This topic is 997 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

## Recommended Posts

i want to know minimax algorithm how really gets implemented.

for example as we know min is a player and max is other player. both of them try maximize their own performance and minimizer other ones. so the main question is:

does both of them have a complete information of every state? and it has to be generated? in most of examples i search it searchs in a tree. i think there should be other ways to search for that. for example in a game like tic tack toe you can simply find all possible states. you can save every state in a 2d array and program that best choice for this state is this one.

one other question can be:

for a game like chess states can be so big. for example for a computer like deep blue did they really found all those states?

many would say minimax is basic of video game ai. we know we have many different techniques for different styles of games but i want to know can it be used for continous open games? for example in a strategic of shooter game computer ai shoulf find the best strategy that makes it hardest for player or like that.

thank you for helping

##### Share on other sites

i have a chess, and several other table games, all using some types of minmax algos. for simplier games, i just give the pointers of the modified array, and they fully recalculating the points of that table. in chess, i store and morph usually special bits to every puppets, like if the puppet is alreday moved (need for castling, enpassant).

i not think a minmax can effectively used in complex 3d games, you must design more advenced techniques to achieve your goal.

##### Share on other sites

for a game like chess states can be so big. for example for a computer like deep blue did they really found all those states?

No. The number of game states in chess is beyond any computational power that humanity is ever likely to possess. Deep Blue used alpha-beta pruning (along with other techniques) since calculating all states is impossible.

many would say minimax is basic of video game ai. we know we have many different techniques for different styles of games but i want to know can it be used for continuous open games? for example in a strategic of shooter game computer ai shoulf find the best strategy that makes it hardest for player or like that.

Minimax doesn't really make sense in a shooter game.

##### Share on other sites

minimax comes from game theory, which despite the name, has pretty much nothing to do with AI in video games - or anything to do with video games at all for that matter.

generally speaking, games try (or ought to try) to implement some form of "expert system", using various algos such as decision trees, finite state machines, weighted decision systems, planners, etc. the best algo to use at each point depends on the type of decision to be made, and the desired AI behavior.

Edited by Norman Barrows

##### Share on other sites

minimax comes from game theory, which despite the name, has pretty much nothing to do with AI in video games - or anything to do with video games at all for that matter.

I pretty much agree with that, except I think of computer implementations of board games as video games. At least some classifications of video games support this view (e.g. https://en.wikipedia.org/wiki/List_of_video_game_genres#Board_game_.2F_Card_game ).

##### Share on other sites

I want to know minimax algorithm how really gets implemented.
The minimax algorithm itself is only about how to incorporate the information you get from one side to the other.

It does not concern itself with how you get that information.

Obivously, in any practical implementation, implementing a minimax algorithm without having code or data to get the information is useless. It's like having a compiler, but no editor for writing the source code. So you see information creation and storage, and the minimax algorithm always together.

##### Share on other sites

generally speaking, games try (or ought to try) to implement some form of "expert system", using various algos such as decision trees, finite state machines, weighted decision systems, planners, etc. the best algo to use at each point depends on the type of decision to be made, and the desired AI behavior.

Please note that "expert system" is a formal term with a very defined meaning. Just to avoid confusion.

##### Share on other sites

Please note that "expert system" is a formal term with a very defined meaning. Just to avoid confusion.

quite true.  "AI system" would probably be more accurate.

expert system is just one form an AI system might take.

##### Share on other sites

minimax comes from game theory, which despite the name, has pretty much nothing to do with AI in video games - or anything to do with video games at all for that matter.

Oof... dude... really? Having included game theory as part of my book on AI and given lectures about game theory in games, I disagree very much with this statement.

The same economic reasoning that goes into Prisoner's Dilemma can be used in agent AI. Many games can be reduced to variations on Rock, Paper, Scissors. Those are just 2 examples.

Anyway, not to get off track... but wow...

##### Share on other sites
It seems to be a bit of a pattern with Norman Barrows that if he hasn't found something useful in his personal experience, he discards it as useless (see another [more outrageous] recent example). I don't doubt that his experience is vast and valuable, but a bit of humility wouldn't hurt.

1. 1
2. 2
Rutin
16
3. 3
4. 4
5. 5

• 13
• 26
• 10
• 11
• 9
• ### Forum Statistics

• Total Topics
633722
• Total Posts
3013547
×