AI too tough?

I'm really looking forward to GalCivII releasing soon. I haven't played a TBS game since MOO2 that I really liked. From what I'm reading, though, I get the feeling like the AI is going to wipe the floor with me. I like this games but I'm also not all that good at strategy games. That said I also don't like "cheating" by having the computer just give me extra money. That feels lame to me.

So how about this as a method for "cheating" other than just handing the computer more resources than the player, or vice versa. At any given time when I'm going to make a move I have several options (more than I can count usually) and I don't always make a perfectly optimal move for my situation ( like I said I'm not that good at strat games). So why can't the computer do that too? Let it have an X percent chance of choosing a sub-optimal move based on its current situation.

I agree with the AI team's assesment of "good AI" vs "hard AI" and scripts versus intelligent decision-making, but it seems to me that if the goal is to emulate a human player it should have more to do with making the wrong decision from time to time than how much resources a computer player has to pull of given strategies.

Let the computer have full use of all of its various faculties, but stick a roll in there that determines whether they'll choose an optimal or sub-optimal strategy. The chances for that roll then change from difficulty level to difficulty level.

I hope this makes sense and that it's useful to someone. I'm not an AI expert and I trust that you guys will do what's right, and maybe you're already doing something like this, but I just wanted to share.
7,303 views 8 replies
Reply #1 Top
I can understand your concerns, but the solution you describe is more complicated than necessary, and I think there is no reason to worry.
You see, on lower difficult settings the AI isn't using all algorithms. That means, without cheating, the AI is weaker than on higher difficulties because it will take far less factors in consideration when making it's decision than on higher levels. Though the way to get there is different from what you describe, the result is pretty much what you wish: The AI is playing weaker and making more mistakes on lower levels - not because it finds the right decision and randomly discards it, but because the decision-making-process is simpler and so it often won't even find the right decision.
Reply #2 Top
I had also worried about this a little bit too. There are multiple ways to enjoy a game like Civ 4. Multiplayer aside, you can either play it to be challenged to your utmost, always struggling to keep your head above water, with one mistep resulting in your defeat, or you can play it more as a "Sim"...just the joy of building your civillization in exactly the manner you see fit. In GalCiv 2's case, just the joy of building your empire, naming your planets what you want, crafting your ships how you like, etc. An AI thats too ruthless and always breathing down your neck even on the easy levels can close off this avenue to you (short of cheats or hacks).


BY all accounts, Stardock has crafted a very smart and challenging AI, which is great for when you want to push yourself, and neccesary for replayability in a single player only game. But something commonly observed in many thrid party previews of the game thus far, is that the AI is very agressive and challenging, even in the "easier" levels. Will this stymie attempts to just play through the game "enjoying the experience", and being able to take time to smell the flowers, as it were?


Id be lying if I said I never just enjoyed feeling like an all powerful emperor without peer, visiting my wrath upon any who opposed me in games like this...especially earlier on when I dont have the skills built up yet to really challenge the AI on higher levels.
Reply #3 Top
im pretty sure that if you just set the ai to the lowest levels then you could do all that you described, stardock wants to make a smart ai for the experts and replayablility but it dosnt want to make it near impossible to beat ig you jsut set it to the lowest it will probably be very easy for anyone to beat
Reply #4 Top
or you can play it more as a "Sim"...just the joy of building your civillization in exactly the manner you see fit. In GalCiv 2's case, just the joy of building your empire, naming your planets what you want, crafting your ships how you like, etc.


Hey, I'm not the only one that does that! Cool...
Reply #5 Top
It's mostly an issue of feeling good about myself when I'm playing. I'd rather play against an AI that has all of the same capabilities as me, but just isn't as smart sometimes. At super low levels I can still see the advantage of the player being given extra money, but I see that as more of a checkbox on the difficulty selection. AI difficulty separated from monetary bonuses or penalties.

I know they're planning on making some of the AI functions disabled at certain difficulty levels, and I certainly understand why they're doing it that way. It's much much much more time efficient and will probably yeild similar results. I know what I proposed is not the most efficient means of programming an enemy AI. I'm not an AI programmer as I said, but I do have a background in computer science and programming, so I have some idea what it takes to write code like that.

figuring out whether a move is a good move or a bad move is really tricky. I think they only reason I care is because the AI is so well documented in the journals that I feel like I already know a lot about how the computer opponents will play. If stardock had never told me that at various levels of AI difficulty that X was happening, I'd have been none-the-wiser. They just told us so much that now I can nit-pick about how they implemented their AI
Reply #6 Top
I think Citizen Yellow Sign hit upon it. Its a matter of the "fallibility" of the AI.

Presumably, playing on a beginner level, youre just that...a beginner...youre going to be exploring the different things to do in the game. You might do things that arent tactically sound, but just because its cool to do so. That kind of gameplay has its own appeal..namely, a lot more freedom. For instance, you might decide you want an empire thats made up of a central world, flanked at nearly equal distances by a ring of barrier worlds. Pursuing a largely cosmetic trait to your empire like that at higher levels would get your backside handed to you...a highly competant and aggressive AI requires efficiency, and precisely used resources to defeat.

So at lower levels, its not so much about letting players cheat by getting more starting money, or some other boon...but just having the AI as prone to poor decisions, procrastination, and whimsy as the player. Not that thats easy, Im sure...making a creatively error prone AI is probably as tough in its own right as a smart, super competent one that has to play by the same rules as the player.

As long as there are settings that wont get me blown out of the water for playing a more whimsical style of game, when I dont feel like I have the stamina for a cutthroat game, Ill be happy.
Reply #7 Top
The one thing that bothers me is that I think at normal the computer gets hamstrung on it's economy. I think it should be just dumber, not actually handicapped. To get the AI using 100% of it's economy you need to set to inteligent, but then you also have a much smarter AI.

Reply #8 Top
I believe there is something inbetween normal and int. that makes the computer dumber and doesn't handicap the economy.