AI scripts

from GalCiv2 Forums
I've noticed that all video games and computer games, from the beginning of time, games that have included settings for a difficulty level, all provide higher difficulty levels by basically cheating. The only way to make the game harder is for the computer to cheat. They have to get special bonuses or something to give them an advantage in order to make the game more challenging for you. The computer doesn't play smarter or better, it just cheats. So I will always play on Normal difficulty level and hope that the computer is playing with the same rules I have to play.

I remember on my 286 7 megahertz computer I used to have this chess game. At the beginning of a game, you could select how long you want the computer to be allowed to think for each move. On this 7 megahertz computer, even if you gave it just two seconds to move, it could beat just about anyone.

Now processors have clockspeeds that are 500 times higher and about a million times more powerful. So let's calculate how much more complex Galactic Civilizations II is compared to chess. Hmm, about a million times.

Now I realize you are never going to program a computer to act like a person. Which brings me to my real point. When are they going to make this game multiplayer?

And....

How long do you think it would take to play a game if you had say, 5 human players and 5 computer players on a gigantic map with everything on abundant and techs set to very slow?

13,531 views 17 replies
Reply #1 from GalCiv2 Forums Top
Now processors have clockspeeds that are 500 times higher and about a million times more powerful. So let's calculate how much more complex Galactic Civilizations II is compared to chess. Hmm, about a million times.

You must be a geezer like I am. I used to play Reach For the Stars on an Apple IIe with an 8 bit 1 Mhz 6502 processor with 64K of ram with an access time around 500ns. Played that game from 1983 until the motherboard died in 1995. Great game. That game certainly used the concept of making computer players smarter by giving them more resources or what you call "cheating".

As you say, processor speeds, bus widths, memory sizes and speeds have dramatically increased over the years. I would conservatively estimate the net total performance increase in "PC" hardware in the last 25 years to be about 10,000. Forget about the specifics of GCII for a moment and just consider the equivilent level of improvement in software overall. A lot of the improvements are in the area of graphics and user interface. Though clearly an improvement how much has this contributed to net productivity? I would say to be generous perhaps it's doubled. This is a subjective opinion, but to say that software improvements have resulted in actual productivity improvements on the order of 10 times would be extremely generous. So PC's today are 10,000 times faster and we can actually accomplish about 10 times the work we used to be able to. Where's the missing 1000 times improvement in hardware performance gone. Well, I believe Mr. Gates wastes a significant portion of it. I also believe a fair amount is wasted in the volumes of c++ (or whatever) code that's since replaced finely tuned assembly code. But most of the hardware performance improvement is simply left in the box. At work I have a 3.5 Ghz Xeon processor whose main purpose is for me to answer email and other standard word processing chores. A couple of times a year I might actually use all of that horsepower to auto-route a board I've designed. It might take me an hour or two. Ten years ago it would take me 4 to 8 hours to do on a 600 Mhz Alpha workstation.

The point of this rather long digression is that GCII is not 10^6 times more complex than your chess program from days gone by, it's probably closer to 10 times as complex with beautiful graphics and a gui interface.

Anyway, you asked a couple of questions based on the following statement.

Now I realize you are never going to program a computer to act like a person.

With this statement you're essentially dismissing the game as unworthy of play without multiplayer. There are certainly people that share this sentiment, but I think there are many more that don't. In my opinion, this is one of the best PC games, if not the best, of all time. If you give it an honest play you will most likely become as addicted to it as most others here are.

Anyway, to address your specific questions. My understanding is that multiplayer will eventually be available, but will be developed seperately in conjunction with an future MMO type game and then back fitted into GCII. Others here can perhaps give more detail than I can.

The length of time for a game will probably be very dependent on the exact mechanics of multiplayer, but I take about 40 hours to play a single player game of the type you describe. I can't imagine multiplayer could be faster (unless your opponents quit the instant you appear to have the advantage).
Reply #2 from GalCiv2 Forums Top

So I will always play on Normal difficulty level and hope that the computer is playing with the same rules I have to play.

Well, for GC2, at normal level, the AI is penalized. If you want an AI at the same level than you, you need to set the AI at intelligent which translate into the tough level.

Sins of a Solar Empire II
Reply #3 from GalCiv2 Forums Top
If you give it an honest play you will most likely become as addicted to it as most others here are


    
Reply #4 from GalCiv2 Forums Top
As peace phoenix said, at normal you're the one cheating. Play at Tough for a fair game.

Also, the AI isn't scripted. And muliplayer is not coming out for a rather long time, if at all.

Oh, and saying the computer isn't playing smarter than you shows you don't know how the GC2 AI works. It's currently able to beat Frogboy (who writes the AI itself) on Tough, sheerly by capitalising on his mistakes.
Reply #5 from GalCiv2 Forums Top
Most game developers will focus their effort on the glitzy aspects--cool graphics and an interesting rule-set. This leaves them with little left over to design a worthy AI. And so they end up taking the short cut of loading up the AI with advantages the human player doesn't have (i.e., "cheating").

Consider the chess example--how long did it take for the best minds in computer science to program a computer to beat a grand master? You really need to hire a strategist and game theory expert to really design a worthy AI, not a programmer.

One thing that I do like about Stardock and Frogboy is that they are truly trying to build an AI worthy of taking on a human player and adapting its own play style to counter him. The fact that the AI still needs advantages at the highest difficulty levels shows they are not there yet, but they are on the right track.

Keep up the good work Stardock!!
Reply #6 from GalCiv2 Forums Top
The AI in GalCiv2 ranges from moronic to unbelievable. I'm still not able to beat 9 opponants on "full" (tough) with all their advanced scripts enabled and no economy bonus above mine. It's a really good step above most games out there.

But the OP is wrong, clockspeeds make NO difference in the ability of programers to make advanced AI. Advanced AI comes from advanced programing techniques and I imagine a huge amount of work. The processors instructions per second and many many other factors (FPU's etc) are significantly more than 500 times more powerful than a 7MHz 286. That aside, that old 286 was doing everything with the cpu, including graphics, game logic and rules etc. Factor in ram size and speed, disk space, turn based nature of GC2 and many other issues and you realise that the limitation of the AI is not the computer but is in fact the budget allocation, time and skill of individuals involved in creating the AI.

What would be the point of hiring 2 Frogboys (obviously very highly skilled) to program even more advanced AI when a lot of people can not beat tough? When they are releasing an expansion? When their budget gives them only a certain amount based on sales? Give 10 Frogboys 10 years to program some AI and no one will be able to beat it (a few maybe, or cheaters?), but what would be the point for us or the company? The AI in GC2 is probably it's strongest plus and is an example of why we dont always need multiplayer - for expansions and sequals I imagine and hope that AI diversity and strength will be a continuing focus.

Oh yeah and if you get bored playing the AI at its smartest without cheating, then come back and start a thread on this, not before.
Reply #7 from GalCiv2 Forums Top
I'm still finishing a 1.3 beta 2 game on the largest map. I think I'm on painful, the AI level where they get a small bonus. The AI is much better than the original release and is challenging to play against.

I consider restarting until you get a good position cheating. I think a lot of people who are complaining the game AI is too easy are doing this.

One area that I see that would make the game much more difficult is AI use of starbases. I see them building starbases, but when I click on them, they never seem to have components added? And this late game when some AI's have giant economies. Would I be able to see the AI components, or are they invisible to the player?

Actually not to get to far off topic, I don't think starbases were thought out well in galciv2 by the developers. Many times it's my starbases that are saving my butt economically and militarily when I'm at a disadvantage. On the other hand, as soon as I am near parity with the top AI players, the game becomes boring because I know I have a huge advantage from skillful use of starbases, so it's only a matter of time before I win. If the AI used starbases even competently, then this would be less of an issue.
Reply #8 from GalCiv2 Forums Top
The AI in GC2 seems to be very good, possibly even the best game AI. Sure there will be situations where it could be better, as in utilizing starbases but overall povides an excellent challenge.
As for multiplayer, how are you going to coordinate a 40hr game with a bunch of people? I used to play Empire Earth online and getting people together for a 30-60 minute game was often a problem.
GC2 seems well suited to single player and I'm quite happy to have it so.
Reply #9 from GalCiv2 Forums Top
I dunno playing with the AI at its toughest setting, while far better than other AI's I've encountered still is no match for a human player. At tough, for me anyway, the game is still a breaze, I'm not sure what the economic benefits are but I tend to play on Painful.
Reply #10 from GalCiv2 Forums Top
What I don't get with this AI is how much better it would be if it just didn't autopilot all the way across to where it wants to go, and not pay attention to what it's passing...

I understand that re-evaluating it's course for a fleet could cause it to cycle, and never get anywhere... But if they could address this I doubt many people would be winning over maso.

Imagine the AI building fast ships, that retreated in a meaningfull way to stay alive. Imagine if it tried to avoid your warships and attack your transports.
Reply #11 from GalCiv2 Forums Top
Giving the AI bonuses at higher levels is not cheating; it stills has to follow the rules but gets a bonus like you get from your starbases mining resources. A cheating AI will have no FOW (e.g., can see where the best planets are before exploring; this was true in GalCiv1) or gets free money or resources, etc.

I got tried of crushing the AI on tough and so for my first Metaverse I played on Painful (AI Gifted) only later to learn on that level the AI only gets a 5% bonus. I recently completed my 2nd Metaverse game on Masochistic (AI Incredible - all Algorithms + 200% Production for the AI) and still crushed the 9 AI opponents, although the early part of the game was challenging. I understand the AIs (partically the weak Terran based AIs, Terrans, Drath, and Iconians) are much better in v1.31, so I look forward to my next game in v1.31 on level Suicidal! Despite the AI's flaws in GalCiv2, it far better than in most games.
Reply #12 from GalCiv2 Forums Top
How long do you think it would take to play a game if you had say, 5 human players and 5 computer players on a gigantic map with everything on abundant and techs set to very slow?


I'd say a couple of weeks max before everyone starts pulling out and it dies.
Reply #13 from GalCiv2 Forums Top
The point of this rather long digression is that GCII is not 10^6 times more complex than your chess program from days gone by, it's probably closer to 10 times as complex with beautiful graphics and a gui interface.


I strongly disagree. I'd guess that GC2 is, from a computational point of view, vastly more than a million times as complex as a chess game.
Reply #14 from GalCiv2 Forums Top
Oh, and saying the computer isn't playing smarter than you shows you don't know how the GC2 AI works. It's currently able to beat Frogboy (who writes the AI itself) on Tough, sheerly by capitalising on his mistakes.


There's no evidence to suggest Frogboy is unusually good at playing GC2 - we only know that he's pretty good at programming it. Lots of people frequently win on the highest difficulty level, and I can routinely beat Tough.
Reply #15 from GalCiv2 Forums Top
There's no evidence to suggest Frogboy is unusually good at playing GC2 - we only know that he's pretty good at programming it. Lots of people frequently win on the highest difficulty level, and I can routinely beat Tough.


True, but that does prove that it's capable of beating a human, and a human who "taught" it how to play, at that
Reply #16 from GalCiv2 Forums Top
The AI in GC2 seems to be very good, possibly even the best game AI. Sure there will be situations where it could be better, as in utilizing starbases


I retract that statement. Last night, two AIs (Torians and a minor) seemed to adopt a "Starbase rush" strategy. The Torian space is thick with developed starbases and the minor has, for some reason, filled MY space with influence bases, the dirty dogs. Didn't really want to attack them as we do a lot of trade, but they left me no choice...
Reply #17 from GalCiv2 Forums Top
There's no evidence to suggest Frogboy is unusually good at playing GC2 - we only know that he's pretty good at programming it.


We have concrete evidence, that a minor knocked him out of the game    

But when it comes to programming he is a god...he wrote GC2...nuth said.