How To Kill a Big Army

[dexters]

Coracle has an AI? where can i download it?

[Txurce]

Theseus, your game has been certified as officially Wild.

Two questions: how do you know how many bad guys are in each stack?

And how do you feel about ultra-early war, given your present standing? What did it get you?

[Alkis]

Another thing you can do is create a new city for the AI to attack. This way you don't risk anything. I was attacked in one game by the French (I was playing Germans) who brought an army of infantry/cavalry/longbowmen of no less than 141 units.

So, I found a new city and placed 11 veteran infantries in it. However, I was proved too optimistic because they managed to destroy it. They lost 40 units in doing so. Then I made another city. I always have settlers ready, sometimes just sitting there for ages of time. I put 14 infantries this time, literally bringing alomst every spare infantry I had. This time they didn't manage to get the city and they started retreating. Meanwhile, my own cavalry was getting one city after the other. They where lightly defended and most of them didn't even had infantry defending them. All that was on Emperor.

Btw, (I don't want to offend anyone but) I don't find the AI especially strong. Come on, we live in an age where a computer program has beaten the World champion in chess, they could surely make an AI better than that.

[DeepO]

Originally posted by Alkis
Btw, (I don't want to offend anyone but) I don't find the AI especially strong. Come on, we live in an age where a computer program has beaten the World champion in chess, they could surely make an AI better than that.

Sorry to go off topic, but this just cries for a response. Without a comment on this games AI (which I find relatively well done), comparing it to Deep Blue is a big joke. You see, in the world of AI, Deep Blue is notorious for killing an otherwise nice benchmark, and Kasparov was very right in his critisisms. Deep Blue was a marvel of computer architecture, not of clever programming. In fact, it's programming was less clever than what you'll find in the Chessmaster x000 series.

They just used raw power to calculate every possible move, 23 turns deep IIRC. No human can do this, as they prune the possible solutions. Kasparov is famous for being able to 'see' 17 turns deep. But human players take shortcuts, use rules of thumbs, see patterns in a play, Deep Blue just did all. It was the dumbest chess program ever made.

It killed the AI benchmark, by 'cheating'... so the next game benchmarks are based on Go (which has far more possible moves at each turn), where Deep Blue could only go 10 turns deep, where good human players see 15 or more.

If you compare this to CivIII, with its hundreds, maybe thousands of possible moves each turn, Deep Blue could plan 3, maybe 4 turns in advance. The current AI uses a lot longer planning, and humans maybe even more (planning for a spaceship victory, or where best to put your Forbidden Palace takes maybe 100 turns in advance) So please, don't use the argument of winning a chess game to compare it with a more complex game like CivIII. If you want the same response time, you should have thousands of Deep Blue's just to play one game... but you can be sure you'd loose, even with production bonuses for you, instead for the AI

DeepO
(BTW, no offence taken, and none intended either )

[Alkis]

Deep0, as you already said, you are off topic. I could get into a detailed discussion about Deep Blue because I am a chessplayer but anyway I will leave that (because it's of topic )

On the subject of Civ3's AI they could easily program it to research only techs who lead in the contraction of the spaceship (not recycling).
Also, they could enter a command like, if enemy tanks etc can approach your cities in one turn, then put equal number of defenders. Or something like that. The result would be that instead of attacking with 100+ units it would attack with maybe 60, but have its cities a lot more defended.

Just two simple examples.

[Meldor]

AI Stacks

Playing a game as Egypt. The Persian were next to me. Instead of a direct attack on my best deffended cities, they came around through France. They had one stack with 100+ rifles and infantry, and a second stack with 50, mostly rifles and archers. I had RoP with both, and scouts stationed to watch the moves. Once Persia got near my border (France had not completed their rail net yet) I killed the RoP with Persia forcing them to advance one square at a time. That gave me three turns to prepare. I stacked all my atry out of their range with Cav cover, and moved infantry claose to my nearest two cities. Once the stack was two squares into my lands I asked them to move, they declare war. Bombed stack with arty and moved all defnders into city they were headed for. Big stack split in two (one with injured making for the border, second for another undefended city). Bomobed them two more turns, and had main stack now down to 75+ units. Still too hot to handle. Took all the tanks I had (I had just got them) and started attacking the Persian homeland. Paid France to come in on my side. I took two Persian cities really quick. The two mega stacks, turned to try and make it back to defend the homeland. Didn't do them any good. They were both toast before they made it past the first few French squares, and since most of the Persian units were caught out of their lands, taking their cities was a lot easier. Would up with 8 good Persian cities.

There were two keys to this strat. Killing the RoP they had with the French by bringing them into the war, and attacking their homeland with enough force to make them recall the foreign guard. One interesting highlight was that I saw a stack that had about 50 units in it pass right by one of my cities that only had 10 defenders. The AI can not handle a two front war. It tries to concentrate all of its power in one spot. This works no matter what level you play on.

As they say, "the best defense is a good offensive".

[DeepO]

Even if Civ's AI could be a bit more reflective on how to use it's units instead of going for one city with all of them, it doesn't change the fact that the AI of Civ is a lot smarter then Deep Blue. Sure, an AI recipe to win a space race should have been included, now it is far too easy to win as a human.

But that's my whole point: we use recipes because we can't foresee all possibilities, so we devise a strategy once, and adapt it to the specific situation. An AI can't do this on it's own, so if it gets such recipes, such strategies coded by the programmer and if it can use them well we tend to think the AI is smart.

To take the comparison back to chess: In chess the possibilities are far less then in Civ. each turn, you have only 8-10 valid moves you can make, with which the opponent again can have 8-10 countermoves. Humans can't see all moves 20 turns deep, and have to use strategies, studied openings with known results, counter techniques and the like. Deep Blue was able to have not one single strategy included, but through raw power win anyway.

In Civ, that would not be possible, because you don't have 8 possible moves, each unit has over 8 possible moves (fortifying and upgrading can be counted as possible actions too). The complexity is so overwhelming, that you need to have a recipe/strategy like AI, raw power wouldn't cut it. Of course, once the player understands these recipes used, it becomes easy to counter them, as the AI can't adapt like we can. So you fool them... as a dog or a little child.

That's why I feel the AI is well done in Civ. Of course it will always be possible to be smarter then the AI, of course it uses some strategies at the wrong places / times, but in general it gives you a run for your money. And having such a vast army appearing at your borders surely is one of those examples where we have to plan carefully, or be destroyed. As such, it's fun and exiting even if it could always improve.

DeepO

[Sava]

Good plan! Although the AI usually keeps spitting units out. What do you do when the next wave reaches you? I guess what I would do is build two forces. One that would take care of existing units, and one that would focus on destroying their ability to make war.