Why can't the computer Ai do this.

I'm guessing the Ai can do this, but it's to much work for Ai programmers Stardock, or anyone who cares. I'm not a Ai programmer, but I have an idea on how to program databases, and I understand linear programming. This won't leave me alone. Their are three programming structures in every program sequence which is it executes a function, or whatever you call it depending on the language, looping, and decision. The decision I'm familiar with the most is the If then else.

I understand when you program an ai the ai really can't predict what the player is going to do, but in all honesty without using my brain I could guess what strategy each player is doing by reading the victory stats. I'm familiar with galactic civilizations. I have espionage, and a chart that tells me who has the best economy, industry, research... If I can read this then it is only right to let the computer do the same. Instead of the computer of having one play style their can be many based on conditions like who has the best economy... screen for each type either for one ai for all players, or each ai player you decide.

Either the ai checks its highest neighbors, human players, or who its hostile in the victory screen. You still need a default strategy until this is set up, or in case of prediction failure. Now remember you are not using any stats the player can't use. Check these stats either at the beginning of each turn, or after so many turns whichever is fease able. These could be set up for sequential turn strategies, or based on how many turns in the game.

Ie If at war then if player ships> computer ships build ships else if player colonies<computer colonies build colony ships. Else If player colonies<computer colonies build constructors.  Else if computer not at war if computer currency > 0 then if player economy>computer economy if planets have blank spaces then build economy.  else if players research... I can detail the logic if anyone wants otherwise you get the point. The purpose is not to suggest strategy, but how to implements it.

I think it would be more constructive on this post to suggest types of strategies, not exactly how it's implemented. If you want to suggest that a language can't read test, or html I already asked a programmer it can be done with about a hundred lines of code. Can be made into a reusable class instead of rewriting it each time. I wasn't suggesting strategy here, so maybe my idea wasn't the best what I'm suggesting is through regularly checking victory graphs the computer could kind of guess the player strategies, and play one of different strategies, and change by how the player plays.

 More like a database. This is not perfect, but better than having one strategy of gameplay for the Ai. I could leave it hear with potentially the best ai ever, but I can't. Now lets say the opponents remembered the average of each type of strategies, of a certain level of turns the player plays. This could be spread to the neighbor concept. One ai for everyone, or a different ai for each opponent. Average would be the most likely played strategies for certain range of turns. If averages came up this could replace default strategies for certain turns. The default would be there when this rule fails. I would like to hear opinions.  

22,670 views 4 replies
Reply #1 Top

If I read that right, you essentially want the AI (In GalCiv?) to be responsive to what's on the victory screen at any given turn?  That sounds like a scaling difficulty.  Barring potential resource bonuses/penalties for a given difficulty setting, it seems like it would remove the need for difficulty settings; thus, your solution could just be to play the game on a different difficulty level.  The persistent memory of player strategy would be a whole different ball game, but with player profiles seems doable.  I think the thing you glossed over is that you're still talking about one AI; nothing you've mentioned would need or even use a different AI for each computer-controlled player.  I believe that the NPC factions use profiles and feed options to a master AI.  If you've a copy of the PC game Worlds of Magic, their AI profiles are in the clear in .xml documents, so you can see examples of exactly what you're talking about.  I don't have FE installed right now so I can't go check, but I seem to recall something similar there (though it's been a couple years and my brain is like RAM - one night's sleep and it's blank again).

 

Programming a complete AI for a game is a massive undertaking, and I've been impressed for years at how frogboy grinds them out.  Programming multiple master AIs would be a terrifying chore that would shatter minds like the sight of a Great Old One on your living room couch.  Or anywhere, really.  There's no good place to run into Nyarlathotep, the Crawling Chaos.

Reply #2 Top

You could do multiple ai's if the resources allow it. As far as the post I would rather an intuitive ai better than multiple ai's, If the resources don't allow it then you could create triggers like abilities. Not to mention we could use the default ai. Basically for when starting out, or when there is no prediction for playing style the same. For the different Ai's. I would rather allocate resources for a challenging Ai instead of different Ai's. The problem here is a lot of people have different play styles.

The Ai is the same for everybody. I've suggested something else that's better in the past. This is dummied down. Instead of predicting you basically do the usual thing just implementing pre planned strategies. I'm just suggesting that you use multiple strategies sets based on the same screens, and tricks the player use to monitor the ai results throughout the game. Yes letting computer do what I call cheating which is arbitrary bonuses is a solution, but to my opinion not a good, or the best solution. As a player I wouldn't mind more information if it would help make a better Ai. As far as current strategies that could be used for default ai. The differences could be default ai. I've nevered played a game where a different ai trumps better ai. really depends on how much resources you have.

Every turn checking is nice, and I support the most, but you could; also check after so many turns depending on how much resource you use.

Reply #3 Top


I'm guessing the Ai can do this, but it's to much work for Ai programmers Stardock, or anyone who cares. I'm not a Ai programmer, but I have an idea on how to program databases, and I understand linear programming. This won't leave me alone. Their are three programming structures in every program sequence which is it executes a function, or whatever you call it depending on the language, looping, and decision. The decision I'm familiar with the most is the If then else.

I understand when you program an ai the ai really can't predict what the player is going to do, but in all honesty without using my brain I could guess what strategy each player is doing by reading the victory stats. I'm familiar with galactic civilizations. I have espionage, and a chart that tells me who has the best economy, industry, research... If I can read this then it is only right to let the computer do the same. Instead of the computer of having one play style their can be many based on conditions like who has the best economy... screen for each type either for one ai for all players, or each ai player you decide.

Either the ai checks its highest neighbors, human players, or who its hostile in the victory screen. You still need a default strategy until this is set up, or in case of prediction failure. Now remember you are not using any stats the player can't use. Check these stats either at the beginning of each turn, or after so many turns whichever is fease able. These could be set up for sequential turn strategies, or based on how many turns in the game.

Ie If at war then if player ships> computer ships build ships else if player colonies<computer colonies build colony ships. Else If player colonies<computer colonies build constructors.  Else if computer not at war if computer currency > 0 then if player economy>computer economy if planets have blank spaces then build economy.  else if players research... I can detail the logic if anyone wants otherwise you get the point. The purpose is not to suggest strategy, but how to implements it.

I think it would be more constructive on this post to suggest types of strategies, not exactly how it's implemented. If you want to suggest that a language can't read test, or html I already asked a programmer it can be done with about a hundred lines of code. Can be made into a reusable class instead of rewriting it each time. I wasn't suggesting strategy here, so maybe my idea wasn't the best what I'm suggesting is through regularly checking victory graphs the computer could kind of guess the player strategies, and play one of different strategies, and change by how the player plays.

 More like a database. This is not perfect, but better than having one strategy of gameplay for the Ai. I could leave it hear with potentially the best ai ever, but I can't. Now lets say the opponents remembered the average of each type of strategies, of a certain level of turns the player plays. This could be spread to the neighbor concept. One ai for everyone, or a different ai for each opponent. Average would be the most likely played strategies for certain range of turns. If averages came up this could replace default strategies for certain turns. The default would be there when this rule fails. I would like to hear opinions.  

 

To add to this.  If something like this was to be implemented in the above strategy, difficulty levels for AI would change how often "polls" to the player's statistics are reviewed.

 

Ie.

Easy AI - Check every 25 turns for a change in strategy

Normal AI - Check every 15

Hard AI - Check every 10

Insane AI - Check every turn

Reply #4 Top

Or instead the first level can only play the default Ai. Then yhe next levels have the checks.