havent totally ruled it out, but we are trying to improve ai first for sure
Well, in the realm of AI improvements that would be pleasing to the eye and to the mind:
1. Curb the AI division spam by some combination of a) lowering minor access to high manpower percentages (National Focuses in particular), b) coding them to stop building new divisions when their equipment total in any required equipment type is getting highly negative. Example of that last: peek in on any Allied minor nation in about 1945 and you will see that they have built probably about 8-10 times as many divisions as they had at their historical peak, the large majority of which are still unequipped, and they are running huge negative balances in multiple equipment types, e.g. infantry equipment. An exception might be made for division templates that have the low priority set, e.g. militia type templates, but even there I'm a bit iffy about it, sensing the possibility of simply changing the problem from massive division spam to massive militia division spam.
2. Code the AI to set/have minimum force levels which it will at all times maintain along certain borders. For example, Germany should never,
never withdraw so many of its divisions from the Soviet border (after the fall of Poland) that it can only place a single division in one province out of three. There needs to be some mechanic to force the AI to always maintain certain overriding defensive priorities, no matter what is happening on other fronts. Similarly, the Italy AI must never be
able to send almost all its divisions overseas, leaving too few in Italy to defend against invasion by the UK....or South Africa!
2b. As a corollary to 2 above, the AI should set or have defensive priorities that are proportional to distance from the capital. For example, Japan should place highest priority on making sure it can defend the home islands, then secondary priority on defending Korea and the nearest islands, tertiary on defending the Philipines once it has conquered them, and so on.
3. The AI seems hardcoded to invite to alliance at the earliest opportunity, and then to call to arms about .000002 seconds after the invitation is accepted. This leads to some awkward and even badly ahistorical stuff, even when historical focus trees are enabled. For example, with historical focus trees, Italy is pretty much always an Axis member before war begins. No problem there. But it is also called to war the instant Danzig or War kicks off the action, which should not happen; they should not be called to arms until either, a) France is about to fall, or b) Germany's offensive in France runs into serious problems, e.g. any advance into German cores, etc. Another persistent example: the instant they are eligible to be asked, the USA is always invited to the Allies. This often occurs as early as the first quarter of 1940. Again, the instant they accept the UK calls them to arms. This has the result of pretty frequently bringing USA into the war too early, sometimes as much as 2 years ahead of schedule.
4. AI major nations, especially Japan and UK, have a lamentable tendency to run their major fleets without near enough, sometimes without
any, screens. This one I suspect is trickier, and I haven't looked into why exactly it is happening. I suspect that the AI is just underproducing screens relative to the number of cap ships it has/builds, and perhaps sending too many of them out in dribs and drabs as tiny (and ineffectual) ASW task forces and to escort duties, not keeping enough of them back to fill out its cap ship fleets. And that latter practice may also be depleting them faster than the AI can keep up with.
5. AI nations should be coded to avoid the English Channel during wartime, and never, ever send transports through it unless their destination is adjacent to a Channel seazone. If Italy is retasking some divisions it has currently in the Low Countries, then they should not rush to Antwerp and hop on transports to pass through, and then to the bottom of, the English Channel. They should also be set to avoid the nastier straits if another path exists to reach their destination. Continuing the above example, if those divisions in Belgium are ordered to go to Egypt to reinforce defenses in North Africa against an Allied counteroffensive through Ethiopia (quite common), they should use overland strategic redeployment to a friendly port that has the minimum number of straits and seazones between it and the ultimate destination, in this case by SRing to say Taranto, then boarding their transports to Alexandria, bypassing entirely the English Channel and Gibraltar, even if the Axis currently hold Gibraltar. Similarly, divisions on the US eastern coast should SR to a west coast port in order to reach destinations in the west pacific, not rush to the nearest port and then transit the Panama Canal (not saying this example is a problem or unrealistic, just that it is an illustration of the pathing logic that would avoid large problems in other parts of the world, especially the Med). Put simply a unit SRing through friendly territory is much safer, in fact nearly invulnerable to enemy action, while one at sea is subject to being intercepted and sunk in each and every seazone through which it will pass. So it only makes sense for the AI to shift priorities from using the port closest to the starting location to using the port closest to the destination that can be reached via SR.
Just my own personal wish list of tweaks to fix AI behavior problems.