Report Questionable Behavior

[r_rolo1]

We all know that the vassal mechanics are a little kinky, but this is actually a good move from Ragnar. Why would he wait for having more soldiers killed and more land lost if he could get a free from jail card?

 

[Lone Wolf]

but this is actually a good move from Ragnar.

And a bad move for Frederick, who most likely will lose his vassal soon.

 

[r_rolo1]

Not necessarily. Even if the vassal is temporary, it will increase his power rating, give , and give acess to more resources. Ok, it will hamper the diplo and make the maintenance high, but this is pretty much a case to case analysis.

Now the big problem is that the AI looks to never make that kind of analysis when thinking on acepting a vassal......

 

[Lone Wolf]

Not necessarily. Even if the vassal is temporary, it will increase his power rating, give , and give acess to more resources.

For how many turns? It seems that it'd be more profitable for Frederick to continue pushing on, taking Ragnar's cities. Unless Frederick had no army at that point.

 

[r_rolo1]

You are only strengthening my point, lone wolf. Capitulating a civ is a complex decision to make, with lots of pros and cons in every turn and the best point to do that it is not obvious from a first glance, even for humans, most of the times. In this case we don't know if Frederick had conditions to push the offensive or not or if it was his "wish" to simply finish the war that was dragging him down.....

 

[BobTheTerrible]

But the problem may just be that the AI saw "capitulation" had opened up as an option and went for it without weighing any of these possibilities. If this was the case, it certainly backfired, since Ragnar only needs 1 more pop to break free.

 

[r_rolo1]

Well, surely....

But my point was not that. My point is that we don't have enough intel to assert that acepting capitulation was a bad idea from Frederick. It might have been, but it might had been the best choice given the circumstances as well.

OFC that , if the AI acepts simply a capitulation just because it can, we definitely need a refinement in that part of the code

 

[jdog5000]

In BBAI, an AI will press on with conquering after capitulation becomes available if it is about to take a "good" city, it has been making steady gains (stagnation == not enough troops), and/or no other player is threatening good cities of the would be vassal. The exact conditions depend on personality.

In this case, since Ragnar is at war with so many other civs and losing badly, he's going to offer capitulation to those who have been killing his units and cities ... loses by Ragnar's vassals don't factor in to Ragnar's decision. Sounds like Frederick was the one who had dealt direct damage to Ragnar, so that's who the offer was made to (could have been other open offers out there).

Frederick would wait to accept so long as he is on a roll or about to capture more high value cities ... but, since there are so many other civs involved, if the human player or another civ is about to capture one of Ragnar's cities, then he'll accept capitulation so his vassal is stronger and not one of his potential enemies.

Hope that helps explain the AIs actions.

 

[joncnunn]

Questionable behavior 1:
As Persia doing an Immortal Rush I was able to lure an Egyptian Swordsmen away from the main battle with a Scout by keeping it visible to the Sword for the entire war.

Since the Swordsmen was the only unit they had that has > 50% chance of winning against an Immortal, it really sped up their conquest defeat.

Questionable behavior 2:
Building an ancient wonder (in this case Temple of Artemis) when next to both Persia (Immortals) & Sumeria (Vultures). Any human playing Egypt in this situation if he didn't reach for the restart button when he saw no Horses would have built units instead of wonders.

I would call their prewar unit choice builds (almost exclusively Archer when next to Persia) questionable as well except they were screwed either way thanks to Sumeria.

Questionable behavior #3:
The Incas were in this game. Shouldn't they have Q rushed their nearest opponent? Or is that something they'd only do if "Aggressive AI" had been turned on?

 

[OnmyojiOmn]

I'm playing Ghandi on Noble, Better AI 0.75, Boudica next door. I declared and razed a small city of hers that was blocking my expansion, then made peace a few turns later. A little bit after that, I noticed she'd founded a tundra city nearby so I took a spearman over and razed that, one turn before she got an archer into it. I made peace again, and she founded another city in that same tundra spot, again with just one chariot defending. Razed.

Shouldn't the AI be guarding new cities better than this, especially when founding next to a civ that's already declared war twice? Boudica had Archery and a decent amount of production in her capital, plus this new city spot isn't that far away. The AI should be smart enough to build enough units for defense rather than settle ASAP in an area that isn't valuable to either of us. For that matter, in this situation the AI should turtle up completely and go straight for catapults.

 

[MadmanAtW]

especially when founding next to a civ that's already declared war twice?

This would require civs to have memory, which they mostly don't... but it would make sense for a civ to take into account whether or not they current have the "you declared war on us" or "you razed one of our cities" diplo penalties with you when considering what to do on your borders.

 

[joncnunn]

A small thing, but still questionable.
This was under 0.75:

Isabella requests me (who has the Christo Redeemer Wonder) switch to Theology.
I accept.
The very next turn I switch back to Free Religion.
I still have the +1 you accepted our civic bonus.

I think the AI needs trained to either not ask the player with Christo Redemer to make civics & religious changes or else remove the positive relationship bonus if they switch right back.

 

[bestbrian]

Referred by Phungus:

Playing LoR (latest build) and observed Sury moving a stack in the BCs with GGs attached to Rams. Isn't the AI coded not to attach GGs to Siege Units? Seems a bit of a waste.

 

[Cybah]

You should simply remove the "PROMOTION_LEADER" from siege units in Civ4PromotionInfos.xml. Don't think this is an AI glitch.

 

[OnmyojiOmn]

The way the AI handles barb cities seems a little weird. A single unit will just park itself next to a city for a huge number of turns while a stack is sent to capture that city, which is often in a marginal area on the other side of the world.

Also, does the AI ever try for the circumnavigation bonus? Naval combat AI is 100% broken if it doesn't, regardless of any other changes that are made. The AI should have galleys out early on, exploring as far to the east and west as they can go. If it thinks it's hit Caravels early enough and nobody else has circumnavigated, it should send one in each direction and they should ignore all other goals. If the AI is unable to remember whether anyone has circumnavigated, then the naval AI is broken. Galleys should be sent in stacks of 2-3 and they should be upgraded to a single trireme (delete the others in the stack) if they happen to be near one of that civ's cities when it hits Metal Casting. Obviously triremes should be sent out if all of the civ's galleys are sunk and there's still coastline to explored.

Sailing is serious business.

 

[Lone Wolf]

The AI doesn't understand the purpose of wonders like Forbidden Palace, building them it cities just right next to the capital.

 

[jdog5000]

Too true ... this is another symptom of the AI building wonders/national wonders in the first city that meets some basic criteria to consider building it, instead of weighing which would be the best city.

Progress on this is definitely in the works.

 

[HardCoder]

Played 3 continents on a huge map with Gandhi and Darius, as Tokugawa to see if I could manage to keep up with their building. As it turned out I did fine, but I didn't switch over to culture fast enough and although everyone had pretty much the same score, Gandhi was headed for a certain cultural victory in the mid 20th century. I built UN and for the heck of it proposed a diplomatic victory. Gandhi voted for Darius (who was definitely not going to win), game over.

Seriously, a computer player should not be abandoning a *certain* victory.

 

[glider1]

Probably already covered somewhere but thought I might just comment anyway. Although the naval AI is seriously better than it ever was and is now possibly the best naval AI ever written for a PC, there are a few little stupidities that I can submit uploads for in some form or other. The AI will occasionally decide to invade amphibiously (some probability of this compared to an adjacent square landing) and that is fine. Sometimes it will invade, capture the city and move it's entire transport and navy into the acquired city and wait. This is not bad either except that the AI mistakes how well the newly acquired city is defended and the city can sometimes be recaptured easily, sinking the entire naval division and sometimes a dozen transports as well. It simply needs to make sure it moves these naval divisions out of the city if the odds of holding the city are small. It can mean a very expensive loss of naval strength to an AI.
Cheers.

 

[CivFanCCS]

The way the AI handles barb cities seems a little weird. A single unit will just park itself next to a city for a huge number of turns while a stack is sent to capture that city, which is often in a marginal area on the other side of the world.

Also, does the AI ever try for the circumnavigation bonus? Naval combat AI is 100% broken if it doesn't, regardless of any other changes that are made. The AI should have galleys out early on, exploring as far to the east and west as they can go. If it thinks it's hit Caravels early enough and nobody else has circumnavigated, it should send one in each direction and they should ignore all other goals. If the AI is unable to remember whether anyone has circumnavigated, then the naval AI is broken. Galleys should be sent in stacks of 2-3 and they should be upgraded to a single trireme (delete the others in the stack) if they happen to be near one of that civ's cities when it hits Metal Casting. Obviously triremes should be sent out if all of the civ's galleys are sunk and there's still coastline to explored.

Sailing is serious business.

Last three games the AI has beaten me to circumnavigation bonus. All large maps.