Castles should have zone of control?

[Zyffyr]

You can get lighter rations, but they are dehydrated, obviously not really applicable in a medieval context. Actual (wet) rations weigh more than the 3lbs. per day I stated; MREs are 1.8lbs and you're supposed to eat three of them per day. And I mentioned the other method of obtaining food from hostile ground: foraging. Stealing it from the locals, in plainer language.

As a specific example of pre-modern military rations, the approved standard ration of the Continental Army in the US Revolutionary War was :
" one pound of beef (or three-quarters of a pound of pork or one pound of salted fish), and one pound of flour or bread per day; three pounds of peas or beans per week, one pint of milk per day, one pint of rice per week, one quart of spruce beer or cider per day, and a little molasses. "

 

[Bannerman Man]

My mod Party AI Overhaul and Commands already has a feature that makes the AI highly prioritize besieging settlements closest to their own settlements, so at least they don't wander deep into enemy territory anymore.

Just out of curiosity, what changes did you make in order to accomplish that?

 

[Captain_Octavius]

Just out of curiosity, what changes did you make in order to accomplish that?

The AI generates lists of potential actions and calculates a score for each action, then executes the action with the highest score. I simply modified those scores based on particular conditions. From my mod page:

Kingdom AI Army Leaders will now prioritize besieging enemy towns/castles closest to their own hometown. Their hometown is usually the most prosperous castle/town that their Clan owns (this is a vanilla feature). The idea is that they somewhat selfishly prioritize taking lands close to their home so that their fiefs becomes safer. Armies won't besiege towns/castles deep into enemy territory anymore. An added bonus is that armies with different leaders will have distinct behaviour.

 

[sanalaismine]

I absolutely agree that castles have to have more incentives in strategical battle. Zone of control for example, is a very good idea. Right now castles have absolutely no purpose as opposed to historical events. Castles should be intimdating and must have a role in wars other than just being conquered. They have no advantage, no control whatsoever.

I would love to see some change on this aspect and I support the OP on this matter.

 

[Bannerman Man]

The AI generates lists of potential actions and calculates a score for each action, then executes the action with the highest score. I simply modified those scores based on particular conditions.

Sorry, I was more wondering about the specifics of what you did.

I know the total target scores aren't strictly based on how close a settlement is, but in native both the distanceScore and settlementImportanceScore increase the closer a target settlement is to the besieger faction's own settlements, so the AI is already favoring nearer targets over further targets (however, a town that is further is often still more valuable/important than a castle that is closer in the case of the settlementImportanceScore). There are obviously other factors that go into the target score calculation that can still lead to a further settlement being chosen over a nearer settlement though (namely garrison strength, food stocks, and wall level).

So I was just curious if you upped the value of the distanceScore to be a more significant factor, or introduced another factor that essentially doubles up on the effect of distance (for instance based on hometown like you say), or reduced the difference in value between castles and towns in the settlementImportanceScore, or something else entirely. Basically, how did you ensure that the closest settlement is heavily prioritized more so than it already is in native?

 

[Captain_Octavius]

My code computes which siege target is closest of all (to the army leader's hometown), and then scales down the scores of the other targets relative to how much further away they are than the closest target. This way they don't always take the absolute closest, if one is a little further away and has a higher score from the vanilla code (due to a lower garrison for example), they might still target that one over the absolute closest.

In practice, usually only the 2-4 closest targets are viable candidates for an Army to siege in my code. Which is exactly how I wanted it. Other armies with different leaders might have a different set of viable candidates though, because their leader's hometown is a different one.

 

[Bannerman Man]

My code computes which siege target is closest of all (to the army leader's hometown), and then scales down the scores of the other targets relative to how much further away they are than the closest target. This way they don't always take the absolute closest, if one is a little further away and has a higher score from the vanilla code (due to a lower garrison for example), they might still target that one over the absolute closest.

In practice, usually only the 2-4 closest targets are viable candidates for an Army to siege in my code. Which is exactly how I wanted it. Other armies with different leaders might have a different set of viable candidates though, because their leader's hometown is a different one.

Cool, thanks for the explanation. Basing it on the army leader's hometown is a nice touch, since native's distance modifiers are not clan specific, but based on the distribution of the besieger faction's fiefs as a whole, irrespective of clan. Which means clan leaders are more likely to go for settlements that will gain them more merit in a fief election because they're closer to their own settlements (acting in self-interest as you say).

Do the scores scale down linearly with distance?

Incorporating the calculating trait could also be neat. For instance, a calculating lord factors in distance to hometown when selecting targets to boost chances of earning the fief in the election, whereas a non-calculating/emotional lord factors in enemy clan relationship values or something when selecting targets instead (vendettas).

 

[Captain_Octavius]

Do the scores scale down linearly with distance?

Yes, rather agressively.

Incorporating the calculating trait could also be neat. For instance, a calculating lord factors in distance to hometown when selecting targets to boost chances of earning the fief in the election, whereas a non-calculating/emotional lord factors in enemy clan relationship values or something when selecting targets instead (vendettas).

Yes all kinds of variations are possible. Valorous leaders might also disregard garrison size more, while cowardly leaders do the opposite.

Right now I only have one variation, and that is that the penalty for not being the closest target is only 1/3 if the closest target doesn't match the army leader's culture but this one does. This way army leaders also have a bias to (re)conquer their homelands over territory with a foreign culture.

 

[Bannerman Man]

Right now I only have one variation, and that is that the penalty for not being the closest target is only 1/3 if the closest target doesn't match the army leader's culture but this one does. This way army leaders also have a bias to (re)conquer their homelands over territory with a foreign culture.

Nice. I know the settlement importance score factors in culture in native, but it will only have an effect of ~10% so it won't be much of a driving force behind behavior.

 

[Aggrobiscuit]

I like the idea, but implementing it now would mean a world map overhaul probably. There's already enough faction imbalance.