Ugly Borders
- Thread starter emperorcharlesV
- Start date
-
We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
It doesn't need to be like that though; IMO just upping the value the AI gives to border provinces, along with provinces in the same state of provinces they are already demanding, would be better, as it wouldn't be so determinant, just an influence.
Say, if border provinces, or provinces in states that have provinces that are already being demanded by the AI are given 200% the value they would normally have, or even 1000%. That could very well solve the problem without forcing the AI to never take provinces that aren't contiguous.
Et voila. Add the various warscore values of province, and I'm pretty much sure that we have the working AI algorythm of Vicky1. The problem isn't so much geography, but warscore. Take EU3: after the asthonishingly quick annexetion of Pommerania, AI Brandenburg is giving a harsh treatment to the filthy teutons (a bit thanks to the opportunistic poles) and wants East Prussia. Now, instead of warring around like expected, the Order went to buisiness and managed to open a trade center in Danzig (beats me what they are selling there). Fact is that all the suddently, the WS (warscore) value of this single province gets over anything reasonable, forcing Brandenburg to settle with 2 useless provinces in Estonia.
In vicky, you can witness similar happenings when France DOWs Prussia in 1871: I don't know exactly how provinces are valued, but somehow just Strassburg+Metz means 80+ Warscore, which is only possible if you occupy nearly all of metropolitan France. AI Prussia is certainly not willing to wage a 10 year war, and settles with some remote colonies, after a few skirmishes.
That's why we had events in Vicky, and that's why the VIP team chose to handle wars exclusively via scripted events....
That is a good beginning but it needs more detail. For example, having the Ai first choose a border province and then add random other provinces in the same state will still sometimes result in strange borders. One suggestion could be to increase the AI's value of a province depending not only on if it borders one of his own provinces but also how many of his own provinces it borders. Include an added ai-value for every provinces that it has already selected in the peace offer and we have a nice boundary line!
Well, looking at the map posted, the Congress of Vienna set the borders in Germany. Fair enough that non-German nations generally went after contiguous conquests.
The AI in EUIII pretty much always demanded contiguous territories, so I have little doubt that the same AI logic can be utilized.
I think part of the problem is that you aren't able to see a map of the things you're asking for.
Like I as the player need to write down all of the prov names I need in order to actually get what I want.
But beyond that, yeah, the AI cares wayyy too much about warscore. I think that if warscore were to be more transparent it'd be a lot better--but currently, I dislike that you need to occupy all of a country to get a single state. Also I think that you should conquer via states, but whatevs on that one.
Like I as the player need to write down all of the prov names I need in order to actually get what I want.
But beyond that, yeah, the AI cares wayyy too much about warscore. I think that if warscore were to be more transparent it'd be a lot better--but currently, I dislike that you need to occupy all of a country to get a single state. Also I think that you should conquer via states, but whatevs on that one.
Indeed, the territories taken by the ai needs to change. The idea put together by andrelvis and me earlier in this post would probably solve that with a not-too-complicated set of changes.
And Otto ftw in that nice map of yours by the way!
I agree, it is very annoying to have to conquer most of China when you only want a couple of provinces from them. Making it a lot harder to occupy provinces but adding to the warscore for doing so might be one way to solve this problem..?
Vector graphics are how 3d rendering engines work. All but the most specialized and obscure 3d rendering engines define a 3d "world" as a set of primitives, be they 2-dimensional or 3-dimensional. a Sprite (In the traditional sense, not the wierd paradox sense) is an example of a 2-dimensional primitive; it's a texture projected on to a 2-dimensional plane, with the 2-dimensional plane extant in either 2-dimensional or 3-dimensional space. A sphere is an example of a 3-dimensional primitive.I think that would easily eat processing power.
Especially with all the fuzz about HoI3 being slow (and single-threaded), I don't think that is a very good idea. Since many Paradox game fans, don't have the newest über newest power desktop, and just imagine the uproar if Vicky2 would be even slower than HoI3 on a average pc.
Secondly, I think it would be a bit more time consuming - since vector graphics is a bit more complicated than pixelbased - and also because Paradox has invested time and money in pixelbased maps (and learning how to make them) and can reuse these maps.
But if/when the Clausewitz engine gets proper threaded (and when average processing power increases) - then I'd love it, obviously
And lastly: I have yet to find vector graphics in games, and on web (anything but flags and other graphics on wikipedia).
A terrain map such as the one we use in Paradox games is generally represented in-engine as a mesh of polygons. These are two-dimensional primitives projected in to a 3-dimensional world. HOI3 renders the terrain this way (as far as I know), as does EU3. Defining the map in a more "native" format could potentially increase render speed; it'd mean fewer superfluous trigons being generated due to the square pixels.
I'm going to go out on a limb here (if a PI dev, or maybe a modder, knows I'm wrong, by all means correct me on this next bit). I'd wager that the map is only read once, at startup; Then, it's loaded into RAM as a mesh, not as a bitmap. The bitmap file is then "discarded", as it's no longer needed. This means the only performance changing the definition format would affect would be load performance. Loading the map is a relatively simple operation to begin with, and converting a vector-based image into a mesh is likely to be faster, especially with the interpolation that seems to be applied to the current map; this could be eliminated with vector graphics.
All I'm saying is it's not cut and dried that defining the map as a set of vectors would hurt performance. The one concern would be, of course, making sure the modders could all handle vector maps; One big way to increase the ease of modding would be to use an open standard like SVG.
By the way, the only reason you don't see SVGs in more places is because microsoft is pig-slow at adapting open standards ; they've only had a decade to code support for SVG (and if Internet Exploder doesn't support it, then web-devs get cold feet about using it. IE's market share has shrunk, but it still amounts to 50% of all users)! But no worries, the next version of IE, IE9, will support SVG, and I hope we'll see a lot of web devs celebrating the fact that 90% (some people will stick with their IE8,7,6, etc) of web users will be able to render SVGs directly in-page.
That said, SVG support is still pretty terrible. This chart gives a graphical representation of what browser support what percentage of the SVG 1.1 standard, which became "recommended" by W3C in 2003.
Note: Sorry if the links make it seem like you're reading a Wikipedia article; I guess I've picked up some wiki-habits along the way.
It'd be nice to get some feedback on this PI. Like if it's even being looked at or not.
I too would like at least a quick feasibility assessment from one of the devs. Even a "yeah, not happening. <one sentence reason>" would be great.
One irritating example of this that shows up even in Ricky is Britain taking a bunch of random Russian provinces due to the Crimean war. Bits of Estonia, Finland, Siberia, etc... It just looks weird, and plays weird.
I think that the 'war with aims' is a system that is supposed to be preventing that, so at least we know that Paradox is looking into it.
This was usually a problem with the AI in some Paradox games not taking connectivity into consideration for peace deals. Barring very special circumstances, it's pretty much never desirable to take a province without sea access which is disconnected from the main territorial body. I agree a good weight metric for peace negotiations (among others) is the ratio of neighbouring friendly provinces to enemy provinces the demanded province would have, but I bet Paradox already considered this and that there are further complications. I also agree that in the previous games, the AI was too stalwart about warscore; improved connectivity for a certain deal (for both AI and human) should offset warscore a little.
I suppose you could also force connectivity through other means (i.e. in the game rules rather than through AI behaviour) in peace negotiations. Judging from some comments here it appears Paradox already have done this, I haven't stayed informed of the latest news.
As for the AI demanding connected but still 'illogical' territories, well I guess there's a discrepancy between what you and the AI consider logical provinces to demand.
If an AI starts a war in the "wrong" area and wins some provinces it will probably end up looking illogical with regards to culture, religion etc. no matter what.
I suppose you could also force connectivity through other means (i.e. in the game rules rather than through AI behaviour) in peace negotiations. Judging from some comments here it appears Paradox already have done this, I haven't stayed informed of the latest news.
As for the AI demanding connected but still 'illogical' territories, well I guess there's a discrepancy between what you and the AI consider logical provinces to demand.
If an AI starts a war in the "wrong" area and wins some provinces it will probably end up looking illogical with regards to culture, religion etc. no matter what.
Last edited:
Frankly, I would understand if the folks from Paradox don't want to encourage people to give suggestions by acknowledging them. They'll do plenty of suggesting regardless and often not too sensibly. The developers may read the suggestions or they may not, they may find it helpful or they may not, but in the end it's all up to them and you shouldn't expect anything.
This was usually a problem with the AI in some Paradox games not taking connectivity into consideration for peace deals. Barring very special circumstances, it's pretty much never desirable to take a province without sea access which is disconnected from the main territorial body. I agree a good weight metric for peace negotiations (among others) is the ratio of neighbouring friendly provinces to enemy provinces the demanded province would have, but I bet Paradox already considered this and that there are further complications. I also agree that in the previous games, the AI was too stalwart about warscore; improved connectivity for a certain deal (for both AI and human) should offset warscore a little.
I suppose you could also force connectivity through other means (i.e. in the game rules rather than through AI behaviour) in peace negotiations. Judging from some comments here it appears Paradox already have done this, I haven't stayed informed of the latest news.
As for the AI demanding connected but still 'illogical' territories, well I guess there's a discrepancy between what you and the AI consider logical provinces to demand.
IMO they should drop the concept that some provinces are worth more than others. The war score mechanic is the reason why the AI will demand scattered provinces - it maximizes the "worth" of the provinces being demanded, other considerations hardly play a role at all.
What the AI should do instead, is: first pick provinces with compatible POP cultures. Then, when all compatible-cultured provinces are grabbed, consider only the military-strategic value of the provinces!! I.e. pick a number of total provinces or states to be demanded, and choose these so that the resulting province transfer minimizes the length of the common border. That means, pick the provinces to be demanded in such a way that it consists of the least possible number of "border sections".
Maybe also modify this so that terrain types are taken into consideration as well:
- a border section that has open terrain on both sides and no river in between has a base value of 1.0
- a border section that has open terrain on the enemy side, and hills terrain on the friendly side, has a base value of 0.8
- a border section that has hills terrain on the enemy side, and mountain terrain on the friendly side, has a base value of 1.2
- a border section that has open terrain on the enemy side, and mountain terrain on the friendly side, has a base value of 0.7
- a border section that has mountain terrain on the enemy side, and open terrain on the friendly side, has a base value of 1.4
- a border section that has a river running in between counts 66% the base value
That should give you good, defensible frontiers and it would bias the AI to pick borders that give it a strategic advantage: Defensible terrain on your side of the border, open terrain on the other side, least amount of provinces, and try to get a river line too.
Basically it's what I do when I demand provinces.
I second that. I have already thought for a long time that it is certainly possible to make the AI prefer provinces that give a land connection. And perhaps an AI is possible that prefers even borders and avoids extreme shapes if possible.