The AI cheats on Noble, correct?

[Blake]

Originally posted by nugog




However Blake you are brilliant at this game, and some of us, like me, are not so brilliant

I don't see how this logic extends to
"I suck at the game"
"Noble AI sucks less than me"
"Therefore Noble AI must cheat"

[Nugog]

Originally posted by Blake

I don't see how this logic extends to
"I suck at the game"
"Noble AI sucks less than me"
"Therefore Noble AI must cheat"

Well truth be told, it probably doesn't.

It was more of a "I have no bloody idea how he does it" kinda comment/reaction.

[Prussia]

The AI probably does cheat somewhat, if only to base it's strategy's on information it wouldn't have, such as maps and locations.

But if the AI weren't cheating, it would be like playing against a drunk monkey playing truth or dare with a troll.

Yeah, wouldn't be too much fun.

[Arrian]

The AI "knows the map" on all levels, and gets large discounts on the cost of unit upgrades (even on Noble they pay something like half price). I think that's about it.

The first significant AI bonus shows up on Monarch (the free worker to start).

-Arrian

[NFIH]

Originally posted by Urban Ranger


You have some advantages and the computer players have some other advantages. If you want to know the exact settings take a peek at HandicapInfo.xml.

I don't know how to interpret all the numbers there since I have no baselines to compare them to and anyway wouldn't know which numbers are significant and which are not.

The advantages are very slight, not sufficient to give them the sort of headstart over you that you described.

For example in 920AD you should have at least 4 or 5 cities. Maybe you could post a saved game over at the strategy section so others could take a look at your game.

Yes, that's what I had--four cities. What I question is how the AI could have not only more cities than that (and far more spread out as well, to mention the maintenance penalty) but also have such relatively large armies. At that particular point in the game, I couldn't have four or five cities AND a large army. It would have to be one or the other. But my point is the AI appeared to have both as many cities, a large army and was also keeping up with me in research. Sounds fishy to me.

[NFIH]

Originally posted by Blake


So the AI has managed to get construction by 920AD? That is an impressive feat!... not. A competent human can have construction (elephants and pults) by 500BC (1000BC if they really try), along with supporting techs like Bronze Working, Pottery etc and maybe some religions for good measure like say Theology so they can all pop out at 6exp.

How do you think poor noble level Alexander AI feels when I hit him with a stack like this at just before 1AD? How big do you think my army would be by 920AD?


A skilled human will destroy noble AI's like they aint even there. A skill human will demolish Monarch AI's. A skilled human playing Inca will take down Deity AI's.
This is actually all due to one big human cheat.
The human has A Plan.

But I don't hear Alexander cry "Planning is cheating!".

Anyway the Noble level AI gets a few bonuses and penalties, the largest bonuses are the barb-combat and the lower unit upgrade costs.
But they pay more in unit upkeep. I think there's another upkeep they might pay more for too, possibly civic.

OK, but how many cities did you have relative to the AI? You had as many or more cities than the AI and a large army and a tech lead?

[NFIH]

Originally posted by Arrian
The AI "knows the map" on all levels, and gets large discounts on the cost of unit upgrades (even on Noble they pay something like half price). I think that's about it.

The first significant AI bonus shows up on Monarch (the free worker to start).

-Arrian

Well, this is all I'm asking about, based on observations from worldbuilder. No idea what Blake's getting so high and mighty about.

[Arrian]

He's being heavy-handed, yes, but he's essentially right. The AI, while good in comparison to early Civ AIs, is still a moron. The advantages on noble are really quite minor.

The main question isn't really how the AI had whatever it had in 980AD... but rather why you only had 4 or five cities and a tiny army.

The best way to look into that is to post a save (4000bc, and then a later one like 1000bc) and ask folks here to have a look and give you tips to improve your play.

-Arrian

[padillah]

Originally posted by Urban Ranger
OTOH, a CPU in a modern PC can perform billions of instructions a second, but can only do things in sequence and only one at a time*. This makes it exceedingly good at crunching numbers. But crunching numbers do not translate well into the kind of fuzzy, simultaneous thinking required of strategy.


* Not exactly true anymore, but still works very well as an approximation.

If you are talkingabout parallel processing then you are still correct in your initial assertion. When humans think of more than one thing we have the ability to relate the two items, computers do not.

IOW, if I asked you to multiply 4*4 and 4*5 you would think (or should think) "4*4 = 16 so one more 4 is 20". You see the relation between the items and use that relation to influence your processing.

A computer would simply do the first problem and then the second problem and give you the results of each.

Techincally it's parrallel processing but it still doesn't have near the networking and associative abilities of a 6 year-old.

It's this abstraction that we have over computers, not processing tricks.

Another example: say we are using a weighted learning neural network (the closest thing to replicating human thinking thus far). The AI is chugging along and builds a few national wonders. It still would take quite a number of very specific failures to make the association that Heroic Epic and West Point should be built in the same city to maximize their effect.

It would need to fail in every scenario when it didn't do so and win in the scenario where it did in order for the weight to overcome simple random determination.

Even parallel processing isn't going to help unless they tie the processes together in some meaningful way.

Even learning is not that difficult for a computer, not anymore. It's association that reigns supreme.

Tom P.

[padillah]

Originally posted by NFIH


OK, but how many cities did you have relative to the AI? You had as many or more cities than the AI and a large army and a tech lead?

Knowing Blake...yes.

Tom P.

[Arrian]

Most likely he's even in cities, ahead in military, and has a "branch" tech lead. In other words he probably is ahead in one area (Construction, for instance) but behind in others. He will exploit that advantage, though, with his stack of 'pults and Oliphants

-Arrian

[Jaybe]

But they pay more in unit upkeep.

I thank Blake for bringing that to my attention. Paying only half as much for troop maintenance as the AI at Noble is not for me.

[Harovan]

Since the chess analogy was pulled:

I'm an old school chess player, I play from my childhood on and throughout my whole life, although not anymore actively (I played tournaments in the 70's and 80's). At the same time I always was very interested in computers. I played against the first chess program I saw in the late 70's IIRC. It sucked and I won both games easily.

But in the years after chess programs got better and better. The first chess program that beat me was Colossus for the Atari XL. Granted, I played far from tournament strength (and IIRC was even slightly drunk) but nonetheless I must admit that I lost, and badly so. I lost a lot more games against computers afterwards and with the today available professional chess programs I have to turn down difficulty to even have a chance.

I also wrote my own chess program in the early 90's just to have an idea how they work. It was written in 8086 assembler and, well, not too strong, but it could play and even win against me if I was not careful. Right now I am writing a chess module for Winboard in C++. Not that there aren't already enough Winboard modules out there to play, but just for sheit and giggles and to compare my module with the modules of others, to tweak my evaluation function etc. etc. It's great fun.

You see, I have seen a lot of chess programs. But I never -ever- saw a chess program, that put itself in material advantage just to make up for the lack of a real brain. I would outright deny to play against a chess program, that plays with an additional queen or something like this, or that can pull a knight out of thin air if it needs it or similar.

Chess programs are strong because they can brute force through a lot of positions in little time. This way they make up for the human advantages, which are his eyes, his ability to think in schemes like typical attacks and checkmate positions, his intuition and feeling of danger.

This however is not - at least not yet - possible in games like Civ4. It has vastly more squares on the "board", a lot more pieces, a lot more difficult rules, it has tile types (the chess board knows only black and white and game-wise there is no difference) and it has a lot more opponents than just the blacks and the whites. In mainstream Civ we unfortunately have to live with a "cheating" AI for at least 10 more years.

Dreaming is allowed, though. I would kill - well, not really, but pay a lot of money as opposed to the measly ~$40 Civ4 cost - for a TBS in the same quality, that would not need to cheat.

[MasterDave]

If a game developer made an AI that didn't "cheat", then we would all scream that the game is too easy! AI is a pushover!! AI never attacks!!!! (oh wait, thats Civ CTP).

Its a fine line for a developer to walk, but I would much rather have a challenging AI than one that never "cheats". So long as they do not do stuff like build three new units in a city the turn I arrive, or consistently win battles at terrible odds, I am fine with an AI on noble level having production or maintenence cost advantages.

I will say that I consistently get my ass chewed up on Emperor level, even as the Inca. I think I have won twice on Emp since the latest patch (nerfing chopping hurt my early game big time ).

[Arrian]

I actually think it will be pretty funny if and when someone finally produces a smart Civ AI that beats humans w/o cheating.

There will be much wailing and gnashing of teeth.

-Arrian