Why, oh why does the poor AI...

[punkbass2000]

You know, Mazarin, that is a really good idea Very creative and original.

[Mazarin]

I discovered that during my AU202 game: I used my only leader to rush the palace in newly conquered territory (costed about 800 shields)...an I was pretty surprised when the next turn, I still needed 50 rounds to finish the palace

[Gwylim]

lets come back to the point made earlier:
The Problem is that the AI is a challenge only in the beginning of the game and as more high the difficulty level the more difficult is the beginning and this is exactly my Problem.
When i play on King its a good challenge at the beginning, but very soon i am the most strong Civ and the game becomes boring.
On Emperor it is even more difficult at the beginning and on Deity it is (for me) too difficult at the beginning to enjoy the game. i have no fun waiting most of the game until i catched up,
so my problem is that up to King the challenge is lasting not long enough and on Emeror and Deity it is too difficult in the beginning.
What i think also much other here would like is an AI which is challenging most of the game and not only at the beginning.
If it is too difficult to program a better AI what about giving them other bonuses later on in the game, or perhaps make them stand together always against the human player once the Player is thebiggest civ?

[ducki]

Originally posted by Gwylim
If it is too difficult to program a better AI what about giving them other bonuses later on in the game, or perhaps make them stand together always against the human player once the Player is thebiggest civ?

One option here (as it seems the ability to exploit the AIs behaviors to your advantage is the real problem) is to alter your behavior.

Try limiting yourself to the OCN.
Try putting certain restrictions on your build options - always Temple/Cathedral first, never barracks, never roads, always irrigate, never Science buildings, never Religious buildings, whatever, you get the point.

Yes, it would be great if the AI could play as well as a human, but that's just not feasible. The fact that the stronger AI gets more "bonuses" as opposed to the human getting more "handicaps" shows that there's no way to make the AI "smarter" or more flexible.

All that we can do is give the AI more "cheats" or try to improve his "playbook" or "recipe book". It's still not going to make him smarter, even though it might make him harder to beat.


With the tweaks to the PtW AI, it's definitely loads better than when this thread was started - check out how many mines the AI builds opposed to how many irrigations, it's semi-scary now.

The AU mod helps the AI some, but if you're an "in-betweener" like me, you're stuck playing a game that's either too easy or too hard; in this case, I think personally limiting your own behavior can make the lower difficulty setting a bit more enjoyable.

Not the optimal solution, but a solution.

[theNiceOne]

I'm with Gwylim.

I'm now playing comfortably on Emperor. I think the difficulty is hard enough during the ancient age, but would like the difficulty to be just a bit harder during the middle ages, even harder during the industrial ages, etc.

ducki has some suggestions, but although I have enjoyed a few OCN games, all his suggestions involves restricting how I can play so much that it no longer will feel like my strategy.

I know that a good AI is hard to program, but what could be easier is separate difficulty settings per age, so that the AI production bonus could be even higher during the later ages and their corruption lower, etc.

Another possibility, which can be modded without the help from Firaxis, is to make a few new governments - governments that are better than any of the current ones. Like a democracy w/o war weariness and with military police. This government would be a very good boost for the AI during the later stages, but would be illegal for the human player to ever choose.

It's not possible (AFAIK) to actually restrict the human from selecting such a government, but it will be considered an exploit to ever switch to it.

[ducki]

Note, I said limit yourself to the OCN - Optimal City Number - not OCC - One City Challenge.


One question - you say that specifically setting certain limitations would make it not feel like your strategy - how much of your strategy is really "your strategy" and how much is an adaptation based on how the AI plays?
In other words, is it really strategy or is it meta-strategy? This is not a pointed question, really. I'm just curious if the adaptations we make to compensate for the AI's bonuses are viewed as "my strategy" or if they are recognized as compensations.

[Pyrodrew]

What i think also much other here would like is an AI which is challenging most of the game and not only at the beginning.

It would nice if Firaxis came out with a patch that would do that... but that is not likely & dreaming will not help.

Therefore, if limitations are not appealing to you, I would recommend using the editor to modify the game. IMO, I have found the following hurts the AI:

1.Railroads - The AI doesn't seem to understand how to change it's strategy when railroads are invented. For example, it still tries to drag its troops through your territory to reach that unguarded city (as if you only had mere roads). Maybe open the editor & delay railoads.
2.Forbidden Palace - The AI doesn't know how to place down its Forbidden Palace in a good spot. So once FPs come into the scene (around Medieval Age) the human begins to gain an advantage of less corruption whose benefits grow to where it becomes obvious in the Industrial & Modern Ages. Maybe delay FPs.
3.Palace - The human knows how to Palace-Jump to further decrease corruption as the game progresses, where the AI does not. Not sure if this can be edited.
4.Corruption - The human knows how to battle corruption with rush-builds, WLTKDs, etc. & the AI does not. Therefore, lowering corruption on a grand scale (using the slider in the editor to ~70-80%) has been reported as providing favorable results for the AI.
5.Artillery - The AI does not know how to use artillery offensively or even in a cost-effective manner. Flagging the AI to not build artillery (so it builds other military units instead) should be of benefit. Removing artillery completely is another option.

If you don't want to touch the editor, the AU Mod has somewhat of a goal of being focused on improving the AI which you may want to try. If you feel that's changing/ruining the game... not much else to do, unless you want to post again about how you would like a better AI.

[Ozymandias]

Originally posted by theNiceOne
Another possibility, which can be modded without the help from Firaxis, is to make a few new governments - governments that are better than any of the current ones. Like a democracy w/o war weariness and with military police. This government would be a very good boost for the AI during the later stages, but would be illegal for the human player to ever choose.

It's not possible (AFAIK) to actually restrict the human from selecting such a government, but it will be considered an exploit to ever switch to it.

Actually, I believe it is, although it requires definite fiddling with the tech tree --

Set some mandatory precursor Advance to a government type, assign it as belonging to "No Era" (i.e., can't be research or -- I think??? -- traded) and assign it only to the AI player(s).

-Oz

[Nor Me]

Originally posted by Ozymandias
Set some mandatory precursor Advance to a government type, assign it as belonging to "No Era" (i.e., can't be research or -- I think??? -- traded) and assign it only to the AI player(s).

They cannot be traded. But out of interest have they fixed the Great Library problem with these in PTW?
Also they would get this at the start.

Originally posted by Pyrodrew
2.Forbidden Palace - The AI doesn't know how to place down its Forbidden Palace in a good spot. So once FPs come into the scene (around Medieval Age) the human begins to gain an advantage of less corruption whose benefits grow to where it becomes obvious in the Industrial & Modern Ages. Maybe delay FPs.

This is probably too radical for AU but I wonder if giving the FP a prerequisite of a courthouse would help.
The AI might often have courthouses first in a moderately corrupt area so gain more from the FP.
It would need to be playtested and I imagine the FP would sometimes end up being built in very odd places or near the capital. It might be better placed on average though.
Also the human would need more investment than just one leader.

[Ozymandias]

Originally posted by Nor Me


They cannot be traded. But out of interest have they fixed the Great Library problem with these in PTW?
Also they would get this at the start.

Oh my -- I didn't know there was a problem re: TGL ...

However, what I meant by "precursor" was:

1. If it's a non-human gov type available at the start, then just use the "No Era" flag uniquely for the advance required for the gov.

2. If not available at start, have the gov type be associated with an Advance not required to complete an era, have it branch from the tech tree wherever you see fit, AND have the "No Era" Advance required.

-Oz

[Nor Me]

Originally posted by Ozymandias
2. If not available at start, have the gov type be associated with an Advance not required to complete an era, have it branch from the tech tree wherever you see fit, AND have the "No Era" Advance required.

This is the part I cannot do. Advances seem to only be able to require advances in the same era and goverments have only one prerequisite which is an advance.

[alexman]

...build Longbowmen when Medieval Infantry is available?

This is a similar problem to building the Guerilla when Infantry is available. Fortunately, it doesn't happen very often, but it does happen sometimes.

It seems that the AI can evaluate which unit is better to build for a certain purpose (offense, defense), and it builds mostly that unit, but it also builds some other units for that purpose, just in case the other units turn out to be useful.

[Nor Me]

shows promise

In the game from which this screenshot was taken, the FP requires a courthouse and allows air trade:

[Nor Me]

Unfortunately of the 5 other surviving AIs, two have the FP near their palace and 3 have not built it in 920AD.
Obviously this needs more playtesting (not to mention a control. How well has anyone else seen the AI FP placement ever?)
But the Persians are looking like a killer AI. (No scared smilies?)

[Catt]

. . . refuse to switch government . . .

. . . when religious and when suffering starvation-level war weariness?

The attached shows two city views of Madrid, the AI Spain's capitol, over the course of a nasty war which ultimately ended with Spain's elimination. Spain (religious) refused to switch from Democracy to either Monarchy or Communism, even as Madrid starved from size 12 to size 3. Only at size 3 did Spain finally go communist. Not only would a switch to communism have alleviated the WW unhappiness, but it would have offered more opportunity to draft defenders (which were sorely needed). One of Spain's foes, Persia (not religious), switched to Communism before any significant starvation occured.

Catt