A Better AI.

[Jaybe]

Originally posted by Aro
And that’s what I get on warlord level. In fact, we need a new level now, between warlord and noble... A mod, maybe...

I am only on my first Warlord difficulty game, having given up at Noble with Warlords 208 (huge/fractal/marathon). I also got 2 settlers from huts, though not on hut #1. Here is what I am considering doing to HandicapInfo.xml so I can try Noble again. I may play 1 or 2 more games before I implement the following:

iAIUnitSupplyPercent: 45 (from 35)
iAIUnitUpgradePercent: 40 (from 30)
iAIInflationPercent: 90 (from 70)
iAIWarWearinessPercent: 90 (from 70)
That's changing the above by two levels, giving Noble the Chieftan values.

I may also adjust iNoTechTradeModifier & iTechTradeKnownModifier similarly, though I don't know what the latter does.

I have already changed iAnimalBonus, iBarbarianBonus, iAIAnimalBonus & iAIBarbarianBonus so I & the AI have the same values (currently -10).

I will probably also increase iUnitCostPercent from 50 to 100. iAIUnitCostPercent is already 100 at Noble level. Paying only half as much as the AI for troop upkeep seems obscene.

[Blake]

New version (07), get it on the Source Forge page.

Changes:

- Workers and other non-combatant units will now respect borders of hostile players as dangerous places to be, even if no units are visible. It should now be pretty darn hard to catch a worker in wartime.
- No more worker gangs, the AI will avoid forming large gangs on one tile and will also distribute it's workers better on a per-city basis.
- Workers should now build railroads on mines/lumber mills as super high priority, since it's a very cost effective build.
- Saner post-biology terraforming, the AI will no longer be obsessed with biology farms. In particular it should aim for only enough food for size 20-21 cities.
- The AI now properly values Coal as providing Railroads. It should be worth about 2 luxuries and maybe a little chumps change.
- The AI will be far more reluctant to use the Woodsman and Guerilla promotions for non-scout units. In particular they'll only choose one of those promotions if they are actually standing on the terrain type when they earn the exp.
- Fixed a problem where an AI stack will wait to heal in hostile territory with no medic, wasting upwards of a dozen turns.
- Slightly reduced the AI's tendency to pillage. If you ask the AI to attack a city, it shouldn't pillage that city.

- The AI will put more emphasis on priests in holy cities.
- Improved AI missionary usage. The AI should do a better job of getting some monasteries and spreading it's religion around to those who would join them (while recognizing lost causes).
- AI's without a holy city will now build missionaries for their adopted state religion, although with somewhat less enthusiasm and they wont spread it to others.
- Added understanding of cultural pressure. The AI should now border push much more efficiently, especially when resources are at stake.
- No more daft placement of Hermitage / Cathedrals / Great Artists. The AI will be far more inclined to build these in culturally pressured cities.
- AI will use Artist Bombs defensively to push back encroaching culture, especially to steal resources. They will respect Friendly relationships to a large degree and recognize that the bombs are not effective against established culture.

This can be summarized as some needed fixes to workers especially for industrial+, some tentative improvements to the tactical side of things and a decent upgrade to the religious aspects. If my changes work as intended you'll also feel a gut-wrenching fear as "Homer" gets popped, however the the target will usually be another AI since AI's tend to found cities more offensively.

The AI should also be more able to hold it's own in border struggles over resources in the early game.

By the way, on the topic of Tech Trading. Try a no tech trading game and you'll see the AI does indeed just tech that fast. "Exclusive" tech trading is not and has never been a crutch the AI uses for faster teching. If you are playing on a "fair" difficulty you should except to be middle of the pack in research much of the time, unless you have a serious edge in research due to terrain and/or wonders.

[Diadem]

Tech trading is in fact a crutch for humans, not AIs. I've been able to keep up in tech at deity games where the AIs researched 3-4 techs for every one I did, simply by beelining for unprioritized techs and trading them away to everyone. This only works with lots and lots of civs, of course. Of course now that Blake changed the research priorities, I don't know what to beeline for anymore

But seriously, cool changes.

I disagree with the Woodsman / Guerilla changes though. I fail to see what the terrain a unit is standing on matters. What matters is the terrain a unit will be standing on. I understand that that is very tough to program in, but in that case I'd just do away with these promotions for AI alltogether. Anyway with a scout I would almost always take woodsman I / II over any other promotion, especially if it's a warrior scout. And Guerilla seems very useful for those resource-guarding units the AI likes to spam everywhere, if the resource in question is on a hill (though taking away this spamage alltogether seems a good idea as well).

[Adagio]

Originally posted by Solver
Yeah, don't be afraid to play on a lower difficulty level now. The new Noble should really feel like Prince before...

I just went down a level, and still having trouble beating the AI... this mod sure has given us a more challenging AI

[Swiss Pauli]

Originally posted by Blake
New version (07), get it on the Source Forge page.

Nice to see an update, but you're certainly progressing a bit too fast for me! Obviously you want to correct anything that doesn't seem very good (e.g. worker gangs), but how about a week between updates, bugfixes excepted?

[zeace]

I think we should decide what we want the AI advantage to be on harder difficulty levels. There's lots of easy bonuses we can give to the AI, but they all unbalance the game in one way or another:

1. Happiness/healthiness cap levels that are affected by difficulty. The AI should always play at Noble caps so we can play easier/harder because of that.

2. Popping stuff from huts. I like that at easy levels you can pop stuff like settlers, while at hard levels you can pop barbs. The AI should always pop stuff as a Noble player would.

3. Production/Research bonuses/penalties. The AI at very high levels should get a 10% or 20% bonus to such things. Not at Monarch hopefully, but at Deity you're asking for the AI to be super human. Maybe keep penalties on really low difficulty levels.

4. Upgrade bonuses. I don't like this bonus as it encourages the AI to build huge armies all the time instead of a specialized army when it is going to war. The result is that my army looks small and so the AIs pick on me. Obviously we would want an AI that understands the high cost of upgrading an army and would only build units it intends to use before they go obsolete. This way a peaceful civ would probably keep a much smaller army around, making my army not seem so bad.

5. Preference for human/AI players. When an AI is deciding who to trade with, or who to go to war with, it could include some preference/reluctance for the human player(s). Many of us will start a war with Monty before he starts a war with us, and we'll adopt Izzy's religion to keep her from declaring a Jihad on us. The AI should recognize that the human is a strong player, and should treat us accordingly. At low levels that means lots of sucking up. At high levels that means making our extermination a priority.

6. Victory conditions. I'd like to see a future AI that is good at focusing on certain victory conditions. Right now an AI will virtually never win a domination/conquest victory, and will absolutely never win by culture. On high difficulty levels I'd like to see an AI decide that it has 45% of the land so it should "go for" a domination victory. It's relationship with other civs would be relative to whether or not you're helping it to achieve its victory. Similarly it'd be nice to see a civ with a big culture lead decide to "go for" a culture win, and start focusing its efforts on getting 3 or 4 cities with very high culture. At lower difficulties the AI should just wander around existing the way it does now. I know this is hard, but it'd be great fun to have to keep Monty from taking too much land while preventing Saladin from getting his last city's culture up, all the while finishing your spaceship.

7. Actual AI differences. I don't know if I like the idea of the AI acting very differently on different levels. It lets you get into really bad habits. Like if the AI ignores Alphabet on some levels but then starts prioritizing it on harder levels it can make the step up seem really hard to someone who has been relying on the AIs bad habit. I don't want to see the AI building cities in bad spots or building bad improvements on squares because noobies will see where the AI did things and will learn those bad habits. That said, I wouldn't mind if a weak AI didn't understand the benefits of specialization. Not building marketplaces everywhere is a powerful strategy that most weak players don't understand so I don't mind modeling that in the AI. But by about Noble or so I want the AI to be playing its best.


Those are my thoughts... If the AI does dramatically improve it might be ok to move the "fair level" from Noble to Prince as otherwise there aren't a lot of easy levels to pick from.

Any thoughts?

[Nacht]

Originally posted by zeace
I think we should decide what we want the AI advantage to be on harder difficulty levels.

Since this is a work in progress it might be a little early for that. For now just choose a difficulty level you feel is right, and you might mod out popped settlers for you or free workers for the AI. Discussing adjusting the difficulty levels at this point (CFC ) isn't very usefull.

[Harrier UK]

Originally posted by Swiss Pauli

Nice to see an update, but you're certainly progressing a bit too fast for me! Obviously you want to correct anything that doesn't seem very good (e.g. worker gangs), but how about a week between updates, bugfixes excepted?

Hey, you do not have to upgrade until you have finished the current game.

In fact in some circumstances that is the best way to go if you find a bug. As it can be pinpointed to a particular release.

[Harrier UK]

Originally posted by zeace
I think we should decide what we want the AI advantage to be on harder difficulty levels. There's lots of easy bonuses we can give to the AI, but they all unbalance the game in one way or another:

Those are my thoughts... If the AI does dramatically improve it might be ok to move the "fair level" from Noble to Prince as otherwise there aren't a lot of easy levels to pick from.

Any thoughts?

Infact your thoughts seem to be different from what I have seen posted on other sites. That is assuming I have not misunderstood you. English can sometimes be a hard language to get proper comprehension from.

Anyway back to the point - a lot of posters seem to be implying that some bonuses should be removed or reduced from the AI, not given to them.

Also a lot of players no longer consider Noble to be the "fair level", but also think the one below it - to still be to unbalanced in favour of the human player. They would like to see an intermediary level - akin to the old noble level, between the two. REducing both human and AI bonuses.

[Blake]

Originally posted by Diadem

I disagree with the Woodsman / Guerilla changes though. I fail to see what the terrain a unit is standing on matters. What matters is the terrain a unit will be standing on. I understand that that is very tough to program in, but in that case I'd just do away with these promotions for AI alltogether.

The theory is that the Woodsman restriction means "almost never" which is correct - there's only two ways a unit will gain experience while on a forest - either it's fighting on the forest, or it pops a hut on a forest. Either way that's a fair indication that it can benefit from Woodsman, either because it's scouting or because it's fighting in forested terrain.

As for Guerilla, that means it'll sometimes pick Guerilla if the city is founded on a hill, if a city is founded on a hill, that indicates there are other hills around. In particular - if there are hills around, there will be cities on the hills, if there are no hills arounnd, there will no cities on the hills and thus no Guerillas.

The goal is simply to ensure that the AI definitely wont pick the promotion if it definitely wont be useful. This is at the acceptable cost of sometimes not picking the promotion when it would be useful.

The other thing is that this code runs nigh-instantly, while looking at the terrain all around the empire would take ages (relatively speaking). I figure it's quick, dirty and somewhat better than nothing.

[Senethro]

Nice job on getting the AI to use G.Artists well. I kept getting sucked back into wars when captured territory was culture bombed back to the losing side.

Also, Russia declared war on me while our entire armies were just outside each others heartland. But I didn't lose any cities because Peter went pillage crazy. Admittedly, it was Macemen+elephants vs. city garrison 2 Grenadiers.

[Kuciwalker]

- The AI will be far more reluctant to use the Woodsman and Guerilla promotions for non-scout units. In particular they'll only choose one of those promotions if they are actually standing on the terrain type when they earn the exp.

What about if it already has Guerilla/Woodsman I (e.g. Gallic Swordsmen)?

[Dis]

I'm really anxious to try this (but NWN2 is out). But I really don't want to go below Noble level. It makes me think the AI is better than me, and it's just something I don't want to accept. But I can pretty much blowout noble (haven't played above noble yet), so I hope I can at least maintain noble level, just won't be blowouts.

[snoopy369]

A lot of it has to do with your play style. I actually am playing at the same difficulty level I was before ... what's happened for me is that the AIs are (generally) leaving me alone more than they were, because their economy is better, which means I can continue to out-Build the AI, just not by as much as I had before. I'm actually better off, because I have played whole games now where I went to war once or not at all. I can build my builder stuff without having to worry too much about military, as long as I keep enough units to keep my power level up...

Now, once Blake kicks the AI's tactical rear end up a notch, I may be in trouble

[Blake]

Originally posted by snoopy369
Now, once Blake kicks the AI's tactical rear end up a notch, I may be in trouble

Heh. Heh. Heh.