AI spawning troops

[Wildcat]

I have noticed that the AI (0n Noble) can field huge armies way out of proportion to their empire size or the time frame of the game. In one game I was overrun before I could even produce my third settler, by the Mongols who fielded 8 chariots and 4 axmen in their advance force. When considering the early date and passage of time in the game this should have been impossible. Well today I have personally witness this troop spawning. In today's game the Russians declared war around 500 BC or so, after killing off there advanced force I approached their last two cities, one size 4 and one size 7 they contained 2 archers each. one turn later each had 3 new chariots. after another couple of turns 6 more spawned and the cities remain size 4 and 7.

I thought the AI was not suppose to cheat on Noble and was not suppose to get free troops except at start on any level. So WTF is going on.

[Hauptman]

was "see enemy moves" on?

pehaps they had them elsewhere and just moved them into the cities.

Yes I have noticed the ai gets armies MUCH larger than mine... but then there cities are poo, so i figure they build units, then some more units, and a few more units before they build that granary. Even on noble the AI gets an upgrade cost bonus. so those 5 warriors he made, get upgraded into 5 axemen even when he's poor. then 5 macemen, then 5 grenadiers etc... its also possible the AI uses that to crank out large armies fast (1 turn to build a charriot vs 3 turns for a horse archer, then blam upgrade.)

Even on monarch I've NEVER seen the AI civs poof units out of thin air. But i have seen superstacks (...stacks) made by a civ with 2 cities...

[Wildcat]

Oh, I forgot to mention they were cut off on a peninsular with no troops around except for those inside the walls.

[Medic911]

although i don't know anything about your credibility, i can't believe the pre-mentioned scenario. BUT, i have experienced, in previous civ versions, the ai cranking out about a dozen units, in 2 turns, with 4 cities in pre-draft tech eras. the computer players can do so much more than any human BEFORE you add all the cheating and stuff, so i think that in terms of techs and units and stuff, the AI should be able to do only what we could. The big factors in higher difficulties should be better tactics and every citizen, worker, and other units are completely utilized to their full potential to maximize whatever the computer needs (at high difficulty). at lower difficulty, the cpu should have lower and lower abilities/reaction times to the maximization of city/citizen and worker potential.

[opaque]

I have seen 2 or 3 units suddenly appear in a city I was preparing to attack (this is on noble), but I assumed these were being relocated from other cities.

[Cort Haus]

Do you have a save anytime leading up to that manifestation, Wildcat? If you had it could be investigated.

OT: Medic, is that a typo in your sig?

[MightyTiny]

Originally posted by Medic911
although i don't know anything about your credibility, i can't believe the pre-mentioned scenario. BUT, i have experienced, in previous civ versions, the ai cranking out about a dozen units, in 2 turns, with 4 cities in pre-draft tech eras. the computer players can do so much more than any human BEFORE you add all the cheating and stuff, so i think that in terms of techs and units and stuff, the AI should be able to do only what we could. The big factors in higher difficulties should be better tactics and every citizen, worker, and other units are completely utilized to their full potential to maximize whatever the computer needs (at high difficulty). at lower difficulty, the cpu should have lower and lower abilities/reaction times to the maximization of city/citizen and worker potential.

I agree, I hate it when higher difficulty levels in games are achieved by changing the rules for the AI's. The only games where the AI actually works better at higher difficulty levels seem to be Chess games.

Of course I guess it's much harder to make several levels of AI quality, and especially to make a tough AI opponent without letting them cheat, but I'd like to see someone tackle the challenge...

By the way, with the patch, there came a "no cheating" option in the custom games - I wonder if checking this means that the AI's also have to play with the same rules?

[Cort Haus]

Originally posted by MightyTiny


I agree, I hate it when higher difficulty levels in games are achieved by changing the rules for the AI's. The only games where the AI actually works better at higher difficulty levels seem to be Chess games.

Of course I guess it's much harder to make several levels of AI quality, and especially to make a tough AI opponent without letting them cheat, but I'd like to see someone tackle the challenge...

Think about it - Civ is a thousand times more complicated than chess. Think map size, tile types, unit types, and then the hundred things that a chess AI doesn't have to think about that the Civ AI must. If you want a more level playing field, try making your moves in the same time the AI is allowed.

We all want a better AI, but although it has some glaring weaknesses, I think this one does a lot of things well. Perhaps the release of the SDK will see some innovation here.

There MUST be AI cheats to compensate for the fact that a 2005 game AI is not yet science fiction, and cannot match a human brain. That's not surprising is it? Yet the nature and balance of those cheats/bonuses/handicaps is a matter for discussion. My view is that the difficulty level gradation is a bit steep in places, and many players feel that some AI benefits are a bit crude & imbalancing for their tastes. eg - very cheap upgrades and tech dogpiling against the human.

[Cort Haus]

Another thing is that the 'right' sort of AI bonus depends on playstyle. An aggressive player will want the AI to always have enough units for a good tussle, otherwise they feel shortchanged. A more passive playstyle might resent the free 'bullying' units for the AI.

So, whatever the designers do they'll be damned by someone, and will probably get "this level is too easy/too hard" depending on how people play it.

[MightyTiny]

Originally posted by Cort Haus


Think about it - Civ is a thousand times more complicated than chess. Think map size, tile types, unit types, and then the hundred things that a chess AI doesn't have to think about that the Civ AI must. If you want a more level playing field, try making your moves in the same time the AI is allowed.
.....

There MUST be AI cheats to compensate for the fact that a 2005 game AI is not yet science fiction, and cannot match a human brain. That's not surprising is it? Yet the nature and balance of those cheats/bonuses/handicaps is a matter for discussion. My view is that the difficulty level gradation is a bit steep in places, and many players feel that some AI benefits are a bit crude & imbalancing for their tastes. eg - very cheap upgrades and tech dogpiling against the human.

I totally agree with you Cort on the regrettable necessity of doing it the way it's done today - but it still doesn't mean I have to like it.

I have actually studied a bit of computer science, and had a course on AI, and the subject intrests me a great deal. I am well aware of the formidable challenges in creating a good strategy game AI. Chess is easy in this regard, because it has such a small set of precisely defined basic rules, and a small, standard "playing field".

But I can't help feeling that enough investment/effort isn't put into AI development in games in general; that's probably because a good AI mainly manifests itself in the LACK of idiocy, and thus isn't an easily marketable selling point like cool graphics. Also, with the hardware side developing at a mad pace, actually getting a game stable on all configurations seems to be a formidable challenge that probably takes up a great chunk of time and funds in game development.

Imagine if, lets say, for the next 10 years, computer hardware development was artificially stagnated - so that no hardware vendor would be allowed to release any new models before that 10 years was done.

After maybe 3 years they'd have done all the development that could be done graphics-wise without better hardware, and by this point I bet stability problems would also be in the past, with well tried and tested game engines and recycled code being used... then, finally, they'd HAVE to shift the focus on playability and AI development, as it would be the only really significant thing left to compete with. And I bet that before that ten years was done, we'd have strategy games with good enough AI's to allow quite high difficulty levels without giving the AI freebees or allowing it to cheat.

You'd probably have great path-finding in RPG's too... (another difficult problem, and the one thing that always sucks, in every game I've played.)

[Cort Haus]

Originally posted by MightyTiny

But I can't help feeling that enough investment/effort isn't put into AI development in games in general; that's probably because a good AI mainly manifests itself in the LACK of idiocy, and thus isn't an easily marketable selling point like cool graphics. Also, with the hardware side developing at a mad pace, actually getting a game stable on all configurations seems to be a formidable challenge that probably takes up a great chunk of time and funds in game development.

Imagine if, lets say, for the next 10 years, computer hardware development was artificially stagnated - so that no hardware vendor would be allowed to release any new models before that 10 years was done.

Yes, the economics do not exactly encourage the prioritisation of 'unmarketable' features like AI. Platform stability is unlikely - perhaps we need State Property implemented in the world to allow production for need (better AI) rather than for profit (better graphics).

* strikes Lenin pose

[Wildcat]

I should have the auto saves, latter today I'll save them to new files and then try to reproduce the spawning.

I do not mine AI difficulty levels (cheats) and long as the conditions are known before hand and they are reasonable. But I have noticed that the AI does not seem to play by our rules at all, have you noticed the early large size of some of the empires while they still maintain high research levels, the huge Armies, and the constant flow of Great people. All this on Noble the supposed cheat free level there is something definitely screwy going on.

[quote:Originally posted by Medic911 although i don't know anything about your credibility, i can't believe the pre-mentioned scenario.]

Observe your own games whether you believe your own eyes as to the AI troop builds is your own problem. There is definitely something fishy going on and along with all the other Civ4 problems it points to poor and rushed programming.

[ghen]

The only time I've seen the AI spawn units is in cities with no defenders about to get attacked. The code for this is actually not about fighting humans but staying alive against barbarians better.

The AI does NOT spawn units in any other case, it builds them just like you and I. On some high difficulties they get bonuses towards production, and on all difficulties they have better upgrade prices.. So the axemen were almost certainly upgraded warriors, and chariots can be cranked out in a capital city with ease. The AI just builds tons of units.

[TheHateMale]

For AI, could there be an (optional) time to plan strategies? I.E. every five/ten/whatever turns the AI turn would take about (an additional) 10 minutes, but they could plan strategies for the next set of turns. Possibly two strategies - one for if they're left alone, and one for if they have to fight.

If you could plan for it, you could use this time to... I dunno. Go to the bathroom, eat, swear out your window at those damnéd teenagers in their hoodies...

[Arrian]

The AI gets really big discounts on upgrades...

-Arrian