Ideas for a killer AI

[alexman]

In the thread titled "Why, oh why does the poor AI...", we identified some shortcomings of the AI in managing its cities. Until these issues are possibly addressed in a future patch, and in preparation for the PTW version of the AU mod, let us use this thread to agree on the best way to improve the AI using the current PTW editor.

So here are my proposals:

1. The AI builds settlers before the city is large enough to produce them: This happens VERY often on Emperor and above. Only the most food-rich cities get to grow past size 3 before the settler is complete in the AI's expansion phase. This more than negates the early-game production bonus on that level. Option 1: Do nothing. The AI is good enough in the early game without improving it. Option 2: Increase Settler cost to 40 shields. This will not affect the AI on Emperor, (since it already wastes at least 10 shields per settler), but it will delay the human expansion.
   
2. The AI emphasizes growth even when it has happiness problems. This causes the AI to assign entertainers, losing productivity, and irrigating more tiles to support those entertainers, losing even more productivity. Option 1: Double the happiness generated by entertainers. If we can't make the AI not use entertainers, we should at least make it use less of them. I personally believe that the human player's strategy will not be seriously affected by this change, but the AI will benefit greatly. For a more detailed debate on this subject, see this thread. Option 2: Never flag "Growth" as a build-often preference for the AI. The AI already builds Granaries when it has hess than two excess food without this preference. Building Granaries earlier would make its population growth even more out of control. Option 3: Make the city governor emphasize production for all civs. This couldn't hurt, but I have not seen any effect from making this change.
   
3. The AI doesn't build harbors early enough. It often can't benefit from luxury trade without harbors. Option 1: Mark "Trade" as a build-often preference for all AI civs.
   
4. The AI doesn't build factories early enough. Even with "Production" as a build-often preference, the AI builds Factories after cultural buildings like Cathedrals and Research labs. Option 1: Mark "Production" as a build-often preference for all civs. Don't mark "Culture" as a build-often preference, since the AI already emphasizes culture over production. Option 2: Increase the maintenance cost for Factories and Plants. The AI does not consider maintenance in its build decisions. By increasing the maintenance of Factories, the AI will at least save some money from building them later.
   
5. The AI builds too many units and not enough improvements. The real problem is that the AI NEEDS all these units because it's so bad in combat tactics. Option 1: Similar solution as with the Factories: increase the maintenance cost advanced buildings like banks, research labs, and cathedrals. The AI generally builds these later than the human, so we can save the AI some cash for doing so.
   
6. The AI is predictable in its research path. Actually, there is a random factor involved, but it is often outweighed by other non-random factors. The human can bet on being the first to get techs like Mathematics, Literature, Polytheism, Printing Press, et cetera, and sell them to the AI for a fortune. Similarly, the human can rely on the AI to research techs like Nationalism and Bronze Working for him. Option 1: Adjust the tech costs to make them appear more balanced to the AI. This is tricky, and requires some playtesting, but in general it involves making empty techs cheaper and Wonder/Unit techs more expensive. Note that the AI does not look forward when choosing the next tech to research. So it there is a great tech (like Electronics) after a lame one (Atomic Theory), the AI will not know to research the lame one so that it can get to the great one faster.
   
7. The AI builds Guerillas when Infantry is available. Credit to Catt for spotting this one. We think it's because Infantry and Guerillas have the same attack and cost, and both have the AI offense flag checked, so the AI sometimes builds guerillas as offensive units. Option 1: Increase the attack factor of Infantry to always be better than that of guerillas.
   
8. The AI attacks with weird units. The AI attacks with units that have the AI offense flag. Riflemen, Infantry, Mech. Infantry.... Option 1: Increase the attack strength of those units. This would at least make them more effective against the human. Option 2: Remove the AI offense flag from those units. Then the AI would not attack with them unless its odds are realy good. For the AU Mod, we used option 1 for riflemen and infantry and option 2 for Mech. Infantry.
   
9. The Babylonians are too agressive for a builder civ. The Iroquois don't take advantage of the best UU in the game. This one is simple. Lower the aggression of Hammurabi to 3 from 4. Increase the aggression of Hiawatha from 2 to 3 and flag his offensive units as a build-often preference.
   
10. The AI doesn't know how to produce units in its core cities and disband them in the corrupt cities. It often uses Wealth when its good cities have nothing else to build. Option 1: Double the value of Wealth. It is still not worth building for us humans, most of the time.
    
11. The AI doesn't know how to effectively place its Forbidden Palace. The AI suffers from high corruption. Option 1: Reduce the corruption level for the AI, by making it play at Warlord level instead of at Regent. This AI Level affects content citizens, their combat bonus against barbarians, and percentage of OCN for corruption. It doesn't directly affect research or production. Option 2: Reduce the AI level to Chieftain! Option 3: Make Chieftain unavailable to humans and make the AI play there. Change this level so that it has 200% optimal number of cities (to simulate the FP bonus), and the rest like Regent.
    
    
12. The AI uses Communism instead of Monarchy as a Wartime Government. Communism is far worse than Monarchy in most cases. The problem is the high communal corruption. Option 1: Increase the free unit support under Communism. Option 2: Add free building maintenance for Communism. Option 3: Increase worker speed for Communism. The trick here is to help the AI when it switches to Communism, without giving such a boost to this government type that it will become a viable option for the human player.

Any other ideas? Different solutions? Comments?
Sorry if you've seen all this before, I just wanted to get all these AI issues together somewhere.

[alva]

Maybe you should contact Korn on this, I know he has done some extensive and excellent testing on these things.
Especially the settler item (shield & food cost)

[Jaybe]

The AI attacks with weird units. The AI attacks with units that have the AI offense flag. Riflemen, Infantry, Mech. Infantry.... Option 2: Remove the AI offense flag from those units. Then the AI would not attack with them unless its odds are realy good. For the AU Mod, we used option 1 for riflemen and infantry and option 2 for Mech. Infantry.

Option 3:Create an additional strategy function for AI units: Offensive Support. Used as defensive cover to stack with offensive units. These support units do not attack unless odds are favorable, and will normally WITHDRAW to friendly territory (or closest friendly units) when offensive companions are gone.

[alexman]

Re: Ideas for a killer AI

Jaybe, thanks for your input, but...

Originally posted by alexman
...let us use this thread to agree on the best way to improve the AI using the current PTW editor.

alva, could you give a link where Korn mentions anything about improving the AI?

[jshelr]

This material is interesting and could prove a big plus for PTW games. You have customers waiting for a product!!!

Just a thought: the early game structure already favors stopping settler expansion early to build military units. If you increase settler shield costs, that would tip the balance further toward the pure warmonger style. Fine with me, but maybe not with everyone.

Would it be possible somehow to lower unit shield costs relative to building costs for the AI but not for the human player? For any given number of units, the AI would have resources left for more buildings. Relative to the human player, the AI would have an even bigger cost advantage for units than they do for buildings to compensate for the poor battle tactics.

[Random Passerby]

The other thread hints at a way to get at the AI's build preferences a bit more aggressively: as noted there, Spain has "build never" for Wealth, and so will not build banks or stock exchanges... but WILL build marketplaces, since they have other uses too. Try checking "Build never" culture and "Build often" production--the AI still remains in love with libraries/universities, but goes for factory, power plant, manufacturing plant right afterwords, only building coloseums and then temple/cathedral later. Still not exactly ideal, but it's an interesting concept: so long as an improvement has ONE function for which the AI does not have "Build never" flagged, the AI will still build it... eventually.

[joncnunn]

A. I propose that nothing be done unless they reduce the number of extra starting military units. We do not need to fight that many more AI military units in early game.

B. Option 4: Increase priority of the happinesses imporving structions and in addition, consider raising the luxary bar when there are too many entertainers present and/or rush completing happiness improvements.

C. Change the AI code to ensure that as soon as practicle 1 harbor gets built.

D. Catherdrials should be built before Factories. But Factories should be built before Reserach Labs. In general make Factories the highest priority for pop 13+ cities until built.

E. Don't increase maintence costs instead work on better combat tactics, this could start with considering all cities defended by the number of units that can reach the city by the quickest time the AI can reach the city.

E. Of course the AI has Nationalism first. Scientific civs are granted it for free. No change needed.

F. Fix the AI to not build inferior units. If you increase the attack factor of infentry it would have a higher attack factor than Calvary.

G. Have the AI consider the odds before attacking and stop making futile attacks.

[DaveMcW]

This isn't really a suggestion but I find it very ironic.

Since we can't improve the AI, we need to heap penalties on those tricky humans until their build order matches the AI. Then, with everyone playing like an AI, we can see who the *real* killer AI is!

[alexman]

If you can't lose to them, join them.

[ducki]

Re: Ideas for a killer AI

Originally posted by alexman [*]The AI emphasizes growth even when it has happiness problems. This causes the AI to assign entertainers, losing productivity, and irrigating more tiles to support those entertainers, losing even more productivity. [*]The AI doesn't build factories early enough. Even with "Production" as a build-often preference, the AI builds Factories after cultural buildings like Cathedrals and Research labs.

I'd be willing to bet my Forbidden Palace that these are related.

He's having happiness problems and losing productivity, so naturally, happiness builds will be more important than production builds until happiness isn't a problem.



At least, I'd be willing to bet that.
I think if you get the AI's happiness problems semi-solved, the production building problem might solve itself.

[joncnunn]

I'm not sure about the other AIs, but the Russian AI did build a FP during the Middle Ages. (In Berlin, the Russians wiped the Germans out early on.)

What the Russian AI took until late in the industrial era to do was build a Harbor.

[alexman]

Originally posted by jshelr
Would it be possible somehow to lower unit shield costs relative to building costs for the AI but not for the human player?

Unfortunately, I don't see any way of doing that.

Originally posted by Random Passerby
Try checking "Build never" culture and "Build often" production--the AI still remains in love with libraries/universities, but goes for factory, power plant, manufacturing plant right afterwords, only building coloseums and then temple/cathedral later.

This is a GREAT idea, but it doesn't work because Temples and Cathedrals have other functions as well: Happiness. The never-build for Culture would work only if there were buildings that produced culture and nothing else. I tried flagging Production as build-often and Culture as build-never and the AI behaved exactly the same as with just Production build-often.

[alexman]

I updated the initial post with a few more ideas. Some of these ideas are very radical. I'm suprised there is no reaction yet...

You better speak up because otherwise they'll end up in the PTW AU mod!

[DaveMcW]

Originally posted by jshelr

Just a thought: the early game structure already favors stopping settler expansion early to build military units. If you increase settler shield costs, that would tip the balance further toward the pure warmonger style. Fine with me, but maybe not with everyone.

I agree. By fixing only the builder aspects of the AI we are making it harder to beat them as a builder. Thus warmongering becomes relatively more attractive. Unfortunately we have no control at all over the AI's military tactics.


Re: giving the AI warlord/chieftain level corruption.

The AI already gets a corruption bonus as you move up in difficulty, by way of the penalty to humans. Decreasing corruption is just as drastic as decreasing the AI production costs. If you want to make a "beyond diety" mod so be it, but I thought the goal of the AU mod was to make the AI play better without cheating.

[Theseus]

1. I would avoid changing Settler cost... this is more for me than for the AI, it just confuses me too much when going back and forth between mods and stock.

2. Double the entertainers' effect. WE just have to agree not to exploit this.

3. and 4. I think trade and production should be emphasized... we do it, don't we?

5. and 6. Leave alone.

7. 8-attack for Infantry, yes?

8. Repeat what we did before.

9. Cool!

10. Leave alone.

11. This is the most difficult one... I'd love to hear from Soren.

12. What are the current fixes in the AU Mod?

We still need to talk about build preferences for the PTW civs.