Apolyton University Mod (PTW version)

[Dominae]

Great comments all-around. I'm especially wary of Communist AIs not being able to "come back" from their government choice (if only the AI could disband!). However, I think we should perhaps test an improved Communism on its merits alone, before considering any ill effects of switches. The AI is hard-coded to prefer Communism anyway (in wartime, which is almost all the time in most games), so giving it an extra fighting chance should be interesting. What I really want to see is an AI Communist engine push back a Republic human player, and eventually make him or her give in to WW. That would be great.


Dominae

[ducki]

Originally posted by Dominae
(in wartime, which is almost all the time in most games), Dominae

What difficulty level?
I'm not seeing this at all at Regent/Monarch.
I did see a lot of war in my recent Deity debut, but I only got to the Middle Ages before my accumulated "stupid mistakes" made me want to start over - things like leaving a new city undefended while at war, etc.
I'm actually seeing "punctuations" of war and peace with very few AI Civs maintaining a constant state of war(in my Regent/Monarch games).

Maybe it has to do with play parameters or difficulty level or human playstyle, but if so, I think the AU should try not to "fix" the AI for any given parameters or difficulty or playstyle, opting instead for the best "generalized" changes.

I vaguely recall some post mentioning that when the human is a warmonger it causes the AIs to war more often - war begets war, I imagine.
I tend to be peaceful unless a)I'm forced to war to expand or b)I'm presented with an opportunity that's too good to pass up - like an undefended AI town or woefully outclassed AI units, like warriors when I have Swordsmen.

Maybe you see more constant war simply because you war.

[Konquest02]

I second Ducki on the point of MP being a huge burden to the AI after war is over. I favor the new improvement more. To answer Player1 on the name, there is only 1 Forbidden Palace, 1 Wall Street, 1 Apollo Program, etc...

As for guerillas, we could make them a bit like the Keshnik of the Jungle (and/or Forest) and remove the movement penalty for them on this kind of terrain.

[Edit]: D'oh, Guerillas only have 1 movement point... but I still like my idea...!

--Kon--

[Dominae]

ducki, I agree that war begets war. But I've never seen (that I can recall) the AI switch to Communism when it is not involved in a war. This is good because the representative governments are better for peace. The point is that if war breaks out during the Industrial age (whether due to the human player or the AI), the presence of MPPs and loads of cash with which to form Alliances, everyone is involved eventually (your examples are not very convincing, by the way, because they're all pre-Communism anyway). In these cases the AI switches to Communism. We need to help the AI when it does so, and for the reasons that it chooses Communism. We cannot change the code to make the AI "like" Communism in the right circumstances; what we can do is make Communism "fit" the circumstances which cause the AI to select it already.


Dominae

[ducki]

Good points on my examples, but they were merely examples of how I tend to avoid war in most of my games - the examples were exceptions to that rule - hopeful illustrations of my attitude to early/ultra-early warfare and questioning whether the lack of "permanent" warfare in Industrial/Modern times for me was due to my playstyle, and whether/how that affects my opinions about Communism, MPs, and World Wars.

And my main worry is, since most of my games (in the age of Communism) are not all-out warfests, once the AI switches to Communism and beefs up his MPs, how much of a hindrance is the added military upkeep once he switches back to a peacetime government?

Maybe some good questions (for me to know the answers to) would be:
1) How long does it take the AI to switch back from Communism?
2) Do they switch back to Rep/Dem or do they always try to switch to whatever their "preferred" govt is? (Maybe we could change the "preferred" govt away from Communism for civs that favor it?)
3) Is the standing army fielded by the AI in wartime actually too large to support during peacetime?
4) What is the "real" cost of Communism to the AI, in terms of corruption, waste, wonders, tech, and anarchy? I know we've all seen the tests showing how much better Monarchy is for wartime, but it's difficult to imagine the real, overall "cost" of Communism in all those categories, over time, and dependent on how much war there actually is in any given game.

[Dominae]

Originally posted by ducki
once the AI switches to Communism and beefs up his MPs, how much of a hindrance is the added military upkeep once he switches back to a peacetime government?

I'm with you on this one (see above). My hope for now is that we can create a Communism that "puts up a fight", or that at least is not self-destructive. Currently, if the AI switches to Communism and the human player can resist the assault, then the human players has won because he or she knows that Communism is not the way to go. But, if the AI threatens the human player enough, WW could be a real problem.

The problem that I see here is with AIs on different continents. Since the AI is horrible at inter-continental warfare, getting it to switch to Communism is a game-winner. alexman's proposals address this by making Communism "not half bad" in comparison to Monarchy or Republic. But even this will not cut it, IMO, because the human player is still better off. Might as well make the AI scary in those cases where it can threaten the human player (i.e. same continent situations), and live with the fact that AIs on other continents can be exploited.

Originally posted by ducki
1) How long does it take the AI to switch back from Communism?

Good question. I don't know. But, I've seen Religious civs switch back and forth in almost perfect unison with the current state of war, so I doubt that AI gets "stuck" in Communism too long if there is no fightning to be done.

Originally posted by ducki
2) Do they switch back to Rep/Dem or do they always try to switch to whatever their "preferred" govt is? (Maybe we could change the "preferred" govt away from Communism for civs that favor it?)

The whole "preferred" government thing makes me uneasy. If Shaka "prefers" Despotism (which he does), and this manifests itself in how the Zulu AI plays the game, then we've got a very good reason why the Zulus aren't all that great. I know the AIs switch to "useful" governments eventually, but if their traits force them into sub-optimal government choices, well...I dunno. Obviously you want there to be variety in the game, but government choices are so important that I would tend to downplay that kind of variety.

Originally posted by ducki
3) Is the standing army fielded by the AI in wartime actually too large to support during peacetime?

Another good question that I would like to see some numbers on. Clearly the AI will never be as efficient as the human player, who can get by with a minimal military (the AI always builds units, because it's designed to be a "jack of all trades"). Ideally the AI will use its units when in Communism, and not be tricked into Democracy when "peace" is declared. But, these are things we have no control over.

Originally posted by ducki
4) What is the "real" cost of Communism to the AI, in terms of corruption, waste, wonders, tech, and anarchy? I know we've all seen the tests showing how much better Monarchy is for wartime, but it's difficult to imagine the real, overall "cost" of Communism in all those categories, over time, and dependent on how much war there actually is in any given game.

I guess you could do a case study to figure this out, but you would have to do many such studies to determine the trends under Communism. Unfortunately, I don't think you would learn much more than we already know: Communism currently stinks. Again, I say the best test is whether or not a human player can "take on" an equally-sized Communist AI. If the AI puts up a strong defense, we've got a better Communism; if the AI whimpers and crumbles after 10 turns, there's still some work to do.


Dominae

[alexman]

Great discussion guys. I don't have much time, but I want to make a couple of points.

-> Making Communism as good as Republic in peacetime is a big no-no. It would be very unbalancing, so Dominae you can rest assured that we will never let this happen. However, don't put Monarchy and the Republic together in the same basket. They are alternatives only in the Ancient Age. Once you get more than 3-4 luxuries hooked up, marketplaces built, and your cities become large enough to take advantage of the extra commerce per tile, there is no comparison. Monarchy is much worse in peacetime than the Republic. I believe we should we make Communism sometimes worth choosing over Monarchy when at war. This is the only way the AI will stay competitive when at war (since it never chooses Monarchy when Communism is available anyway). But Monarchy or the modified Communism would still not even come close to the income generated in Republic or Democracy.

->To answer ducki's question, the AI does not stay in Communism for long when it's no longer at war. Also, it switches to Communism only when it can no longer be in Democracy due to WW.

-> "Prefered" government means that the AI likes other civs better when they are in that government. So it affects attitude, and has nothing to do with their own choice of government. The same is true with the "shunned" government

I'm in the middle of doing some tests to observe AI behavior. Hopefully I will have some more information about these matters soon.

[player1]

I've seen that some AI civs avoid Communism and stay in Monarchy.

So, I'm curious.

What if you put preffered gov. for Monarchy and shunned to be Communism.

Will it make Monarchy gov. choice of war?

Anybody tested this?


P.S.
It's not solution, but if Monrachy is "so better", then make AI use it instead of Communism.

[ducki]

Originally posted by Dominae
But, if the AI threatens the human player enough, WW could be a real problem.

Might as well make the AI scary in those cases where it can threaten the human player

Ideally the AI will use its units when in Communism, and

Again, I say the best test is whether or not a human player can "take on" an equally-sized Communist AI. If the AI puts up a strong defense, we've got a better Communism; if the AI whimpers and crumbles after 10 turns, there's still some work to do.

Dominae

Sorry for excerpting so many quotes, but these, I think, finally made me realize why I don't like the Increased MP as the "main" fix.

It assumes that the AI will be on the defense.
(Edit: Your quotes don't assume this, but the specifics of MPs does - stationary resistance quellers and defenders.)
It requires that I, as the player, take the fight to the AI.
At least, for Increased MP to be "efficient", anyway.
If we assume the AI is the aggressor, then he's got to build twice as many units - the necessary MPs, and the requisite "invasion" force. Which, I think, will only make the post-war non-Commie switch an even worse leech on his bank account.

I know I don't have "the answer", I'm just pointing out what I perceive as a flaw in the "Increased MP" fix.


Is there an instructional thread on running debug mode? I think I might have a go at testing out the Kremlin, extra MPs and other Communism "fixes" next time the baby's sleeping.

Cheers!

[Dominae]

Originally posted by alexman
However, don't put Monarchy and the Republic together in the same basket. They are alternatives only in the Ancient Age. Once you get more than 3-4 luxuries hooked up, marketplaces built, and your cities become large enough to take advantage of the extra commerce per tile, there is no comparison. Monarchy is much worse in peacetime than the Republic. I believe we should we make Communism sometimes worth choosing over Monarchy when at war.

I'm sorry if I made it sound like Monarchy and the Republic should be considered together generally. I was only considering them as wartime alternatives. For extended war, Monarchy is better than the Republic, but for strictly military purposes, Communism should always be better than Monarchy (basically we're agreeing!).

alexman, I think the difference I see between our visions of an improved Communism is that you want Communism to be an economic alternative to Monarchy, while I do not. If a player wants the "best" of both worlds (economy and war), Monarchy is fine; if they want a strong economy, the Republic and Democracy are fine; and if they want to fight, Communism shoud be fine. I think you're proposing that Communism should also be a useful empire-building government too (i.e. by causing WLTKD in all cities so that Corruption is reduced).

An interesting solution would be to raise the aggression level of the AI when under Communism. It's a wartime government, so might as well make them bloodthirsty. Again, this is unfortunately not possible with the Editor.


Dominae

[Dominae]

Originally posted by ducki
Sorry for excerpting so many quotes, but these, I think, finally made me realize why I don't like the Increased MP as the "main" fix.

It assumes that the AI will be on the defense.

When the AI is at war, it "thinks" both offensively and defensively simultaneously. Raising the Military Police would have the effect of keeping the cities happier, regardless of if the AI is being attacked or not. The human player can play with the Luxury slider when on the offensive and leave cities largely undefended (although this may not be optimal), but that is not designed to do this. Also, the fact that the AI is able to draft units that it already uses offensively (Guerillas and Infantry) means that its "defensive" efforts will potentially be put to use offensively as well.

So, although I admit I was thinking in terms of the human player being the aggressor, the changes I'm proposing are not affected by this assumption.

By the way, the increased Military Police is only a part of the "solution". Here's what I'm thinking about for Communism:

1. Increased Military Police to 5 or 6.

2. All Spies and Diplomates are Elites.

3. Immunity to various forms of Espionage.

4. Increased modifiers for Propaganda.

5. Increased Draft Limit to 3 or 4.

6. Increased number of Free Units per City, etc. by 2.

7. Increased Worker Rate to 4.

The specific numbers are just ideas. Note that the "feel" of Communism would remain the same, but the AI would be much stronger when choosing it.


Dominae

[MiloMilo]

Regarding communism: this is attacking the problem obliquely, but how about reducing the likelihood that the AI will need to switch to communism in the first place, by reducing war weariness? This can be done by giving, say, temples the 'reduce war weariness' ability. This way the AI could withstand a 20-turn war without having to switch to communism.

Obviously this will help the human player, but maybe it will help the AI more?

[MiloMilo]

Another idea (I'm just full of them today! ):

How about giving factories the 'reduce corruption' ability of courthouses? There tend to be more factories in core cities, so it will have a negligible effect on corruption there. But when a civ switches to communism, those important core cities will have an instant bonus against the increased corruption.

A side effect may be that the AI will see factories as being a better deal, and they will be earlier on the AI build queue. (I mention this because it was mentioned on the 'Poor AI' thread.)

For balance, maybe take the ability away from police stations.

[ducki]

I already never need to leave Democracy or Republic. though I have had to sue for peace a couple of turns early once.

I kinda think that WW reduction would benefit the "Grey Side" players like me much more than any other playstyle or AI.

Grey Side - one who realises the power of the Dark Side(warmongering) but refuses to inhibit his Builder tendencies, thus becoming a belligerent builder, as opposed to a true warmonger or true builder.

[Nor Me]

The problem with war weariness is that the human is far better at combatting it than the AI. Better strategy means less casualties and leaving the luxury slider on 60 percent for a few(maybe 10?) turns in republic is often better than switching.
If I were to do it, I would change Democratic WW to low. But then I'd switch to Democracy as soon as I'd researched it and stay there for the rest of the game. Still it would help the AI.

[Nor Me]

There are non-Grey Side players out there?

[ducki]

I actually still like to play White, and still haven't really learned to embrace Dark.


I was actually thinking of experimenting with being a rabid warmonger as Carthage.

[Nor Me]

I'm still a builder at heart. But I've learned to see a horseman rush as just another way of building.

[Catt]

I spent some time running an experiment this morning. I set up a debug game on an editor-generated standard map (middle of the road all options), regent level, random civs, with the human starting on a one-tile island. The rules had been modified to: (1) provide communism with an MP limit of 12; and (2) start everyone in the Industrial Age (didn't help to shorten the process). Using the ol' beer can on the spacebar testing method, I hoped to see how the AI handled the modified communism.

It was only one test; it turned out to be a game without war until relatively late (almost tanks, and not until 1300 AD +), and the war concluded quickly enough that I didn't have an AI in communism for hundreds and hundreds of years -- an AI ended up building the UN and winning, shortly after the final combatant had been eliminated. Nonetheless, I got to see some behavior that leads me to make a tentative conclusion.

Contrary to my expectations, the AI did not seem to make a concerted attempt to build up and garrison cities with up to the maximum MP limit (12). Instead, the AI seemed to be content with a modest (by AI standards) garrison, even if a few additional units would eliminate some waste. See the screenshot below -- it's not the best example, but it's one I happened to grab. The city in question currently has 4 MPs but several unhappy citizens. It suffers from a modest amount of waste. Not pictured in the screenshot are nearby (and RR'd) areas of the AI empire that show numerous units "hanging out" -- i.e., not guarding radar towers but rather moving from turn to turn (i.e., "AI patrolling"). You can just make out that a nearby city is building a courthouse instead of a unit (although it too suffered waste, only had 4 or 5 MPs, and didn't enjoy WLTKD). This situation was not unique -- I saw many instances where the AI was not pumping out units (was in fact building various city improvements) to reach available MP levels, as I expected it would, both during the war and after its conclusion.

Again, although only one limited example, it certainly produced a result quite different from what I expected, and, if one sample is to be trusted, seemed to imply that raising the MP limit will only have incidental effects -- if the AI is cranking out units but is unable to use the units (is having to make intercontinental invasions and transport them with a limited navy, for example) and chooses to garrison cities rather than "patrol" as it does so often, then maybe an increased MP will be utilized. I simply saw nothing that would imply that an increased MP availability induces markedly different behavior from an unmodified MP level -- and therefore that a modified MP level may not be a real help to the AI (even if for wildly different reasons than I hypothesized).

Catt

PS: I saw some other interesting things which I'm posting in the "Why, oh why . . ." thread.

[Dominae]

Catt, you test provides some information that is definitely good to know. Most importantly, it shows that the AI will not use Military Police to enter a WLTKD. This is a definite shortcoming, but not a major problem. As long as the AI uses MP efficiently to alleviate unhappiness, everything is ok. We're trying to help the AI here, not make it "smart" (which is impossible for us to do, really...).


Dominae

[Pyrodrew]

Why even have any AI Civs build artillery at all?

Rarely (if ever), do I find a case where artillery in a city actually helps an AI more than another defensive unit would have. The AI building artillery in cities is like putting extra candy in cities for the human player when he/she conquers those cities. It only makes the game easier. Flagging 'Never Build' for artillery would "help the AI". If the AI could build ships, but never be able to move ships out of the city it was built in... would we want the AI to still be building ships?

[Dominae]

Making Artillery-type units have 1 Defense means they cannot be captured. Consequently, the AI can use them liberally (however ineffectively they do so), without the human player exploiting the weakness of being able to easily capture stacks of AI Artillery. Without "teaching" the AI to use Artillery effectively, I think this is the closest thing to a solution to the Artillery "problem" (i.e. that humans use Bombard effectively while AIs do not).


Dominae

[player1]

Of course it makes human's Artillery attacks much easier.

If you lose Arty defense, they can't be captured (and not even destroyed without having at lest as many unit as human has artilleries).

[Catt]

Originally posted by Dominae
Catt, you test provides some information that is definitely good to know. Most importantly, it shows that the AI will not use Military Police to enter a WLTKD. This is a definite shortcoming, but not a major problem. As long as the AI uses MP efficiently to alleviate unhappiness, everything is ok.

My concern is that I'm not convinced that the AI uses MP efficiently to alleviate unhappiness. I think it uses whatever units happen to be available, but will not "consciously" take advantage of an increased MP level -- in other words, the number of units garrisoned in each city is determined based primarily (if not totally) on other factors. A screenshot below shows a Persian city from my test game (modded with up to 12 MP) -- despite numerous units "patrolling" the homeland, far from any fighting (which is all on another landmass), only four units are garrisoned in Tarsus, which is using 3 entertainers. Three additional units would eliminate the need for entertainers, and well more than three additional units are quite readily available.

In other words, I think the otherwise interesting proposal to increase MP limits under communism will have only incidental and minor effects, because the AI (from what I can see in one sample) will not take advantage of the increased MP limits.

Catt

[Pyrodrew]

this is the closest thing to a solution to the Artillery "problem"

Well, I would still say just remove the "problem" and not let the AI build them. Most (all?) of the AI's use of artillery consists of only sitting in city. A 10 defense unit would be better than a 1.
Nevertheless, I am impressed with the creativeness & I guess it keeps things "as much like the original as possible" while at the same time "helping the AI."

[MiloMilo]

Suggestion re: artillery--

In addition to giving them a defense of 1, I've lowered their hit points, so a regular has only one hp and a veteran has two. It makes just about zero difference, but it comes closest to making them indefensible-but-not-capturable, which I think is the goal.

This is contrary to everything written so far, but what of the price of artillery was lowered, so the AI would think they are more valuable and produce more of them? Maybe having more in the field would induce the AI to use them more effectively?

What if they were flagged as offensive? Would they then bombard units instead of tile improvements?

Don't citizens, buildings, and tile improvements have some kind of inherent defense value? Maybe if these values were made to be higher than 10 then an AI artillery would see an infantry unit as a better target? I know we can't mod this it's just a thought.

[Mad Bomber]

Originally posted by MiloMilo
Suggestion re: artillery--

In addition to giving them a defense of 1, I've lowered their hit points, so a regular has only one hp and a veteran has two. It makes just about zero difference, but it comes closest to making them indefensible-but-not-capturable, which I think is the goal.

This is contrary to everything written so far, but what of the price of artillery was lowered, so the AI would think they are more valuable and produce more of them? Maybe having more in the field would induce the AI to use them more effectively?

What if they were flagged as offensive? Would they then bombard units instead of tile improvements?

Don't citizens, buildings, and tile improvements have some kind of inherent defense value? Maybe if these values were made to be higher than 10 then an AI artillery would see an infantry unit as a better target? I know we can't mod this it's just a thought.

The defense of buildings and Citizens are able to be modded, its just that the values are too high as they are (16 as of v1.29f). Your idea of flagging artillery as offensive and defensive has merit as I have seen others report that the AI will use them offensively if flagged as such. As for the defense value, I am thinking that 2 or 3 may be more appropriate, but I have yet to test this in a mod.

[alexman]

Well, so much for the idea to make ships carry cruise missiles...

The plan works well for the human, but the AI "cruise missile" strategy does not allow loading the missiles, even if there are units available to carry them.

So what happens is that the "naval transport" ship goes to get the missile, but the missile doesn't want to load into the ship. The ship stays in the city waiting, and the cruise missile ignores it.

Apparently a missile needs to be a nuke in order for the AI to know to load it in a ship. Oh well, look at the bright side... one less change to the mod!

[alexman]

Also, the Small Wonder implementation of the Kremlin does not work as expected.

Even though it does not show up in the list of improvements in the city, it actually continues to work even after the civ switches out of Communism.

[alexman]

It's worth testing whether flagging artillery as AI "offense" instead of "artillery" will do the trick.

I recently tried something similar with ironclads, giving them a low attack and a high bombard strength, and the AI just didn't attack with them often enough. It just ran away. But it's worth another test.