Pathfinding: Putting in the yo-yo

[Capt Dizle]

just kidding Dexters...

[dexters]

I stand behind the fact that you are a moron if you continue to be insulted. Your inability to understand my last message is certainly indicative of your inability to discern insult directed at you from a general comment.

But if you insist on being included in the band of whiners, sure. You are a whiner.

Edit: Nice of you to throw in that "just kidding" message. Have a nice day.

[Capt Dizle]

What a riot some people are.

Anyway, this is a big problem with the way the AI is programed.

For instance say you build a defensive wall along your frontier consisting of maybe 3 of your best defenders. The AI can't conceive of blasting through so it will not attack.

OTOH, if you open one tile here they come. Let the AI get about 100 attacking units up there in 3 or 4 stacks. Close the hole - they all retreat. Thats not good. Any stack could blast through.

So, it's a clear flaw. Should be corrected.

[Capt Dizle]

And as far as not exploiting this, how? What would you do?

Not build a defensive wall? Not open a hole? Not close the hole?

Not build any military units at all?

I mean, to what extent are you going to handicap yourself to simulate a credible AI attack?

[nato]

Well Catt, some people can't help but cheat and then whine that they can cheat.

These are the same people who blame everyone but themselves for their own failings.

I've personally never used the undefended city

Theres a big problem with this.

As a rational person, I want my border and front line cities to be well defended. They are most in danger of being attacked, so I logically want walls and multiple defenders.

My interior or core cities however are not in danger of being attacked. Building walls and multiple defenders is a total waste of precious resources. As a rational person, I want to go as light as possible on defense, as in no walls and only one or even zero defenders.

This only makes sense to do. In Civ terms, I'd be foolish to defend Kansas City, in the very center of the US, as heavily as Seattle, on the very border with ever predatory Canada.

Since I'm aware of the AI, I never keep a city totally undefended, but there is no way I'm stacking all 50 or so of my cities with 3 or 4 defenders plus walls. Thus the AI always does its futile parade, walking past the most logical targets to get to the interior, never making it.

So it is dumb to take advantage of exploits, but that still leaves legitimate problems from non-cheaters.

For instance, another problem I have is the AI is too easy to keep at peace with. There are times it should realize I'm a clear threat and declare war. I could fix this by declaring war myself, even though that is very much against my in-game interests, but ...

At that point I'm no longer playing Civ as best I can; I'm trying to limit myself and coax the game into some sort of script. Its no longer a game where you behave optimally to win, its like a story you are trying to write.

Thats ok, but its something different from a challenging game.

[jshelr]

This topic is not a whine!!!

I think this thread is targeted at a very significant potential improvement in the AI war-fighting capability. Attackers have to use speed and target a city reachable in one turn as a priority, particularly late in the game. Like others above, I've seen the AI arrive at my door with a stack that should have, at least, taken out several cities. But instead of targeting a border city, they headed for the lightly defended interior while my railroads allowed artillery plus armor to chop them down to nothing. While I'm no programer, I think the AI could be taught to blast away at the border first and move inward as victories mount up. Equally important, Bismark should stop trying to attack with 4 panzers -- it's either 50 panzers on no attack at all. Strategically, I've got a similar big problem to jimmytricks small problem: why should I make the stupid move to attack when I know that I'm invulnerable late in the game due to AI offensive deficiencies and that a SS victory is virtually certain? (Turning off the SS eliminates a critical balancing factor in the game.) I'd like to see the game make life more difficult for the player -- which is not a whine.

[ShredZ]

..as heavily as Seattle, on the very border with ever predatory Canada.

I live just north of the 49th, and let me tell you, if you so much as flinch down there, we're ALL over you!

[nato]

Don't I know it!

[MiloMilo]

jt and others I think you are blowing this way out of proportion.

FIRST of all, regarding the yo-yo trick: you can do this to a human in MP, and it will work just as well. If I attach 3 units to your one, I can stop you from going, say, west. Being smarter won't let you get around this. You could declare war, but if you aren't as powerful as me you wouldn't do that. So what's the problem? I've done this occassionally: if I see an ai settler beelining for a good city site and it's one tile closer than my settler, I'll delay it for one turn. And then STOP. After that I don't care. Spending a whole game doing this to all the ai settlers isn't civ3. It's stupid tedium.

I haven't noticed the coast-tile-galley exploit; I haven't ever bothered to try it. I still won't now, because it seems boring. If I was a real king directing real fleets against real enemies, would that be a tactic I would use? Of course not, it would be incredibly silly. (That's what makes it an exploit as opposed to a tactic.) So whatever.

Regarding chokepoints: If I have a 4-tile-wide chokepoint that I want to block, I'll put 4 warriors there. If I don't want to block it, I won't. It's that simple. Why would I move a warrior in and out, in and out of the line? What reason could I have for that, in the context of the game? I may know as a person sitting in front of a computer that it will cause an illogical reaction in the computer, but from the point of view of the leader of a civilization, it makes no sense.

Likewise with undefended cities. I understand that it may not be extremely efficient to station troops in cities far from the border, but sometimes I like to be a bit inefficient. Pull the troops out of my capital? But then I have no troops there! It doesn't have an effect on the game, but it does have an effect on how I envision my most advanced cities. Just stick swordsmen or cavalry there, units that can't be upgraded. I generally have two units stationed in my core cities, and I have NEVER seen ai stacks go for them instead of a border city, even if I have 15 crack defenders in the border city.

The point is, if you can't not use exploits like these, you're breaking your own suspension of disbelief, and improving the ai won't ever help that.

[Catt]

Originally posted by jimmytrick
Catt, I just don't think you understand my point. When I am playing the game with almost certain knowledge of what the AI will do I am left with two choices...

1. Make a good move or..
2. Intentionally make poor moves.

Its that simple. It is not just a question of exploiting the AI.

Maybe an example or two would help. I guess I'm not seeing in my games the same certainty in AI moves that you do. Although my interior cities are much more lightly defended (one or two units perhaps) than border cities (potentially LOTS of troops), my AI civs are making reasonable attacks on border or near border cities rather than marching inwards. I can certainly see how this would be annoying if you need to put a bunch of troops inland just to stimulate the AI to "logically" attack a border post.

I should also say that I am seeing much better tactics from the AI recently -- taking a turn or two more to get to its target border city, but taking longer because its sticking to the high ground (even with its two move units). I have also seen the AI attack a fortified chokepoint en masse, when much more lightly defended opportunities presented themselves -- don't know why, but it was indeed a pleasure.

I am simply hoping that Firaxis will consider putting a little randomness in there. I just don't want to be sure what the AI will do. I want to be forced to plan for more possibilities and have the possibility to be caught short when I fail to do so. It would make the game more fun.

Oh, I definitely agree that making the AI a much more challenging, fearsome foe (although I'm not sure randomness will do it) would be great! But again, this thread started with a detailed exploit of breaking formations to cause the AI to continuously waste turns in futile movements -- that is what I was initially responding to and I pointed out a very easy fix that didn't even require any precious programming time.

Admittedly the thread morphed into more of an "AI is illogical in certain circumstances" discussion, but asking for a much more "intelligent" AI, especially in warfare, is quite different than asking for the yo-yo to be fixed. I have to believe that coding a phenomenal AI strategic and tactical warfare capability is a daunting (if not impossible) task -- we do send our real life warriors through years of study and training after all, and their tactics and strategies are still incessantly second guessed / studied by historians and others .

Originally posted by ShredZ
I live just north of the 49th, and let me tell you, if you so much as flinch down there, we're ALL over you!

Although the Pacific Northwest is lovely, I stay down here in the SF Bay area precisely becuase of this possibility [eventuality? ]

Catt

[nato]

jt and others I think you are blowing this way out of proportion.

Well, speaking for myself and my post here, you are right, its not a huge deal. Overall I am very impressed with the AI. It is so good at building armies and actually massing them into decent forces!

I just find the parade towards weakly defended cities a drawback. When you get to the point where you feel guilty defending your frontline cities well, even though its the logical thing to do, because you know the AI won't handle it well, its annoying.

Same with the thing I wrote about declaring war to help the AI out (because logically he should declare on me). Not using exploits is one thing, but second guessing yourself and working against yourself makes it un-gamelike.

Not a big deal, and inevitable with an AI I suppose ... just something I've noticed.