (i.e., computer actually knows where all units are on the field, and even how many wits the opponent currently has to use). Of course, building a good AI off this information is going to be impossible to beat, since they are essentially map hacking you, but then comes in the probabilities for bad moves.

As already stated, in each state of the game (with no uncertainty) the agent should have a prioritized set of moves. To make AI easier you could reduce the chance of playing moves high up on this list with lower ones, this doesn't necessarily mean it's making a "bad" move, just one that is not as good. Of course coming up with this state-dependent list of priorities is still a huge challenge in its own

i.e., is it good to attack their base with my soldier now? Or should I take out the sniper instead? Of course this should depend on many things, the most trivial one being (for example), "well if the base has 2 or 1 health left, I should probably go for that".


These are mainly just talking points though and a first step at how to simplify the whole problem. I write AI for multi-agent systems myself, and am not expecting this to come out for Outwitters anytime soon

[getuliojr]

There are many ways to build an AI and I don´t think it is that much complicated. You just have to set the goals and on each new version of the software making it better as you learn along.

1. You can create a game based on played games. You choose only one type of map and get all results that has lead a player to victory and save his movement to a database. And put the computer to get copy this moves. It probably will go farther away and probably will find in the dabase a move he can take based on the actual state. BAD: you would need to download a lot of chunk of information to the mobile in order to have it data to play offline, otherwise even single player would require a internet connection.

2. You could write te goals that ends the game and all moves and attacks and create to achieve the goal and make the computer simulate, like a tree. How much should a computer simulate before playing ? Well in easy mode it can be like 10 moves ahead or 20 seconds of processing, in hard could be up to 30 moves ahead or 1 minute of processing. When each one reachs the computer stop, and execute the best one at the time que processed that would give him 70% of chance of victory to 100% of chance of victory based on certain simulated cenaries and plays it.

Next turn it restart and do all over again. NOT that much hard, but work is necessary. Right now I think the game is fine online only and should be put effort to make it even more better, fix the problems it already have, and then think about an offline mode (maybe even sell offline mode por 0,99 upgrade) to cut the time necessary to develop the feature.