Why are people so quick to bandy around the term "cheat"?

[Moonsinger]

Originally posted by Thrawn05
Paragraph 4: I would pay $500 for a smart AI.

Very few people could afford $500 for a game. A lot of people don't even earn $500 a week. Even if I'm a milliionare, I would never pay $500 for a game. I would never be able to sleep at night if I spend $500 on a game while many people in the world don't even have enough money for food.

Besides, I have had more fun against computers then other people in MPs games, why?

Me too. I have more fun against the computers than any human players. For one thing, the computer doesn't abandon its game because it isn't doing well. And the computer doesn't get too emotional either. Sometimes the human players know that they couldn't win, so they go on a suicide mission to prevent me from winning. Their slogan usually is "if you hurt me, I will make sure that you are going down with me". There is no fun in that!

[DeepO]

Originally posted by Thrawn05
Paragraph 2: Mmm, both Chess and Civ are TBS. Both have units with certain abilities, both can capture/kill a unit. Basicly, Civ is just an advanced form of chess. In fact, all war games are an advanced form of chess. I wouldn't be surprised if the basic underlining AI code in Civ3 is similar to most chess software.

It's not about being the same genre, it is not about handling the same kind of units. Chess on a 100x100 board instead of on a 8x8 would also be impossible for AIs.

Let's calculate a little to show you what I mean. Algorithm: a simple Min Max tree, where every possible move and countermove gets calculated, and after a number of plys the end result is that move that will mean that whatever the opponent does, it is the best favorable thing to do. (This is a classical chess-AI, only there they use an extra technique called alpha-beta pruning).

For chess, you have around 8 possible moves (i.e. not forbidden, and not overly stupid things like moving your king into reach of a enemy piece) each turn. A tree 20 plys deep would consist of 8^40 leaves (end nodes), each of which have to be calculated. 8^40 ~= 10^36 . A big number, but with enough CPU power, and some pruning techniques this is within the current possibilities (e.g. DeepBue).

For Civ, it is a lot more difficult. you don't have 8 possible moves, you have, let's say, an average of 50 units that each have 10 possible moves. These alone will give 10^50 of combinations for 1 turn(!). Your opponents have the same kind of troops, so with 5 opponents, the total number of combinations can total 10^300 for one ply. If we build the tree again, 20 plys deep, this would get to 10^6000 combinations, which, even with pruning and all the world's computers in parrallel, is uncomputable.

Of course, the calculations here are rude approximations, but you can't say that I favored the Civ case: in more then one game, there are more then 50 troops running around, and, to make it easier, I didn't take into account that most of them have more then 10 different moves (e.g. with railroad each tile with RR is accessible, and can count as a move). Plus, the chance instead of the fixed behaviour of units in combat means that it should even be higher, and more spread out.

So, I hope this shows that you cannot, from an AI point of view, consider Chess and Civ of the same class. Plus, just upgrading your computer from a 500 MHz to a 2 GHz won't do any good either, that's just a factor 4, where you need a factor of 10^1000 to be noticable.

Another conclusion is that this also means that underlying AI of Civ cannot have the same principle as the one of chess, or it wouldn't get anywhere. Which brings us to our initial point: smart AI for Civ is almost impossible to build without either spending decades on its development, or without 'cheats'. I chose for the latter.

DeepO

[Meldor]

Let us also add into the mix that chess is never more complex for the AI than the mid game. The openings have more pieces, but more stupid moves, the end game has a lot less pieces and fewer good moves. In civ3, almost the entire game the for the AI, things are gietting more and more complex.
In chess, there are only 6 different units, you can't re-build units, or upgrade units, or research new units. There are no resources to capture in order to be able to use certain units, no luxes to make people happy, no wonders, no commerce, no cities to manage, etc. Each of these things adds to the complexity of the choices the AI must make. The research and othe factors limit the choices so as to make it a lot more managable, but it is still a huge task. And then, on top of everything else you through in a Rndom map of a different sizes, each combination needing slightly different strats. My hat is off to them for the job they did.

[TacticalGrace]

comparing chess to civ is like comparing apples and oranges

[Moonsinger]

DeepO: You did a great job in explaining it! Me lazy and don't really want to get involve with the Chess algorithm debate, because it's very time consuming. However, I think there is something Firaxis could do to make a better AI for civ3. For example, let's go back to Chess again The basic open moves in Chess are usually the same. In Chess Master, it usually starts the game with a sequence of moves randomly pick from the collection of famous Chess players in the world. Well, we could apply this same technique to civ3. Instead of program the AI to calculate all possible combination of moves which is impossible with today technology, we just program it to randomly emulate the moves of the world famous civ3 players (like you, Thrawn, or even me).

For example, the AI civ would randomly pick my strategy or your strategy out of a library of strategies from the world top players. I think the AI would become very deadly right there, no doubt! That's just a small window to what artificial intelligent is really all about. The AI is making its decision based on a library of knowledge, the bigger the library of knowledge, the more intelligent it may seem. Of course, if anyone really want to make a real AI for civ, they could try to make it learns the moves and the styles of the human players, especially when it's losing against them (which would prevent it from making the same stupid move again), and add that to its library. Of course, we (except Thrawn05) don't really expect Firaxis or anyone to program the AI that way since we couldn't afford $500 for the game.

[Moonsinger]

To Meldor: Whille I was typing up my long post and was answering phone call at the same time, your post got accept before me...since our ideas are similars, just want you to know that I wasn't repeating your idea.

[vmxa1]

Originally posted by TacticalGrace
comparing chess to civ is like comparing apples and oranges

More like apples to USA food distribution network.

[Moonsinger]

Originally posted by vmxa1
More like apples to USA food distribution network.

I'm not even aware that we have a food distribution network.

[DeepO]

Moonsinger: indeed, some opening strategies to pick from would maybe make the Civ AI more intelligent, but what makes you think these are not already included? I'm pretty sure that Firaxis has some good players as well, plus they have access to the specific game rules from the start.

As starting strategies vary wildly with the situation you are in, my guess is that one of a few recipes gets selected, and followed. It's the only way I know of that will give good results with a limited amount of time. After that, there will most likely be a differentation between some long term goal planning, and planning on how to fulfill a particular goal (rules, I'm sure. No MinMax tree here).

Other comments on Civ being even more complex than what I already described: sure. But I thought my example was bad enough, and I didn't want to make it worse Besides, from a MinMax point of view, a city is just a unit, which again has a certain number of choices it can make each turn. It would only make the number of combinations higher...

DeepO

[vmxa1]

Originally posted by Moonsinger


I'm not even aware that we have a food distribution network.

Ok maybe I should have said the food distribution system in the US. You know my grand kids point nits out when I say it is 6:30 instead of 6:28. I must have been wrong to think my meaning would have been understood. For the record I did not mean to say the gov runs the food system. They do howerver have one of their own, so I guess I was correct actually. Sorry

[Moonsinger]

Originally posted by DeepO
Moonsinger: indeed, some opening strategies to pick from would maybe make the Civ AI more intelligent, but what makes you think these are not already included? I'm pretty sure that Firaxis has some good players as well, plus they have access to the specific game rules from the start.

That's true. However, it doesn't hurt if they decide to include our strategies in the game. I wouldn't mind to collect a small royalty from this game. Think of it as a small reward for loyal fans who play civs day and night.

[Thrawn05]

Originally posted by DeepO
So, I hope this shows that you cannot, from an AI point of view, consider Chess and Civ of the same class. Plus, just upgrading your computer from a 500 MHz to a 2 GHz won't do any good either, that's just a factor 4, where you need a factor of 10^1000 to be noticable.

Another conclusion is that this also means that underlying AI of Civ cannot have the same principle as the one of chess, or it wouldn't get anywhere. Which brings us to our initial point: smart AI for Civ is almost impossible to build without either spending decades on its development, or without 'cheats'. I chose for the latter.

DeepO

I only said 2GHz because that's the best out there at this point. There is no difference between a 500MHz and a 1GHz or even 2GHz. HOWEVER, a 10GHz (which I am currently saving up on) has a noticable difference. By that time (I think in about 2 years), bus speeds will be at least twice as fast as the ones Intel has out now (~500MHz bus I think? not sure).

You basicly made my point. In order to make a "hard" AI, you need to allow it to cheat. You would think ever sing PONG, companies would be working together in order to make a smart AI class that can be applied as easily as DirectX (which I think a headache in itself anyway ).

Hey DeepO, lets get cracking, we'll make an AI with O(0) and make millions!

[Thrawn05]

Originally posted by Moonsinger


That's true. However, it doesn't hurt if they decide to include our strategies in the game. I wouldn't mind to collect a small royalty from this game. Think of it as a small reward for loyal fans who play civs day and night.

I've been asking Firaxis, that by the time Civ5 or Civ6 comes out (which in these seems bleak ), to release the code of Civ3. Rebellion released the AvPGE code within a year, and most of us modders have been happy ever since with that.


Although I've kept this on the back burner, since I considered the fact that all of the "hate this" threads that poped up in the begining of this year, and didn't want to add to the burden.

[DeepO]

Originally posted by Thrawn05
Hey DeepO, lets get cracking, we'll make an AI with O(0) and make millions!

Believe me, if I thought it would be possible, I would have started already If you ever get a serious idea on how to do it, you know where to drop me a note

DeepO

[Thrawn05]

Originally posted by Moonsinger


I'm not even aware that we have a food distribution network.

OT: USA basicly feeds the world... on paper. Most of the time it never gets there. Usualy stolen from local warlords.