Exciting AI improvements.

[Senethro]

Amazing blake, I've been playing with only some of the changes and its made for a tough Monarch game. After conquering 4 Zulu cities I subsequently lost 3 of them to Chinese hordes. Multiple mega-stacks of well trained units. I'd forgotten what it was like to lose a city.

[Kuciwalker]

before...after

Wow.

[kiky]

This is getting better and better!

That before - after, looks really scary.


Does the AIs know if their city is in danger ?
Say AI is at war, 'sees' enemy units coming and outnumber AI defense. Will AI start to whip military units?

[binTravkin]

Will AI start to whip military units?

I think Blake already said, he's planning it.

[Blake]

Once again I have improved the AI performance. I have finished my rewrite of the worker automation code (and now I am tired). The AI performance has gone up across the board. The AI should now also be less "cottage spammy".

Fancy your chances of out-building this Arabian production-behemoth?

EVERY AI's score has increased. This is mainly because I finally got the food balancing code right.

I solved the worker issue by just making them build lots of them. Seems to have worked out.

You can now download for play and testing:
CIV1.61_ImpAI_v0.5.zip

Basically to summarize what it does. GREATLY improves worker automation, the AI (and your automated workers) now properly balance out food (Assuming irrigation, they can't irrigation chain properly) and attempt to specialize the city, to a degree.

The AI city governor now uses the emphasize buttons much better. It grows A LOT faster, and a lot BIGGER. Mid game performance should be much, much better. The AI also has tweaked builds and research in order to improve it's general builder performance, for example workboats, workers, granaries, lighthouses are all more valued.

What the mod doesn't do is improve Military AI. I have not touched any military related code, the unit mix, the tactics, are all the same, so you can still split the AI open like a ripe melon. You might face a little more whipped/drafted resistance though, and generally they'll have a stronger economy capable of producing larger armies (so maybe expect army sizes of a difficulty or two higher).


CIV1.61_ImpAI_v0.5.zip

On the agenda:
Military AI.
Rework AI bonuses.
Special cases - building up fishing villages and such.

Source Code

[binTravkin]

Cool! I think everyone here's thankful to you, Blake!

[Swiss Pauli]

Originally posted by Blake You might face a little more whipped/drafted resistance though, and generally they'll have a stronger economy capable of producing larger armies (so maybe expect army sizes of a difficulty or two higher).

Excellent work! I look forward to being worn down by huge AI armies.

[Arrian]

Question on whipping units and drafting... especially in AI vs AI wars, will this not result in a CivIII-esq draft implosion, wherein the AI civs draft like crazy, wrecking their cities, and then of course make little to no gains in the war b/c their drafted units and the enemy AI's drafted units slaughter one another?

Obviously it's a fine line. On the one hand, it's bad for an AI to get crushed b/c it's unwilling to sacrifice some pop in the face of an existential threat. On the other, it's a bad thing if the AI will ruin itself (or set itself back significantly) due to being in a war, any war, particularly a war that it could win or draw w/o doing such a thing.

Is there any way for it to gauge different levels of war?

-Arrian

[Blake]

Well they wont draft if it causes unhappiness. Since each draft causes +3 unhappy, they'll have trouble making many drafts before hitting the happy ceiling (ie even a size 10 city, with happy cap of 20, can only be drafted about 4 times).

And when an emp+ AI invades my empire... it causes my empire to implode, with maybe 50% of pop converted to units. But that implosion of population into units turns the tide of the fight and the AI is driven back and eventually conquered...
And lets face it. When is the last time an AI has broken a choke or fought off with an attack? At present they always just fold. If nothing else, making invasion more of a grind would make war generally less profitable.

War is hell. Or it should be anyway.

[Quillan]

I've been following this thread closely since it started, but havent chimed in until now. All I can say is that's some stupendous work, Blake! My hat is off to you. I haven't tried your improved code yet (too busy playing Company of Heroes and a campaign of Rome: Total Realism at the moment), but by the time I'm finished with that, you'll probably have even more done. I'm definitely looking forward to trying the finished product.

[DaveV]

Blake,

Here's a case where I disagreed with the governor (emphasize food, hammers, commerce all turned on). In the specific case of building a worker or settler, I'd rather give up the commerce and work a 3-yield tile to get the unit built faster and go back to growing.

By the way, is there a way to tell for sure we're running your code (version number or something)?

[Quillan]

Two things spring immediately to mind from that picture, Dave. First, all you could get is 2 food 1 hammer from working one of those grassland forest squares. Would the extra 1 hammer a turn make a difference in the production of that settler? Second, try turning off "Emphasize commerce" and see if the worked square moves.

[DaveV]

Quillan - I could make the settler in 13 turns instead of 14 with the worker on the grassland forest. I guess it's debateable whether making the settler 1 turn earlier is worth giving up 39 commerce (actually, slightly less than that, since the new city will generate one extra turn of commerce). If it lets me beat someone to a good city site, it's worth it in my book.

[Swiss Pauli]

Originally posted by DaveV
By the way, is there a way to tell for sure we're running your code (version number or something)?

Running it as a mod is the best way of tracking it, I think.

I'm playing a test game and the Monarch AI are doing a good job of dogpiling me, although I have been provoking them. I had a sweet start (FPs + riverside-plains-wheat) so they're still taking a beating, but aren't the pushovers the were previously.

[Blake]

Originally posted by DaveV
Blake,

Here's a case where I disagreed with the governor (emphasize food, hammers, commerce all turned on). In the specific case of building a worker or settler, I'd rather give up the commerce and work a 3-yield tile to get the unit built faster and go back to growing.

As I said previously, I've only tweaked the AI usage of emphasize buttons, I have not touched the general Governor code. I'm probably going to wait for the next Warlords patch before really tackling that but I MIGHT improve it before then.

Oh and actually I think the AI will work cottages while training workers/settlers... I think I'll make it turn off emphasize commerce. It also sometimes trains them in size 2 cites for some reason even though I thought I stopped them doing that (i know sometimes te player wants to train workers/settlers in small ciites but I think it's best for the AI to grow to size 3, especially sicne the AI delights in training such units at a couple of food to go for the next growth...).

By the way, is there a way to tell for sure we're running your code (version number or something)?

Hmmm.... having blue circles on is probably the easiest way .