And the AI?

[victor charlie]

As an old HOI and HOI2 player I like the extra detail being put into the new development of HOI4.

But with so much more going on, especially the naval being upgraded, how will the AI handle all this?

Of course I know not as good as a human player but will the AI be able to give some of a challenge and make logical moves?

Invasions of unimportant countries and islands and a Japanese fleet sitting off Ireland is what I recall from an earlier HOI and detracted from the AI's play.

 

[bwc153]

I feel like the naval and air game have more depth while also simplifying complexity due to the addition of strategic air/naval regions. As such, I feel the AI probably will handle things better due to actions within those regions being automated and less micro intensive. If a Paradox game has taught me anything, it's that AI is bad at unit micro.

 

[ILoveLamp]

Yeah, the land combat will definitely be much harder to program for, this is where the AI will be at it's weakest for sure. It looks like the battle planer system will try to help even the odds a little, by incentivizing less micro and rewarding the player for following through pre-planned movements. I'm not quite sold on it yet tho. I'd really like to see how a Barbarossa sized operation plays out - whether or not you can actually plan out complex pincer moves, spearheads, etc.

 

[Axe99]

My bet is that the air system will be the easiest for the AI, the land the hardest and naval in-between (in no small part because the air system will have the smallest number of regions, the naval the next biggest and land the most), although in HoI3 the land AI was better than the naval, and naval combat was less complicated there. That said, I think PDS' AI has grown in leaps and bounds since HoI3, and apparently the weather simulation chewed up a fair bit of processing power, so hopefully at least part of that's been let loose for the AI instead. I'm also not sure all the extra detail is something the AI will have trouble coping with - as long as it can build sensible fleets, divisions and air wings.

 

[EntropyAvatar]

I think that air and naval AI will be easier, since the AI will mostly be making decisions about how to allocate forces between regions and between roles. This shouldn't require a lot of spatial reasoning (which IMO is the hardest thing to do in AI) but simpler reasoning about connectivity of strategic regions and relative forces.

I am somewhat hopeful that on land, the battleplan system will help the AI by explicitly breaking the problem into different scales. At the small scale, it has to be able to follow a battle plan, which basically boils down to maintaining or achieving a certain front line. It can be fairly unimaginative about this, just keeping a reasonable distribution of forces along the line and attacking or giving up provinces in a reasonable order and redistributing forces as the situation changes without shuffling them around too much.

At the higher level, the AI has to be able to formulate (and revise!) battle plans. This seems like a harder problem than following a battle plan, but at least at this level it doesn't have to think about exactly how to move individual divisions. One could imagine a bag of possible operational maneuver types where access to each maneuver is scripted to be dependent on having the right forces and force balance. Then it has to choose a set of compatible maneuvers, fit them to the map, and as the plan is executed, periodically evaluate whether the proposed plan still scores well enough compared to alternatives.

There's no particular reason why this division between operational planning and lower-level plan execution couldn't have been implemented in HoI3, but from observation the planning part of the AI, if it existed, seemed fairly rudimentary. I think the design of HoI4 kind of forces this division to be more well-developed.