PTW - New Stuff, Mysteries, and Thoughts

[vmxa1]

Well no more counting on the hordes to go for the town, where you can cope. Better have some units to handle them out on the range now.

[ducki]

Also, in my first attempt at a Monarch game, the AI actually sent out a pair of spearmen to pillage my roads and improvements during a war. Completely ignoring the forces that were 2 tiles away and approaching a town.

That's new for me, at least the pairing of two spearmen by the AI. I lost 4 swordsmen to those two. And they were regulars against my vets! Damnable RNG.

[Soren Johnson]

Originally posted by Dominae
I have to mention that Jeff wasn't "entirely sure" what changes were made (not his responsibility), he just knew no major changes were on the agenda. Perhaps Soren did some nice tweaking for his own amusement (and ours!).


Dominae

well, no major changes, but plenty of little ones, which make 1.04 Ptw AI somewhat different from 1.29 AI. Probably the biggest change actually came with the 1.04 patch. I finally solved the problem of city governors always over-emphasizing food, which means they will switch to super production mode once they hit pre-acqueduct size 6.

[FrustratedPoet]

Originally posted by Soren Johnson Firaxis
I finally solved the problem of city governors always over-emphasizing food, which means they will switch to super production mode once they hit pre-acqueduct size 6.

Very glad to hear that.

Alexman in particular will love this.

[ducki]

No wonder I'm losing out on a lot more Wonders lately.

[alexman]

Glad to hear Soren is still around!

However, I think the AI needs to go into super-production mode, building happiness improvements, when it starts using entertainers, not when it can't grow any more. See here for more info.

Actually, this AI problem comes from its refusal to use the luxury slider.

[Soren Johnson]

Originally posted by alexman
Glad to hear Soren is still around!

However, I think the AI needs to go into super-production mode, building happiness improvements, when it starts using entertainers, not when it can't grow any more. See here for more info.

Actually, this AI problem comes from its refusal to use the luxury slider.

the situation you outlined in that thread is something of an outlier. The AI's priority is to build riflemen to help defend the city because it is a capital. Otherwise, I would expect to see it build a temple pretty soon. Under normal game circumstances, the city would probably have a luxury or two, could use its defense for military police, etc, etc... I guess what I am saying is that if the AI is in that position so late in the game, it has already lost.

[Soren Johnson]

Originally posted by alexman
Glad to hear Soren is still around!

However, I think the AI needs to go into super-production mode, building happiness improvements, when it starts using entertainers, not when it can't grow any more. See here for more info.

Actually, this AI problem comes from its refusal to use the luxury slider.

hmmm... after looking at it again, the only thing that worries me is that the AI chose wealth. I almost never see it do that in my test games, so I'm not sure what is going on, to be honest. The AI is very different when it only has a few cities, so I am not surprised that starting it on a one city island caused it to fly off the rails.

[alexman]

Thanks for the reply. That makes sense.

Do you agree that the AI would benefit from using the luxury slider (like human players do) instead of entertainers? It would make it irrigate less and have more productive cities until it can build those happiness improvements.

[Soren Johnson]

Originally posted by alexman
Thanks for the reply. That makes sense.

Do you agree that the AI would benefit from using the luxury slider (like human players do) instead of entertainers? It would make it irrigate less and have more productive cities until it can build those happiness improvements.

the luxury slider is very, very tricky. the AI basically only uses it to balance out war weariness, and there is a maximum level it will never go over. the danger was that the AI could get used to the luxury slider and deemphasize acquiring luxury goods and building happiness improvements, allowing a good portion of the treasury to disappear.

[Dominae]

Thanks for your input Soren. A couple of quick questions: what do you mean by "super-production mode"? Will the AI start mining everything once a city reaches size 6, and will it destroy existing improvements in doing this? Will it revert back to "super-food mode" after Construction (or, once the city builds an Aqueduct)?


Dominae

[vmxa1]

Thanks for the info, the changes to 104 explains some of the things I have seen. I would expec that any unusal situation could create havoc with the AI.

[nbarclay]

Here's a new one: in my first PtW game, the first overseas AI to find me (or one of the first) seems to have actually sold OTHER AIs (with one exception) contact with ME! In the past, I've always been denied contact until I was the one to pay for it. Assuming what I think I saw represents a pattern, this is a giant leap toward the AIs not treating the human differently from each other in their negotiations!

Nathan

[nbarclay]

Originally posted by Soren Johnson Firaxis

the luxury slider is very, very tricky. the AI basically only uses it to balance out war weariness, and there is a maximum level it will never go over. the danger was that the AI could get used to the luxury slider and deemphasize acquiring luxury goods and building happiness improvements, allowing a good portion of the treasury to disappear.

Thanks for the info. It's really great that you're willing to post here! And I like what I've seen so far of what you've done with the AI for PtW.

Would it be possible to counter such a tendency by having the AI (1) base its luxury aquisitions and happiness improvement building on what the situation would be without the luxury slider and (2) then re-optimize the luxury slider setting as it acquires luxuries and completes happiness improvements? That would keep its main behavior essentially the same as if it weren't using the luxury slider but still benefit from the slider in some situations. (In connection with that, in some situations, it might need to preemptively convert some people to specialists in order to avoid, or at least slow down, growing to a point where it couldn't feed everyone without happiness problems.)

Nathan

[alexman]

Good idea Nathan. Effective use of the luxury slider would go a long way towards improving the AI.

Since we are talking about the AI, here's another area that could use some help: don't build settlers when their estimated completion comes before the estimated city growth to size 3 (settler pop cost + 1).

The following screenshot is from a standard emperor game in debug mode, and ALL 7 AIs did exactly the same thing: they immediately started building a settler in 4000 B.C. because they had enough units already (the free ones). The result is that the AI wastes several turns of production waiting for the city to grow to size 3. Why not build a temple or granary instead?

This tendency of the AI is exploitable even on lower difficulty levels. You attack a ancient UU civ during their expansion phase, making sure you trigger their Golden Age, prompty sue for peace, and you waste their Golden Age because they try to build settlers, not realizing that their boosted production completes them before their cities are large enough. (Credit goes to Sir Ralph for pointing out this dirty trick).