Diplo AI - Approach Calculation (4-7 Beta Patch)

[Recursive]

Is there a way to tweak these values ourselves?

I agree that too much expansion or wonder spamming could accrue diplo penalties but I don't think it should be that high. I had -200 from building wonders by late game and I wasn't even the highest wonder builder in the game.

Building wonders are already restrictive and need a lot of planning/decision making. (ie. no. of policies needed, policy finishers, city plot/terrain requirements, production penalty from previous wonders).

If I manage to beat the AI to a wonder despite all this, it's a big achievment and so satisfying and ultimately fun.

It seems anti-fun to be punished for "winning a race". A race where I am blindfolded, running one legged, with a weight attached to me lol

Again I don't think I should get NO diplo penalty, but just not so extreme.

You can adjust opinion weights in the DiploOpinionWeights.sql file.

I suggest updating to the latest beta, the penalty is far more reasonable.

 

[JamesNinelives]

The instances of the AI declaring aggressive war and basically not doing anything either to invade seriously or build up their military were constant, and one of the weakest points currently. If the AI is going to war, they need an actual army that can win a war and to then actually use that army from the get-go. Especially against a human. The Huns and Sejong both DoWd me without a clear plan of assault.

I've seen this complaint from time to time and while I do empathise (it's annoying) it is sometimes I have done myself on occasion. E.g. if an AI is leading the pack and ahead of me on tech (so I don't want to engage) it's advantageous to me to join a war against them that is already being fought by others. Both for diplomatic reasons and so that I can pillage/kill any wandering units they send.

I guess it maybe it's different when AIs do it because they underestimate how much of a threat the human is. Just wanted to say it can be useful sometimes, particularly if one of the AIs is in the lead rather than the human.

 

[MontyJava]

Certainly that makes sense, James, but I'm talking more about AIs starting wars rather than dogpiling.

Later in the game, this issue died down. That final war, I definitely got invaded, hard, on all three sides, and a mid-game invasion by Hiawatha saw him throw a massive army at me from the get-go, which was fun and intense.

It might just be certain scenarios the AI doesn't handle well. Early game wars were very weak on the AIs' part, while Industrial and later wars it handled well. I think it's just struggling to build up a large enough force to actually do something, while also keeping its own territory protected. I'll need to play through a few more games to suss it out. Not sure these diplo changes are directly reflected in this issue anyway.

@Recursive I have the newest patch. I'm probably just going to stop building more than a couple wonders, or otherwise will have to dig into AI diplo tables to see who is going to react negatively first and heavily base my diplomacy around that. Some AIs don't care, and for some it severely undermines the relationship without my going overboard. Where is that table in the files, for things like warmonger/wonder/religious difference attitude flavors?

Also it doesn't help that the UI doesn't clearly say what's a wonder. I'm still (re)learning the game/mod, and I have a few times accidentally built a wonder or thought it was just a national wonder. Such a huge oversight by Firaxis on this one.

 

[Recursive]

@MontyJava I'm not ignoring you, just haven't gotten much chance to reply as life's been busy and your posts are long. I appreciate the detailed feedback!

 

[adan_eslavo]

Maybe I will add some things noticed in my latest game:

• Aztecs were very angry at me, because of my warmongering, so they declared war on me. The problem was, they were on the other side of continent and between us was partially Rome and Assyria and about 4 CSs who weren't my allies. So I just waited and couple of turns later they just suggested white peace... pathetic
• I declared war many times to Persia.
  • First was joined war with Denmark+Me vs Persia (Denmark asked and I waited 10 turns). Persia focused on Denmark and moved almost no forces to their weakest city, so I captured them with only spearmen. Accepted peace treaty. They didn't capture any of the city of Denmark.
  • Second war Russia+Me vs Persia (Russia asked me) and later on Denmark vs Persia at the same time. Again, passive defence of Persia allowed me to take another city on the north. Same war finale.
  • Third and last war was started by me. Later another war Ottoman vs Persia, Denmark vs Persia, Germany vs Persia started all at the same time. Persia didn't have easy life in that game and ended up eaten by me (5 cities) and Ottoman (1 city). The bad thing was it didn't build up its defences even with great threat from all neighbours. And not to mention that it had its special UU (!) in the early game and didn't use its advantage. It was not any problem.
• I had very good terms with Assyria which 2 times asked me about fighting Zulu. It was nice cooperation North-South.
• What was annoying: many, many, many times I was asked for joined war by nations from the other side of continent. Even If I wanted, I couldn't reach the target. It was really annoying.

Persia should focus fully on land units because most of their territory were land cities (only one coastal on inland sea). But they had huge navy. What for?

 

[JamesNinelives]

• What was annoying: many, many, many times I was asked for joined war by nations from the other side of continent. Even If I wanted, I couldn't reach the target. It was really annoying.

I actually like these requests because they give me opportunities to influence the diplomatic landscape.

I tend to 'join' the fight when the target is someone I dislike/a rival for techs and wonders. Even if I don't participate in the fight, it encourages conflict between other people which is to my benefit overall. Of course, if a big/militaristic nation is asking me to declare war against a smaller nation then I don't want them to get any stronger so I actually tip off the the target in the hopes it will boost their defense (and garner favour between us).

 

[adan_eslavo]

I actually like these requests because they give me opportunities to influence the diplomatic landscape.

I tend to 'join' the fight when the target is someone I dislike/a rival for techs and wonders. Even if I don't participate in the fight, it encourages conflict between other people which is to my benefit overall. Of course, if a big/militaristic nation is asking me to declare war against a smaller nation then I don't want them to get any stronger so I actually tip off the the target in the hopes it will boost their defense (and garner favour between us).

Valid point.

 

[Recursive]

A T300 review of new game. I've aimed for a bit more brevity, and I've divided things up into clear wars. You can ignore the spoiler unless you want more details.

Spoiler :

T088 - City founding mostly over

T084 - Dido DoW Gust - (Neighbors, Dido +50% mil, buildup)
T107 - Dido Pea Gust

T092 - Huns DoW Port - (Neighbors, +40%, mild buildup)
T098 - Inca DoW Port - (Neighbors, mil score ==, dogpile?)
T119 - Huns cap 1 city Port
T121 - Inca Pea Port
T128 - Huns Pea Port

T116 - Gust DoW Hiaw - (Neighbors, Gust has 5% lower mil score, Hiaw buildup not Gust???)
T130 - Dido DoW Gust - (again, neighbors, 10% higher mil)
T134 - Huns DoW Hiaw - (now neighbors, 5% lower mil)
T139 - Hiaw Pea Gust
T152 - Dido Pea Gust
T158 - Huns Pea Hiaw

T157 - Huns DoW Port - (now neighbors, Huns +80% mil)
T179 - Huns Pea Port

T189 - Gust DoW Hiaw - (again, neighbors, == mil, no buildup)
T191 - Huns DoW Hiaw - (again, neighbors, dogpile)
T210 - Hiaw Pea Gust
T215 - Huns Pea Hiaw

T201 - Plyr DoW Port - (she had it coming, citadel bombs and all)
T204 - Huns DoW Port - (again, dogpile, neighbors)
T211 - Plyr cap cap Port
T211 - Port recap cap - (ships capturing cities is about my least favorite aspect of this game)
T212 - Plyr cap cap Port
T213 - Plyr Pea Port
T226 - Huns cap 1 city Port
T240 - Huns VAS Port

T217 - Dido DoW Isab - (Not neighbors, Dido +40%, mild buildup)
T232 - Dido Pea Isab

T230 - Gust DoW Baby / Dido / Lizy
T251 - Dido Pea Gust
T252 - Baby Pea Gust

T238 - Huns DoW Gust / Inca / Isab
T250 - Huns Pea Gust
T253 - Huns Pea Isab
T259 - Inca Pea Huns

T243 - Dido Dow Inca / Isab / Sejo / Poca (Poca gets hit. Are neighbors both, +35% mil, huge buildup)
T255 - Sejo Pea Dido
T258 - Dido Pea Isab
T266 - Inca Pea Dido
T266 - Dido cap cap Poca
T270 - Dido cap 1 city Poca
T275 - Dido cap 1 city Poca - (Poca ded)

T244 - Lizy DoW Inca / Isab / Gust (Isab target, are neighbors)
T256 - Lizy Pea Inca
T266 - Isab Pea Lizy
T268 - Lizy DoW Inca / Baby / Hiaw / Sejo / Poca

T272 - Lizy DoW Plyr - (like she didn't have enough on her plate???)
T272 - Huns DoW Plyr - (neighbors, legit though no actual invasion, mad buildup but mil==, very weak assault)
T275 - My first DP in this mod, wow, with Inca!
T276 - Dido DoW Inca / Plyr - (going after me I guess)
T278 - Plyr cap 1 city Huns for 1 turn
T284 - Inca Pea Lizy
T285 - Plyr knocks out Port, raze city and all its god-forsaken citadels
T288 - Lizy Pea Plyr
T290 - Sejo DoW Inca / Plyr - (I am target, never actually invades me or does anything)
T291 - Sejo cap 1 city Lizy - (and that's why, army is on other side of territory)
T294 - Plyr cap 1 city Huns - (for keepsies this time)
T294 - Dido Pea Plyr
T297 - Inca Pea Dido
T297 - Plyr pea Dido / Huns
T297 - Hiaw pea Lizy
T298 - Baby Pea Lizy
T299 - Plyr cap 1 city Sejo
T299 - Sejo Pea Lizzy

Discussion:

The T084 Dido-Gust war was sensible. She apparently did very well but never captured a city.

The Huns' T92 war against Portugal was likewise smart, as was the Inca dogpile, and far more successful for the Huns, which took a very nice city (a spot I was planning to settle as a priority).

Then the warmaking got ridiculous. The T116 Gustavus war against Hiawatha was nonsensical. He had a LOWER military score (I was scared of Hiawatha at this point) and had done zero buildup. Dido decides to dogpile against Gustavus, and the Huns dogpile against Hiawatha. A fun little war with no clear results.

The T157 Huns attack on Portugal was very sensible and well executed.

The T189 Gustavus war on Hiawatha was about as sensible as the first though at least he tried building up this time (to mere parity).

My T201 war produced a reasonable amount of warmongering penalty and was over quickly.

The T217 Dido war against Isabella made absolutely no sense despite a clear military advantage. She was nowhere near Isabella outside of a cross-ocean invasion. Nothing predictably happened. Guess this is a "diplomatic war"?

The T230 Gustavus war against presumably Babylon (despite the DoW going to Lizzy) was at least prefaced with a buildup and a marginal military advantage, but bringing in Dido against him, who has been the military powerhouse of the game and a neighbor of his, looks misjudged. I watched some of this, and a meager effort to invade Babylon was attempted, though he didn't seem to put too much pressure to bear on Babylon. I didn't see the Dido side of the war.

The T238 Huns war against Inca (DoW went to Isabella) made perfect sense and they had a clear military advantage going in, though without any buildup beforehand. I watched a bit of this, and there was at least a good deal of fighting at the border.

The T243 war of Dido was against Pocatello mostly it seems, though Inca got the DoW. Dido did a huge buildup and had clear military superiority over Pocatello, though not with the Inca included (they are all 3 neighbors) - Inca seems to have not gotten involved much. The outcome of her totally wrecking Pocatello was a foregone conclusion.

The T244+ war of Lizzy against Isabella was sensible, she had a rapid buildup and clear advantage going in, and none of the other civs were nearby, but she got wrecked with a quickness. The T268 war against Sejong when she was already at war and heavily beat up made no sense, and Babylon was just a few tiles shy of being her neighbor to boot and quickly marched an army through Sejong to get to her and mess her up more (one instance of a DP actually resulting in real help). The T272 DoW against me was sheer pointless madness, I guess a diplomatic war. We are not neighbors (for now).

The final bit of insanity is Sejong's DoW against me. I suppose he figured I was preoccupied with the Huns and Dido, however Dido should have been completely discounted as she was half a map away. It just opened Sejong up for me to march to his capital. I had a stack of negative penalties from him by this point, so it made sense in terms of basic diplo mechanics as well. I suppose we can't fault the AI for not realizing a human will take a DoW as an open invitation to their capital. It made sense if I was an AI.


Conclusions:

Overall, the AI is choosing DoW targets wisely vis-a-vis DPs, using the weakest link in the DP chain as their DoW target, even if the intended target is another member of the DP. The AI is doing a mixed job of actually preparing for wars or for choosing wars with a clear military advantage for itself. Gustavus and Lizzy were particularly atrocious at this, while the Huns and Dido managed their targets very well. Unsurprisingly, the latter group has expanded well into dominant positions. Gustavus and Lizzy appear to be insane from a strategic-diplomatic point of view. Something needs to be reined in there - the Huns should not be the sensible moderates in the room. It was nice to see some civs were totally peaceful, like Isabella, Pocatello, and Babylon, while others were constantly stirring things up, like Dido, the Huns, Gustavus, and later Lizzy. Inca took one opportunistic dogpile early on and kept to themselves otherwise.

Sejong's DoW against me, while it made sense in terms of our relations (lots of tensions), ended up being a huge blunder, and seemed more for diplomatic points than anything else.

The instances of the AI declaring aggressive war and basically not doing anything either to invade seriously or build up their military were constant, and one of the weakest points currently. If the AI is going to war, they need an actual army that can win a war and to then actually use that army from the get-go. Especially against a human. The Huns and Sejong both DoWd me without a clear plan of assault. The Huns had at best a fifth of their military nearby me, and sort of just shuffled 3-4 units at me at first, easily repelled, then just shot random 1-3 unit piles at me after that, also easily repelled. Sejong DoWd me and literally sent nothing at me and had his closest city on my border guarded by like a single pikeman and hwatcha. The path to Seoul is wide open right now for me.

It seems sometimes the AI does a huge military push with clear advantage, and sometimes just flubs it, even when clearly aggressing. I've been wrecked by dogpile AIs (prior game) and had them just sort of sit around waiting for me to attack back.

On map, I'm Russia. Portugal was my southern 2 cities, Pocatello was far West, now Dido.

Game ended poorly after I took out Seoul. Warmonger penalties were extreme at that point (I think -200 for one), plus I was again getting wonderspam penalties in the -40 range despite a very middle of the pack amount. Seems very unbalanced on the latter.

I fought an endless stream of defensive wars after that point, alongside sanctions, against England, Spain, Attila again, and Hiawatha. I still managed to turn them all away and get my economy into shape for a late-game run at a time victory (I generally ignore VCs, and with so many AIs on the map the game usually doesn't get far enough along anyway).

And then the grand finale: a triple team from Hiawatha, Spain, and Babylon out of nowhere (friendly entire game, he just flipped on a dime late-game). That was too much to handle, and I peaced out with half my cities.

Never really got a chance to cool down the warmonger penalties. It seems you have to either have a great strategic position and keep a massive army around, or be very sparing with conquest. Being in the middle of civs makes you an endless target, and taking more than a couple cities makes your diplo penalties ridiculous.

Certainly, I could have just sat back and tried to build away post-Portugal conquest, but I doubt the end result would've been different. Isabella was gobbling up everyone around her and I was next in line, Hiawatha hated me all game especially after Portugal, and Babylon flipped last-minute from lifelong-superfriend to worst enemy which is obnoxious.

Not quite sure how to handle this mod now. This is the second game where a giant dogpile ended the game short (well T450 in this case at least). It's great in terms of AI aggressiveness, but actually managing diplomacy is a lot more... chaotic and key aspects seem locked out for me (DPs and trading DoWs).

At least I have unwinnable wars again. I just wish there was a clearer path to avoiding those outcomes other than "never take territory yourself".

Sorry for taking so long to respond to this! The detailed feedback was highly appreciated.

You highlight three issues that I plan on working on further:

1) AIs declaring war without sufficient preparation. Part of this issue should be solved by my most recent fixes. As for the rest of it, it's likely some of this is due to brokered wars - which, due to how they work mechanically, necessitate a DoW without preparation beforehand. This has always been prone to issues and exploits. I'll do what I can to improve this in my diplo interaction/trade logic rework.

2) Humans having difficulty getting DPs and using other advanced options. The problem here is that the interaction/deal logic has a lot of problems. AIs will check for DPs with each other every so often automatically, but humans get no such notification when the option becomes available. In addition, deal logic is...problematic. This is another issue I plan to address in an upcoming rework.

3) Lategame AI war pileup. Currently AIs get a large aggression boost when someone is close to winning the game; this is their "panic mode". If this mechanic prevents a player from winning the game, this is in fact working as intended - the AI is playing to win. I can look at other ways AIs can work against other players (particularly if they have no chance of winning themselves), but the mechanic applies to both human and AI players and prevents the endgame from being too boring and predictable.

I could add a diplomacy option for disabling the lategame panic mode without disabling victory competition entirely, though, if anyone wants that.

Thanks again for your feedback, and sorry for taking so long to reply.