Adjusting unit classes

[Nerei]

Based on talks with George Geczy on Discord here is a few ideas for unit classes. First of yes I know chances of it getting into the current games is basically nil but there is always Next Gen.

Currently we have a limited number of unit classes. This leads to some issues if we want the AI to have a somewhat realistic composition of its armed forces.

Lets say we look at world war 2 US. The US has a colossal pile of capital ships. Everything from light cruisers like the Clevelands or Atlantas to large battleships like the Iowa or Montana are in this group and a lot of these classes where build concurrently. Making the AI build a varied capital ship navy will with such a list be hard as you will have to convince it that there is not just one "best" design but a lot of designs to go for. Even if it manages to build a varied navy it could end up being a varied battleship only navy if that is what the math tells it is best.


Lets instead assume these are split into multiple subgroups. Having the AI then make its navy consist of say 10% battleships, 40% Heavy cruiser and 50% light cruisers would be easier. It would involve making it build build 50% with tag X 40% with tag Y and 10% with tag Z.
Sure there is some complexity in doing that and yes it is not an ideal solution but it should be easier than making it dynamically pick a varied fleet that is varied without it going for trash or ending up with just say battleship sized capital ships.


Yes having too many classes will be confusing but it does not have to be done by adding more unit classes. To me that might actually be a sub-optimal choice.

Lets say it is instead done by giving each design a 32 or 64 bit int variable to represent subclasses.
If these subclasses are assigned as a geometric progression it would mean it is possible to assign designs multiple subclasses by adding numbers together and determine what these are simply by this number.

E.g. if the geometric progression starts at 1 a Guided missile cruiser has a subclass value of 7 it would mean it has the classes 1, 2 and 4. This could be say light cruiser, heavy cruiser and battleship respectively and the vessel would be a candidate for all of these classes when the AI decides what to build. In this case it could facilitate the AI going from a world war 2 style navy with different sizes of gun armed capital ships to more modern guided missile cruisers.
The advantage in this situation over say just making a separate guided missile cruiser subclass is that you do not need to teach the AI to pick that over say heavy cruisers as that class of warship dies during the cold war.

Likewise it would be possible to have both medium and heavy tanks exist together until heavy tanks goes obsolete as medium tanks morph into main battle tanks simply by tagging these MBT's as both medium and heavy tanks.

It would also be a way around things like trying to make the AI build Type 26 frigates and not just Type 45 destroyers by making the Type 26 a patrol boat.

Yes it is not a perfect solution but I personally think it might be a good way to make the production AI appear smarter without having to go through the tough task of actually having to make it significantly smarter.

[Rosalis]

The idea can be ok, if they want to research 20k+ units in what category they fit. I think SR2020 had this in some way as there was more variety in compositions by the ai. Your example of navy makes it pretty useless to me. Only a couple of regions can benefit from this. The rest may be happy they get a design if they research something. As player only thing that mathers is get the most subs (still have too loose one to an ai ASW heli). Those roles can be improved alot and biggest competition for the US navy is Russia which sends it ships 1 by 1 against way to many enemies. Availability to ships on the market is a much bigger issue in my opinion. Same with the F35, who gonna produce it? Not US no mather your proposition, unless the superpowers export a % in peace time and more regions get the design. The whole navy thing is pretty useless. Embargo and economy wars could make it more important instead of this and in return actually produce ships.

Special forces is a much better example i think. Instead of going for hard infantry units, the soft units can be made more important, because of deadly cloose combat and used for that scenario. But still that doesnt change the fact that those special forces need to know what there role is, else the only thing your doing is make the ai weaker. I guess if you look at the game as a museum it may be worth it. Or the famous vodniks light infantry, instant repair, fast production, good to disrupt enemy lines, whats not to like? I can see it being used in your proposition before moving to navy with much greater effect. Thats how i play, give the ai a balanced military and economy. It should be the game doing this, but atleast im able to do it.

Start simple i would say and expand in less priority areas when you see fit. Adjusting unit orders is a bigger issue i think, since that is almost not happening. Infantry does it well and tanks, but the rest.. ai loose tanks and infantry and moves artillery and aa to the frontline where it can be seen by the enemy. In the final patch atleast planes are given roles again instead of sitting at the airfields for 5 years.

So far the % doesnt really add up on missiles, since ai always prefer to build nuclear even if it doesnt have a missile silo or build one (iran/north korea). Let alone use them. Meanwhile its eating all the strategic bombers and missile launch subs of US and Russia in this strategic pool feature. Even seen US out of capital ships, checked and yup 50 capital ships in the strategic pool for when that rare moment happens. One way to balance, but yeah. I prefer ai cheating in that rare case. For those that want to know lock minister control on those units and ai wont move them there anymore, not that the ai can use them, but atleast its easier to distrubute them when you log in as US.

[Nerei]

So a significant part of your argument is I should have picked a different example? If I pick ships it is bad but if I had picked infantry it would be okay?
There is nothing that prevents this system from working on any type of unit class. That is the point and why I suggest it over throwing in more unit classes like say leg and motorised infantry. See I used infantry in this example.


It should also be said you say start simple then bring up a more complex issue. Tactical AI at the level of detail this game deals with is anything but simple which is what you need to work on with your artillery and AA example. Just try and make a list of what you need to consider and possibly test to do that. That list will "not" be short and some of it will be quite complex things. There is also the entire performance issue which if you tank will annoy some people.


I also love how you argue if people want the game as a "museum". Lets say that is what some consumers want. Is that wrong? Are you "seriously" arguing your way of playing is the right one and theirs is wrong?!
It is fine to have an opinion but you seem to have a problem with something going against yours.

Not going to comment further on this as you seem to want to touch just about everything. If you think it is worth mentioning why not make a topic about it?

[SGTscuba]

I do think this idea really has merit. For us, it is quite easy to make distinctions but AI very much could do with some extra "hints" in the various directions it needs to go and any additional information it can have will help it along.

Somehow I also think the AI needs to understand in depth a bit more about its requirements somehow, and tying it in with this would make sense. For example, the USA would need more longer range ships, and tankers to support it as its likely enemies are an ocean away. For somewhere with only local enemies such as Germany, the AI would be better focusing on shorter ranged, but more capable ships. This likewise should transfer over to land and air units, for example a country with lots of jungles or forests or mountains would be better off with a bias towards leg infantry, but a country with wide open plains would be the opposite. Maybe this could be handled on a per battlezone case, and then the AI would sum this information to make its overall decision.

It may also facilitate using different sized facilities which have limits on what they can produce. This would allow the AI to make some other decisions about what it can actually afford to support and maintain.

[Rosalis]

So a significant part of your argument is I should have picked a different example? If I pick ships it is bad but if I had picked infantry it would be okay?

With the difference that infantry units are actually on the map, most of the navy is not. Also producing navy is considered very low priority by the ai. There are more issues then this. Its all connected.

It should also be said you say start simple then bring up a more complex issue.

Start simple in let ai produce special forces, let them prefer cloose combat areas, stay there untill a city or other cloose combat forward is being attacked by other friendly units and move on to the next. Ships dont win wars, so less priority in improving the ai. Game also prefer long range artillery, you need to find arguments why it should produce other types of artillery. Cost is a good one, defensive for mortars, offence and missile. So 4 types and your done here with potential to connect other parts of the game.

Tactical AI at the level of detail this game deals with is anything but simple which is what you need to work on with your artillery and AA example.

Yes i know, but it helps if those units survive for your museum. There are plenty of games where LoS is actually taken into account. No LoS for a friendly tank or infantry or even a city? No need to move forward. Damage output is 0.
Removing ai stance and implementing the basics for units shouldnt be too hard. Not every improvement for the ai tanks the game. I find this a very weak argument, you should be able too implement the basics else find another way to represent the units. They got order to capture or atleast move in enemy territory but will never complete their mission, you think that doesnt tank the game? Maybe with 5k more different artillery and aa you will notice it will tank the game.

I also love how you argue if people want the game as a "museum". Lets say that is what some consumers want. Is that wrong?

No, but if you like museums go out the door and visit those. The game shouldnt suffer because someone want to see units that are easily killed in the game. There are enough features for the player already and issues left that prominent community members are gonna request more.

Im not saying anywhere that my way is the way to play, im just saying you can play like you propose here. I dont have to open topics about the issues i mentioned, they are already there or already known if you play the game, like my arty and aa example.