ORIGINAL: fatgreta1066
I'm playing the 1939 Campaign as Allied. It took Germany until November 25 to conquer Poland. It seemed to me that the AI did a good job of cycling units in and out of position when making important attacks (I.e. against Warsaw). It seemed to not do a very good job in figuring out how to bring Poland down. By comparison, I played 5 times as Germany and conquered Poland in 2 turns 4 times, 3 turns once.
Germany was in good position to invade France by late March, 1940. I held on to France until early August. I thought the AI diffused it's initial strikes a bit too much, not creating a breakthrough when it might have. To be fair, I did a pretty gamey thing by placing the entire French navy (including its Med fleet) on raiding duty against the Norway convoys, as well as some RN units. That cost Germany a number of MPP. I also sent an extra British Corps to France, in addition to the BEF.
Now that Germany has French ports, we will see how the AI handles the Battle of the Atlantic, and the war in Africa.
This post caught my eye so as Poland and the invasion of France and the Low Countries were a portion of the AI that I had spent quite a bit of time on fine tuning.
General speaking, any optimizations here for the AI would not only be good for this part of the game but also for every other part of the game when it comes to effective AI combat.
For Poland, the goal on my end was to attempt to recreate the 2 turn victory that most human players can achieve... essentially this was the bar.
The challenge in Poland for the Axis AI was that while there are many targets and many towns between the Axis start lines to Warsaw, how to ensure the Axis AI would focus enough of its assault not only on Polish front line units, but to also focus its drive towards Warsaw as quickly as possible.
This was a tricky one since there are so many targets it is easy for the AI to get distracted but after a lot of watching the AI play and determining what most human players would do, we did at one point have the AI (if the rolls were good and the weather cooperated) capture Warsaw within two turns 50% of the time. Part of this was achievable through more tightly focused scripting but also by adjusting certain behaviour such as having the AI to hold back its Tanks (not always and only under the right conditions) after the initial wave of attacks. This allowed it to breakthrough a bit further with its tanks and position itself closer to Warsaw for the follow up turns.
This one general behavioral change, combined with many many others, seemed to improve AI assaults quite a bit and also played out quite well in other parts of the map such as France and the Low Countries and so on.
Unfortunately we also then found that Poland was too easy to defeat in general so there were changes that led to requiring the Axis to use most of its frontline units to achieve quick victories over Poland, i.e. rather than being able to move half of those units towards France right away, and this then led to the AI also being less likely to capture Poland in 2 turns on a regular basis.
As it is now, the results will vary from game to game, and although it is extremely rare for the AI (I've run countless AI vs AI debug games during development) two turns is still an outside possibility but 3 or 4 turns is more likely. In some cases where the rolls and weather combine to frustrate the AI a longer time frame sometimes occurs as it did in your game here.
That being said, on average by the time of Barbarossa things seem to even out and the fact that the AI was able to capture France by August shows that made up some of that delay in Poland as an August capitulation against a human player in France is still a pretty good showing for the AI... at least from all the tests we've run.
Based on this, my guess is that Barbarossa will likely start near the historical timeline as well for the Axis AI.
