Oh lord, it's hard to be humble... but you can help!

[Arrian]

Hmm... though I have only progressed to Monarch thus far, I have found that mostly I either 1) fail early or 2) win easily. The AI has real trouble with the industrial and modern ages.

The only game which gave me trouble at the end was my first successful monarch game (persian). I made the mistake of signing a few MPPs, and got sucked into a couple of wars. Wars which, from a strict military standpoint, I won. However, my MPP's kept me at war too long (basically, I had to stay in for 20 turns) so that war weariness kicked in HARD. It ruined my science rate, as I had to bump luxuries to 40%. Even then, my capitol blew up (nuc. plant) on turn 17 of war. The end wasn't in doubt, however. It was merely an annoyance.

My problem, however, is that I LOATHED the "gang up on the human" in Civ II. Not that it really hurt me at all, because the AI was really, really stupid, but I just didn't like the idea that the strongest civ couldn't have friends. It bothered me that I had never, in the whole game, attacked anyone and they all hated my guts. I think that the 2nd and 3rd strongest should probably align with each other (which I have seen happen) to take down #1 (which I haven't). Perhaps a tweak to overall AI agression might do the trick. But even then, as mentioned above, the AI just can't fight an intercontinental war. It will show up, bombard some things, and off-load some (often hopelessly obselete) units next to a city. I've never seen the AI pull off a real invasion.

There are two main problems with the AI that result in the "oh, this one's over" syndrome (usually sometime during the industrial age):

1) Bad choices by the AI on research (it should emphasize tech that will get it more tech via wonders -> astronomy, theory of gravity, theory of evolution, computers).

2) The later in the game it is (tech wise), the worse the AI is at war. It can have the same units you do, and you will thrash it.

I still enjoy the game very much, mind you. But I agree that it gets a bit tedious once you know you've essentially won, but must spend hours dealing with (at the very least) scores of workers cleaning pollution or (worse yet) having nothing to do.

-Arrian

[vmxa1]

Arrian I understand what you are saying. If you ever played Hearts, ganging up is the only way to go if you are not in first place. If you are in first place you try to put the one with the worst score out of the game so you win. If not, you do not want to keep dumping hearts or the queen on that person or the game ends and you lose. So you try to give points to the top dog and then the second one and so on. If the AI see you in first it should not care if you have been its best friend, you are now its mortal enemy. I had that last night with China. I saved them, but in the end they had to attack me once I became number one. Most strategy games have that period where the game is to be won or lost and once you get over that hump, you are never headed again. It is the best part of the game. some times I will quit once I am past that point as it is just perfunctory after that. I would rather start a new game unless I care about the score. I may care if it the first time I have play on that level, depends.

[jed]

Another idea, though I'm not home so I can't look at the editor to see if it is possible:

Perhaps you could create some new minor wonders for the AI to use. You make them very cheap to produce, and they become available when certain technologies are discovered. Of course, you don't ever construct them yourself, but assuming the AI will construct them, they will get boosts in the later part of the game when they need it.

[Pyrodrew]

Originally posted by David Weldon
The reason I think no civ bonuses, bad starting places, etc. only delay the ascension rather than prolong the period of conflict is as follows: even though their effects last all game long, that just means you need more cities or more improvements to achieve a critical mass. Let's say we're talking about commercial, and I can get 4 turn techs in the Industrial age with 20 towns. Without it I need 25 towns. Ok, so I have to wait until I have 5 more productive towns and then I'm in essentially the same position of dominance. Without those 5 towns I'm behind the AI, with them I'm guaranteed victory. Developing them is just a lot of tedious micro-management until they're ready to be useful. All this did was extend the 'underdog' period, not the 'competition' period.

If >= 5 more cities is endless dominance & < 5 more cities is "behind the AI" there is no "period of competition" at all in that scenario. We may have to agree to disagree here. It is my contention that something which limits your success "throughout the game" also limits your success within the "period of competition" & thus keep you there longer. Conversely, if you improved your Civ to have all 6 Civ Bonuses & made grasslands even better you would probably find the "period of competition" to decrease.

If you don't like the extra towns concept because of corruption, think "without universities... with universities..." or anything else along those lines.

But universities take time and/or $ to build & if you're in the "period of competition" that keeps you there longer until x universities are built. At any rate, although less effective on larger maps, I would suggest improving the palace. By significantly improving the palace's culture you lower the value of all future culture improvements. Suddenly adding a Cathedral doesn't extend your culture lead as much over the other AI Civs due to the early fixed palace culture everyone has. Yet, this might make a culture victory too easy (since we cannot edit the culture victory requirements - might have to turn off culture victory) & it doesn't really apply to production or science since those are based on the # of citizens in the city (those winning/leading will tend to have more citizens in their capital).

If you weaken those improvements which usually occur after the "period of competition" in the dominance phase, that would decrease the lead you have over the other Civs... it doesn't extend the "period of competition", but atleast your lead won't be as rediculous & large later. Example: lower factory production from +2 per square to +1 per square.

Originally posted by ArrianMy problem, however, is that I LOATHED the "gang up on the human" in Civ II. Not that it really hurt me at all, because the AI was really, really stupid, but I just didn't like the idea that the strongest civ couldn't have friends. It bothered me that I had never, in the whole game, attacked anyone and they all hated my guts.

It is NOT about "hating your guts", it is about trying to win. Likewise if we played the game Risk no matter how "nice" you were to me I would be stupid to let you (or help you) destroy the last other weak player, only to let you keep your huge superior lead & have you finish me off later. It's not about "friends"... it's about winning.

Nevertheless, I believe most of us agree an AI seeking to balance the power when 1 player takes a strong lead is best. Improving the AI so it doesn't build obsolete units sounds 2nd best.

[David Weldon]

If >= 5 more cities is endless dominance & < 5 more cities is "behind the AI" there is no "period of competition" at all in that scenario.

Perhaps that's the problem, there's virtually no period of competition. I don't like being in a huge hole and essentially counting on the stupidity of the AI to be able to even the score (the 'underdog' phase), so I don't really enjoy playing Deity or even Emperor all that much. I know a lot of people consider this to also be a period of competition, but I just get angry about all the B.S. I have to go through before I arrive at a "fair" point in the game, at which point I get to play hard and then win. It's a psychological thing, I guess. I have much more fun on Monarch, but then it's all over way too quickly.

Above Monarch, I don't go from loser to winner via 5 cities, I actually do it via Steam Power & Replacement Parts. I have my entire empire railroaded before I even get the next tech after it, and as soon as I have Infantry I invade with Cav/Inf/Artillery and raze 2 of the core cities of my main competitors. Game over in less than 10 total turns. I put up with anger for 10-15 hours of gameplay, and then feel like I finally have a fair shake for 1-2 hours, and then normally quit and start again. I guess my tactics could be shortening the competition phase, but I want to be able to play my best. *whaaaaaaaa....*

I'll definitely try slowing the game down, but I'm afraid it will just extend all of the phases by roughly proportional amounts, while not really changing their relative durations. Anyway, I think the ideas are good ones whether or not they help me with my little problem.

The palace idea, on the other hand, is brilliant! I'm not sure exactly what I'll edit, but it definitely has the power to help smaller civs stay more competitive. Are there any other things that aren't proportional to population that could be used in this way?

As a side note, does anyone know if the AI ever builds an FP?

[jed]

I had a chance to look at the editor, and here is my suggestion. As I understand it, you'd like to play on Monarch, but have the AI get stronger through the game to keep it competitive throughout. So why don't you take 4 improvements/small wonders/wonders that you won't miss (solar plant, coastal fortress, iron works, pentagon and cure for cancer are some possibilities that come to mind) and convert them into cheap factories for the AI to build. They would become available after acquiring a tech half-way through each age. They should cost 10 shields, give a 25% production bonus to the city, and cost zero to maintain. You could adjust the 25% to 50%, and move them around in time so as to adjust the timing of the AI handicap. Assuming the AI's are smart enough to build them, they will get a handicap that grows throughout the game. Of course, you must refrain from ever building them yourself.

[David Weldon]

Another good idea! Now we're gettin' somewhere...

I very much like the idea of adding bonuses to the AI as time goes on, rather than giving them all their bonus right at the beginning. I don't even feel like it's "cheating" too much because it's quite obvious that a human can manage his empire so as to increase production and science much better than the AI can. It's kind of like training wheels for the AI, but without giving them a period of absolute dominance in the beginning of the game. A much more elegant solution than the current Deity handicapping solution. Very nice.

Will the AI build new improvements if I make them available? Or will it even try to take advantage of modified improvements if I just edit a few as you suggested? Will it ever defend its cities with enough units? Ooops, sorry about that, not sure how that last one slipped through there...

I had some other ideas:

-Remove the railraod production/food bonus since humans seem to be more aware of it's importance than the AI is. Automated workers will continue to clear Jungle in border towns rather than go and RR the core cities. Stupid stupid stupid...

-Alternatively, reduce the cost of RR so that the AI can build a lot of them very quickly once it starts trying. This might let them get a good defensive network up to speed before I can mount a successful invasion.

-Increase the rush buy cost of shields (if possible) because I'm pretty darn sure that I make more use of it than the AI does. This would also have a greater effect in the later ages because that's when more gold is available to do this kind of thing. This might backfire by making unit upgrades more expensive, though. I'm not sure how that cost is determined.

-Give all AI civs "Commercial", and don't play a civ with that strength myself. This benefit will help the AI more when it has a larger empire (i.e. at least it is delayed until the late Ancient Age).

-I would suggest not playing Industrious, but I hate the massive micro-management required to use normal workers to improve everything. Any trait that saves 1/2 of my unit suffling is a great thing! Perhaps I can make "super workers" that are even better than Industrious workers for the AI. I'm a little worried that this might have a much larger affect on the game than intended, however...

Side Point:
I have been afraid to install the patch because of all the issues people are having. I would love to hear from anyone who has played with it a lot to see if it has any affect on the issues we've been discussing here....

[jed]

I haven't had a chance to test yet whether the AI will take advantage of modified improvements, but I think it should from what others have reported when they modified units. I'll test it tomorrow night.

Of course, using such a mod, if you happen to capture a city with an AI bonus "factory" in it, you must sell the "factory" immediately.

I would recommend installing the patch. I've only had limited experience with it (finishing a saved game) but it really seems to do no harm at any rate (except for the ugly pop number, but that's a minor issue for me.) As usual, the minority of people with problems is much louder than the majority that has none.

I believe that the AI might be less vulnerable to diplomatic exploits post-patch. The patch certainly got rid of the air superiority and precision bombing bugs (and other assorted bugs), as well as the IFE and palace hopping exploits. Also slightly changed some of the units to make them more sensible. I really wonder if it changed despotic pop-rushing in any way. I haven't had a chance to test that yet.

The most noticeable change from the patch is that techs have been jacked up in price, especially the later ones.

In the game I just finished, the AI's *did* gang up on me at the end (I was just trying to finish off a space race peacefully, at my son's insistence), but they were so far behind, that their attacks were not any serious threat. I only installed the patch half-way through the game, and by then the AI were too far behind. So I have some hope, but no really solid evidence, that the patch does improve the AI somewhat.

I'd love to hear from experienced players on what effects they think the patch has on the AI, based on games that started with the patch. I'm surprised we haven't heard more about this.

[TJW]

I'm sorry if i'm about to sound stupid but, to stop the AI building obsolete units, could you use the editor to make units like the longbowman upgradable to rifleman-or something like that, doesn't that simply eliminate the longbowman from the list of units it can build, meaning no more hordes of them in the modern age...

[David Weldon]

TJW:
I think you're right about letting things upgrade, but there are potential pitfalls: what if the AI or a human just wants some cheap MPs? What if the AI doesn't have a resource necessary for the upgraded unit? These will eventually be overcome, but I don't necessarily think it is an easy fix. In any case, this needs to be fixed up either through AI tweaks by Firaxis or through moding by us.

Jed:
I remember a chat transcript where Dan or Soren mentioned that buildings are of three types, and that one of the types is always destroyed when you take a city (notice there are never any temples in any town you capture). The other types have some chance of being destroyed or surviving. We could make the AI helper buildings of the first type, so that you never see them. You could also make them un-buildable by the civ you were about to play, so you didn't have to see it in your build queue all the time.

I'm not sure that increasing the cost of late tech will help here. It seems like that would just let the human with a tech edge have more turns of potential conquest before the AI gets the same tech to defend itself. I guess I'll try the patch, but I probably won't be able to play much in the next week or two...

[jed]

I've been thinking about this project of giving the AI late game boosts some more and was able to do just a little testing. So here are my latest comments.

First of all, I don't know how to make it so that the human can't see the fake improvements that I suggested meant only for the AI. Kind of annoying to always see them on your build list. I set one up to be available without any pre-requisites (for testing). The AI didn't build it right away like it should have (it cost zero); it seems like the AI is hard-wired to expand early on with no improvements. I'm not sure whether or not it would build them eventually--I was too impatient to keep testing.

But I had some other ideas which are perhaps simpler. On the difficulty levels page of the Editor, one can try to focus the advantages so that they are a little less for the AI in the early game, but stronger in the late game. Specifically, don't give too many starting units, because that's an early game advantage. Give more free support per city, because that's more of a late game advantage. Set government transition time low--that's a late game advantage. Cost factor of course is the big AI advantage, and I guess it's basically one that isn't weighted more to late game or early game. I would set number of content citizens to one just because I like the game more that way, but it is actually more of an early game AI bonus.

Now for the big one: percentage of optimal cities. Set that to half of what it currently is. I.e. if you're starting with emporer, make it 40 instead of 80. But to make the corruption bearable for the human, change the optimal city number in the world size tab to make it twice what it is now (32 instead of 16 for a standard map). This has the effect of making corruption about the same for you (although you now won't be able to build the forbidden palace until you've got 16 cities, instead of 8), but the corruption is much easier for the AI. This should be a big late-game boost for the AI.

Another thing to do: I am pretty sure the AI will build small wonders, even if they're modded a little. So first mod the Iron works (I never use it anyways since I never get the chance). Change it so that it becomes available with steam power automatically (no resource requirement). This should give the AI at least one strongly-productive city. Don't lower the shield requirement--we want to force the AI to build it in a productive city. Get rid of its pollution though. Of course, the human must not build it. Second, change the Pentagon into another forbidden palace (again available only to the AI) that costs very little to build. Make it's pre-requisite feudalism or some other middle-age tech. Hopefully, this will help the AIs build large but uncorrupt empires. And these are two small wonders I won't miss.

These mods should impact the rest of the game minimally. Of course, if you want to make other changes fine, but I think these are relatively simple ways to improve the late-game AI compared with the early-game AI.

[David Weldon]

Wow Jed, you really stepped up to the plate! Wonderful ideas there!

I have a question about the % optimal cities, from the way I read your post it implies that the AI doesn't suffer from this modifier. Is that true?

Also on the free support per city, I'm not convinced that it's much of a late game advantage. Generally by the late game 20-100 gold per turn won't mean a whole lot, but in the early game an extra 3-4 free units can really be big. I would absolutely prefer this to giving them free units to start, though, because at least that way they at least have to _build_ their bonus units.

I like the sm. wonder production increasers, but I'm a little worried about the FP. Has anyone ever confirmed that the AI even knows how and/or where to build an FP? I have never seen any proof that an FP has ever been built by the AI, and it seems like that would be a difficult thing for the AI to understand where to build. The whole trade off between building it in the middle of corrupt cities, but at the same time building it somewhere with enough production to not take forever seems difficult from my perspective.

In any case, I've been contributing some to Vel's mod, and I hope to get as much of this kind of stuff in as possible. Most of the work we are doing is aimed at general gameplay that everyone will experience, whereas this stuff is only really necessary at Emperor and Deity. It would suck if we added some things and then made chieftan too difficult in the late game or something...

I mention the mod just to let you know that I hope to steal some of your ideas... (even if they don't get in his mod, they'll certainly make it into my computer!)

[jed]

David, please steal away. I'm very intrigued by Vel's project myself and am tempted to join the effort. I'm a little skeptical about trying to change so many things at once though. It's hard to control what's going on when you change too many variables at once. So I'm going to make a simple mod to try to simply address the weak late-game AI issue (which now that you've raised it seems to me to be the biggest problem with Civ3). I'll post my mod here so people can discuss and test it. Hopefully we can eventually balance it so that the game is equally challenging throughout all the ages. My philosophy is different from Vel's mod though--change as little as possible while addressing the late-game AI issue.

About your questions:

1) Yes on the difficulty tab in the modifier, I am pretty certain that only the player suffers from the optimal city percentage, just as only the player gets only 1 content citizen on Emperor/Deity. By the way, I can't remember if this was adjustable pre-patch. I'm using 1.16.

2) I agree free support isn't much of a big deal.

3) I hope the AI knows how to use the FP. I'll look around in one of my saved games to check, and get back to you on that.

Anyways, I'll try making a simple mod soon and post it here.

By the way, on a separate subject, I've noticed from other posts that you're quite bothered by micromanagement involved with workers. I have a couple suggestions. First, you actually spend a lot more time clicking if you have your workers work in big teams. To illustrate the point, imagine that you have 10 workers to do 10 jobs that each take 10 turns. If you have them work in 10-man teams, you have to click on each one each turn for 10 turns, for a total of 100 clicks. If you have each one work separately, you just click on each one once at the beginning, and 10 turns later they're all done, for a total of 10 clicks. So don't work in teams if you don't like clicks. (Admittedly, team-work will tend to lead to slightly better play.) Secondly, after the patch I think the automate worker feature works better. I automate them after I've linked up my empire with a rail network, and nothing too terrible happens. Vastly improves the micromanagement.

[jed]

OK, I checked it out--I investigated a bunch of AI cities in an old finished game and found that they did indeed build the forbidden palace, along with wall street, battlefield medicine, & intelligence agency. Whether or not they build the FP wisely is another issue (the one I found sure wasn't exactly placed where I would have put it) but at least they built it. Giving them an extra one should help somewhat then.

[jed]

OK, I made a mod to try to boost the late game AI. I tried to keep it as simple as possible, changing as few things as I could. Don't know if it's properly balanced, but testing should determine that. Here's the changes I made in the editor:

1. In the world size page, doubled the corruption level for all world sizes. (E.g. for standard 16 changed to 32).

2. In the difficulty levels page, halved the optimum city percentage at every level. (E.g. for emperor 80 changed to 40).

The effect of these changes is to leave corruption based on number of cities unchanged for the human, but half as onerous for the AI. Of course, the Forbidden Palace will not be buildable until twice the number of cities as usual are built. E.g. on standard, one will need to build 16 cities before the FP is available, not just 8.

3. In the Improvements/Wonders page, I changed the Pentagon so that it is a zero-cost version of the Forbidden Palace. The only other difference is that I checked all the religious / militaristic / expansionistic / scientific / commercial / industrious attributes (instead of just religious). This has the effect of making the Pentagon automatically one of the required wonders to trigger a peaceful golden age. So it makes it considerably more likely that an AI will get a peaceful golden age after building the Pentagon. I was originally hoping to reduce the number of cities required to be built before the Pentagon was available, but I didn't see anything like that for the FP, and I don't think it's changeable. A little testing showed that the Pentagon will indeed now be available exactly when the FP is (i.e. at 16 cities on a standard map as modded above). Of course, the human player should not build the Pentagon! It's meant to help the AI!

4. Also in the Improvements/Wonders page, I changed the the Iron Works into an improvement with the following characteristics:
Steam Engine Prerequisite, 50% production boost (a 2 in that slot), 0 cost, 0 maintenance, -1 pollution, +50% science, taxes, and luxuries, +1 Happy face, +1 culture. (Edit: it also lowers corruption, and of course it doesn't require any resources to build.) Also, I checked all the expansionistic/scientific/etc. attributes. This basically guarantees that (assuming the AI builds the things, which I think it will) it will trigger a peaceful golden age by the early industrial period if it hasn't already had one. (I think it's a bad sign for an AI if it hasn't had a golden age yet, so this is a good way to make sure they all get one at a reasonable time.) Again, the human should not build the Iron Works! The Iron Works seems like a huge boost, and it is, but I wanted to put in something strong to make sure there was an effect. The -1 pollution makes the Iron Works an anti-pollution improvement.

The fact that the Iron Works and the Pentagon improve culture means that they will automatically be destroyed if you capture AI cities containing them.

Now possibly these changes will have the effect of making the late game harder than the early game. If so, we can always scale back. If the late game is still too easy, we can increase the effects. Anyways, I intend to test (although unfortunately not the next few days), and would appreciate hearing any comments.