A Better AI.

[Wodan]

Have you guys noticed that the AI often founds new cities during wartime? Ecspecially just a few turns after you declare war on them.

If there were a routine that told the AI not to found a city within X turns of someone declaring war on them, I think that would be helpful. X could be 5, 7, or 10, not sure what is best.

It just seems like it is never a good idea to found a city within 5 turns of someone declaring war on you.

I wouldn't say that. If they already had a settler on its way, or had a settler close to completion and he doesn't feel the need to stop production everywhere to pump out units, then I'd say it's better to go ahead and start the new city.

Unless you're saying he's founding a new city where you (his new enemy) can easily conquer it? Which is an entirely different problem, that of selecting the city site. If, OTOH, he founded the city in the opposite direction from you, I don't see a huge problem there.

Wodan

 

[Wodan]

Why, did you run it like I said and didn't have it crash at all? If that's the case, then maybe I could try something else.

No, I didn't run it. I just found it strange that you made no mention whatsoever as to what the problem was. Thanks for the explanation.

Wodan

 

[Wodan]

Blake,

Here's something I just noticed which I think could use some attention. Take a look at this screenshot. Germany has 4 cities. I spammed caravels and have just finished scouting the whole area (I could probably have loaded WB to take the screenshot, sorry, didn't think about it).

Anyway, what Germany did was found his cities on all the 2 food resource spots he could reach. He has so far ignored strategic resources (except for the one iron which happened to be in one of the 2 food spots). He has also ignored luxury resources and single instances of food.

So, he has ignored Elephants, Dye, Horses, and the single instance of Rice anywhere on the whole map.

He has also ignored distance-from-capitol penalties. He is dead butt-last on research so he's probably being cripped by this. He is a little isolated by the map so that probably didn't help. Only recently has he been contacted by the other AIs.

Anyway, seems to me there is too high of a priority on the 2 food resource spots.

I admit this is somewhat subjective, but if it was me, I would have immediately settled the Elephants/Rice, followed by the Iron. And, for the iron, I would have settled on the other island, even though it would mean only having 1 Fish instead of 2 Fish... the payoff is by having usable hills for some hammers instead of a total coast/ocean city.

(Horses on island map I would have left until much later, even if it meant I wouldn't get it, so that's the one that I agree with.)

Wodan

 

[civzombie]

"I wouldn't say that. If they already had a settler on its way, or had a settler close to completion and he doesn't feel the need to stop production everywhere to pump out units, then I'd say it's better to go ahead and start the new city.

Unless you're saying he's founding a new city where you (his new enemy) can easily conquer it? Which is an entirely different problem, that of selecting the city site. If, OTOH, he founded the city in the opposite direction from you, I don't see a huge problem there."

Yes, he's founding a new city where I can easily conquer it.

My solution was not intended to be optimal, only hopefully easy to code. The solution you describe is certainly better, but probably harder to program.

 

[kniteowl]

Blake,

Here's something I just noticed which I think could use some attention. Take a look at this screenshot. Germany has 4 cities. I spammed caravels and have just finished scouting the whole area (I could probably have loaded WB to take the screenshot, sorry, didn't think about it).

Anyway, what Germany did was found his cities on all the 2 food resource spots he could reach. He has so far ignored strategic resources (except for the one iron which happened to be in one of the 2 food spots). He has also ignored luxury resources and single instances of food.

So, he has ignored Elephants, Dye, Horses, and the single instance of Rice anywhere on the whole map.

He has also ignored distance-from-capitol penalties. He is dead butt-last on research so he's probably being cripped by this. He is a little isolated by the map so that probably didn't help. Only recently has he been contacted by the other AIs.

Anyway, seems to me there is too high of a priority on the 2 food resource spots.

I admit this is somewhat subjective, but if it was me, I would have immediately settled the Elephants/Rice, followed by the Iron. And, for the iron, I would have settled on the other island, even though it would mean only having 1 Fish instead of 2 Fish... the payoff is by having usable hills for some hammers instead of a total coast/ocean city.

(Horses on island map I would have left until much later, even if it meant I wouldn't get it, so that's the one that I agree with.)

Wodan

I'm Assuming you're playing on an Archipelago map, The AI has always been bad on Archipelago maps, I Don't think Blake's AI improvements were meant for Archipelago Maps, I think he improved the AI for more standard Games Settings. Eg - Continents.

 

[a4phantom]

Thanks Blake.

 

[Mergle]

The problem is, is that these boni get exponentionally bigger / more important as time goes by. That makes each difficulty setting disproportionate from each other.

The low levels are just too easy. I consider Noble the 'normal' difficulty setting (don't know if the AI still gets any bonus here). Prince is however, much more difficult than Noble especially if you're not a war mongerer and are on a standard to large map. This also partly accounts for the situation where you are often quickly overtaken technologically at around 1600 - 1700 AD while they were far far behind you a couple of decades before.

The trouble isn't that people don't want to play below their level, but the trouble is mainly that the next step is too big. When playing on Noble you can get away with anything, if you want a peaceful win you can. But on Prince you must start war asap and knock out your main rivals before they knock you out.

I just won a game on Prince with Blake's latest AI. I never fought a battle except against some animals early on and to take one barbarian city. (I built the great wall to help against those pesky barbs!). I had an excellent starting position at the end of a pangaea map and was able to play exclusively peacefully as, of all people, Montezuma! I was a cultural powerhouse, and could have easily won by culture but did it by spaceship in the end. I didn't do an excessive buildup of force, but I did a *lot* of diplomacy, tech-trading and gifting, whilst staying neutral in the conflicts raging around me. Interestingly, my closest AI rival was probably Ghandi who despite being half conquered by both Tokagawa (who was #2 to me in score and #1 for most of the game) and Saladin at different times (actually being Tok's vassal till Saladin beat him [ghandi] up allowing him to break free!) raced up the tech tree at a comparable pace to me.

So up to there at least, no need for war.

A possible solution is to give the AI a bonus that either grows with time (at the easy levels) or shrinks with time (at the levels Prince and up), but only on standard maps or bigger. On small maps the bonus should not be affected by time.

I jsut don't see the need to fiddle with the difficulty levels yet. It may be that when Blake's done it might be possible to make Noble truely "equal" by removing the (few() boni the AI get on that level, but beyond that I can't really see the need.

 

[Wodan]

I'm Assuming you're playing on an Archipelago map, The AI has always been bad on Archipelago maps, I Don't think Blake's AI improvements were meant for Archipelago Maps, I think he improved the AI for more standard Games Settings. Eg - Continents.

Regardless if true, in no way does that change my feedback.

Thanks, however, for the comment.

Wodan

 

[DrewBledsoe]

I just won a game on Prince with Blake's latest AI...............
I jsut don't see the need to fiddle with the difficulty levels yet. It may be that when Blake's done it might be possible to make Noble truely "equal" by removing the (few() boni the AI get on that level, but beyond that I can't really see the need.

No, all the pretty decent Monarch lv players who have repeatedly reported stuff like the Industrial Era being entered in around the 14th Century AD, are just lying of course and should be berated, and at best ignored. The game's completely and utterly balanced, and anyone who says otherwise must go and hang their head in shame, as they aren't even fit to play settler lvl these days

Yeah right.....

 

[Mergle]

No, all the pretty decent Monarch lv players who have repeatedly reported stuff like the Industrial Era being entered in around the 14th Century AD, are just lying of course and should be berated, and at best ignored. The game's completely and utterly balanced, and anyone who says otherwise must go and hang their head in shame, as they aren't even fit to play settler lvl these days

Yeah right.....

In the context you cut, you would see that I was replying to someone who said that it was only possible to win on Prince by being aggressive and taking someone else out early. That's not true - my counterexample proves it untrue.

I said nothing about Monarch (which is the level I usually play btw - I dropped down one to test some things out). It does appear that the AI boosts have reduced the ability to play a purely peaceful game on Monarch - I've had to fight a couple of defensive wars on my last game. In that sense, the "balance" between war- and builder- strategies has shifted. But no sign of needing to do rampant conquest.

Your point on industrial age in 14th century seems irrelevant to me. Of course, if you have a better AI with the same bonuses, it will research tech faster than before (if it doesn't it's not really better is it?) If you don't want the AI to have those bonuses...drop down a level .

Current noble doesn't have any really significant AI bonuses. If the AI was as good as you, you would expect to win 1 in n games where n is the number of players. So if you usually play on standard maps, that's about 14% win rate. I suspect many people would see themselves currently as needing to drop down a level if they only won 1 in 7 games, but perhaps they could look at it in that light.

I honestly don't see why so many people are getting worked up over the difficulty levels. No offense, but I'd rather Blake devoted his full attention to making the AI the best he can rather than worrying about what boni the AI should get on what level. The ideal is a noble (with perhaps its minor boni removed) being the greatest challenge it can be.

 

[Harrier]

I honestly don't see why so many people are getting worked up over the difficulty levels. No offense, but I'd rather Blake devoted his full attention to making the AI the best he can rather than worrying about what boni the AI should get on what level. The ideal is a noble (with perhaps its minor boni removed) being the greatest challenge it can be.

I agree with you on this point - Blake should concentrate on improving the AI ( a never ending task - but thanks Blake for attempting it).

What myself and I think others are saying, is that it would be great if we could have one or two new levels around about the old Noble level, that makes it a more progress leap when going up levels.

Dropping down from Noble a level is not currently right - because of the extra bonuses the human player has at the lower level ( plus happiness/health etc)) I personnally do not want those. A lot of players say that the current Noble (with version 208) plays as the old Prince Level plays. And a lot of players found it difficult to move from Noble to Prince, even in vanilla Civ4.

So what is wrong with suggesting we have one or two interim levels between Warlord, Noble and Prince. Where AI bonuses are slightly reduced to make the transition betwen levels smoother.

Keep the higher levels as they are for the gamemasters, if that is what is wanted. Just make it slightly easier for others to progress up the levels.

Of course it is obvious this fine tweaking of AI bonuses can only be done when the AI tweaking has been completed. When ever that maybe.

 

[aelf]

I honestly don't see why so many people are getting worked up over the difficulty levels. No offense, but I'd rather Blake devoted his full attention to making the AI the best he can rather than worrying about what boni the AI should get on what level.

I don't think anyone ever said that Blake should stop and meditate on the handicaps first. Firaxis should be the one doing the adjustments, if they are interested in keeping the game balanced. For my part, I originally made my comments because the thread is called "A Better AI" and I thought I'd share about the negative (and temporary) effects that this has brought to the game in its current version. I don't know why opponents of AI bonus adjustments like to implicitly accuse the other party of rejecting the AI improvements.

You are not concerned because you are playing on Prince and find that you can still win somewhat peacefully. However, thinking outside of yourself, don't you feel that there's something wrong with the game if it ceases to be 'Civilization' and becomes 'Warmongerisation' from Monarch onwards? Some need for war is good, but in the game's current state, it seems that the emphasis has shifted towards constant war. In 6000 years of human history, sure, there has been lots of room for war. But the game only has a limited number of turns to represent that 6000-year period. Having to fight wars for most of the few hundred turns you have, while conforming somewhat to historical tendency, does not make for a game that has balanced war and empire-building elements, which I believe Civilization has always been intended to be.

 

[Iustus]

New downloads, both Warlords and Vanilla are posted today, 11/17/2006 on the sourceforge downloads page

Overall Strategy:
- Cultural Victory: the AI now may decide to try for this victory type
- Dagger: sometimes, the AI make make a big military effort somewhat early

Military Strategy:
- larger stacks should be more likely
- units are split off to heal while the main stack moves on
- quite a few minor changes

Military Tactics:
- when attacking, with low odds, sacrifice units are sent first
- The AI will pillage a lot less.
- When sieging cities, the AI will attack with all of it's units if it thinks it can win.
- AIs use promotions more sensibly, they will try to make a medic in large stacks and are less likely to use "conflicting" counter-promotions (ie Shock+Cover)

AI Tech Preferences
- Techs that give buildings should be valued more highly than before
- Meditation and Polytheism are closer in value, which one an AI chooses to do first will be more random
- AIs will prefer techs that give them their UUs and UBs

Worker AI:
- automated workers correctly will improve over city ruins
- all known bugs with city ruins and 'leave old improvements' fixed
- other minor tweaks

Diplomacy:
- AIs will be less likely to accept peace when they have a large advantage in land area

City AI:
- AIs will try to spawn a prophet in their holy city if they do not have a shrine yet

Governor changes:
- the default case of the governor should now be one you usually use
- food has a higher priority when cities need to grow
- if you pick emphasize, it will be more likely to do what you wanted
- in general, emphasize nothing is a better choice than emp food/prod/commerce
- emph commerce will also emphasize gold (also pick emphasize research if thats why you want commerce)
- emph great people rate should actually do what you expect
- emph research should work commerce+food tiles when appropriate
- the governor knows how close you are to the happy cap of the city

Cheat mode: (= chipotle, to enable)
- lots for AI debugging, including messages created for all AI trading (see how much or little the AI really trades)

Enjoy!

-Iustus

 

[Mergle]

I don't think anyone ever said that Blake should stop and meditate on the handicaps first. Firaxis should be the one doing the adjustments, if they are interested in keeping the game balanced. For my part, I originally made my comments because the thread is called "A Better AI" and I thought I'd share about the negative (and temporary) effects that this has brought to the game in its current version. I don't know why opponents of AI bonus adjustments like to implicitly accuse the other party of rejecting the AI improvements.

I guess because if you improve the AI's intelligence it is inevitable that a pure building strategy would be less viable. You're no longer going to be able to outdistance the AI by so much simply by building if it can do the building nearly as well as you. So if the AI is improved, a shift towards more war is inevitable if you hope to win. At present, the shift is even more marked because the AI is still weak at warmongering while the building is improved, so war-heavy strategies will be favoured.

You are not concerned because you are playing on Prince and find that you can still win somewhat peacefully. However, thinking outside of yourself, don't you feel that there's something wrong with the game if it ceases to be 'Civilization' and becomes 'Warmongerisation' from Monarch onwards? Some need for war is good, but in the game's current state, it seems that the emphasis has shifted towards constant war. In 6000 years of human history, sure, there has been lots of room for war. But the game only has a limited number of turns to represent that 6000-year period. Having to fight wars for most of the few hundred turns you have, while conforming somewhat to historical tendency, does not make for a game that has balanced war and empire-building elements, which I believe Civilization has always been intended to be.

I usually play monarch and by temperment am an extreme builder. Personally, I'm enjoying having my low-army strategy more challenged. But I honestly can't see why people finding monarch etc. too warlike can't drop down levels. Obviously, if you crank AI bonuses up it's going to distort the game. It shows how relatively weak the original AI is that the bonuses they get in unmodded Monarch don't make them unstoppable. And yes, I don't want to play a game that is so slanted it eg forces war strategies (that's why I don't play diety unmodded). And yes, Blake's mod above monarch is pretty much forcing war -based strategies for the reasons given above - just like unmodded does above Emperor.

But it's just about finding the sweet spot where AI bonuses are enough to compensate for its failings. That sweet spot will obviously vary with the player, and with Blakes improvements it will lie at a lower difficulty level than before. Higher levels will become redundant except for a very very few who like a hopelessly slanted game.

(I think Harrier UK's points about Warlord-Noble are well-made, though - that's about getting the *player's* bonuses improved against a stronger AI).

I'm not going to comment further on difficulty levels as I think it's detracting from the real purpose of this thread.

 

[Doomed_UK]

and there was me about to report a bug in the last version - I will download this one and test it on my save first.


(English had built pyramids, adopted representation early, later on they adopted universal sufferage, as soon as they changed to sufferage the game crashs to desktop when I press end turn if I use the .dll, works fine without it.)

 

[Arlborn]

"- Cultural Victory: the AI now may decide to try for this victory type"

Blake, did I already told you how much I love you? THX for putting effort on it only because I insisted

Going to play right now I just won a cultural victory with your other version in noble and was thinking in go back to Prince(my normal level), but now humm, let me try Noble again

I will put a mix of enemies that are peaceful(cultural?) with others that are aggressive(dagger?)

 

[Arlborn]

**EDIT** Aff, is it what he meant with the think of the debug mode(Cheat mode: (= chipotle, to enable)
- lots for AI debugging, including messages created for all AI trading (see how much or little the AI really trades))? LOL..Why not do that avaible only when you actually press Z to activate the cheat mode??? I have that cheat mode on for a Long time and only used once lol

Is this a bug?? I didnt even meet them yet!! Look the screenshot:



Never saw that menssage before use your last version Blake.

ANd this:



0.o

 

[cymru_man]

Take the cheat code off then you twerp

 

[Arlborn]

Take the cheat code off then you twerp

Ye I did, but mind you this is in my civilization for months and I only used 1 for fun, I even completely forgot about it..

But the thing is, isnt it better then to only show that messages when the code is actually activeted? That is pressing ctrl+Z..

 

[kettyo]

A whole lot of thanks to Blake for responding that much to the player comments like the problem of too weak AI stacks (my complaint) and various AI winning strategies (awesome)