Like the rest of you, I think a semi-realistic logistics system is the biggest gap in HOI. I'm also a fan of systems that allow you to tweak if you want to dive in, but for the most part, "just work".
The system that I'd propose is similar to Railroad Tycoon 3's economic model.
I'll explain based on how HOI2 works, since I can't map details to the unknown of HOI3. In particular, I want to avoid the factory vs IC debate as it's mostly orthogonal to this issue. I speak only of supplies here, but this method can be extended to the other resources, notably fuel.
Here's the proposal:
Units currently consume a certain amount of supplies each day. Change this a very small supply storage capacity to each unit that is consumed at different rates based on the units activity. For example, a unit with no mission vs anti-partisan vs moving vs fighting would consume supplies at different rates.
Each province that has IC produces supplies. The rate at which a particular province produces supplies is determined based on the amount of IC in that province.
In order to get supplies from factories to units we use "network flow". (Check it out in wikipedia if you're math inclined and transportation theory) Factories are sources and units are sinks. The supply stockpile unit proposed earlier in the thread would be both a source and a sink. (Although I don't know if I like the extra complexity that unit adds.) Each province is a node in the graph. Each day each factory puts some supplies into the network. Each day each unit consumes some supplies in order to keep its storage capacity full.
Supplies would have a speed with which they move through the graph. This speed would be affected by the infrastructure in the province. (Although note that the speed should really be along the edge of the graph, so the infrastructure here would actually be the quality of the connection between provinces) Supplies would be destroyed en route in a province with partisans. This implies that the amount of supplies produced does not correspond with the amount of supplies that all units receive!
Ports should be sources and sinks. This allows convoys to move goods from one port to another. In order to really convoy, there should be a setting to specify how big the "batches" are. In other words, are supplies always moving, or do they sit and wait to reach a critical mass before moving the next step along the graph. This system could be unified with movement across ground too. Would convoying across the ground minimize losses due to partisans? (Here I'd really want default settings for peace time and war time broken out by land vs sea. Perhaps allow the player to tweak if they really want to)
There would be a map mode that lets you visualize the flow of supplies. Railroad Tycoon 3 does this by showing cars full of supplies creeping from the sources to the sinks. An alternative would be to show the graph and colour each edge based on % of supplies lost or supply movement speed.
Offensive supply would simply temporarily increase the unit's supply storage capacity.
I haven't mentioned HQ units in here. They don't fit easily into this model. However, if you add a supply speed penalty for too many units stacked in a province (I can imagine that distributing supplies at the last mile when you've got 20 divisions fighting over the last case of bullets would be a nightmare), the HQ could reduce that penalty. Therefore the HQ is only effective when there are many divisions present. This seems reasonable to me.
The problem I have with the current system is that it's not discoverable. There's infrastructure, TC, ESE and some other stuff that gets combined in a way I don't understand (and don't want to understand!) The only way I observe the effects is if I win a battle or not.
With my proposal it's obvious. "Oh, this division is running low on supplies, I should put it in reserve to build up its supplies before I put it in combat. No supplies means no bullets, I get it." A naive player can just up his supply production rate and still succeed. An experienced player can see that he's overrunning his supply rate with his offensive in North Africa and pause while his really long supply line catches up. Meanwhile, Monty's short supply line is reinforcing his army... An expert player can dive in and discover that Warsaw (for example) has a high partisan rate, perhaps if he stations another Garrison there he can improve my efficiency and use those IC for new production instead. (Can partisan activity be related to province manpower? Putting a large garrison in Warsaw makes sense, other provinces, not so much)
As a bonus, the "magic numbers" ESE and TC are no longer needed! The player now also has a strong incentive to improve the infrastructure in key provinces. The faster the supplies move, the fewer there are in transit, which means less time for loss and more time for consumption. An encircled army is also possibly still in supply. As long as there's a heavy production centre inside the encirclement, some amount of supplies can still be produced and delivered to the trapped army. However, those supplies cannot get out. This (mostly) fixes the encircling of the capital problem.
The big problem I see with my proposal is the computational complexity. Optimizing network flow is a hard problem. It's probably doable as you can make several approximations, but there's a trade-off between this and other features when it comes to CPU time. Personally, I'd prefer to see more CPU time spent on logistics, but I'm biased.
In some other thread, I saw a suggestion for a food resource. That would fit nicely in with this model too. Infantry needs food and supplies. Armour needs food, supplies and fuel. Same with a carrier (don't forget about the navy and their supply lines!) How cool would it be if a fleet met up with a convoy (you can see the movement of supplies in that map mode, remember?) and restocked its fuel from the convoy?
I'd also like to make the observation that optimizing supply networks can be fun. Try Settlers 2 some time (or the open source remake, Widelands).
