ORIGINAL: Torplexed
I've noticed the AI in this game and its sisters seems fond of making a lot of attacks at less than optimal odds when it might have been better if it just stood passively on the defensive and conserved its strength. This has the effect of leaving weakened units here and there stuck out on the front line which you can punish further or maybe destroy in your turn.
Unfortunately this is actually purposeful. Generally the AI knows a good attack from a bad attack, but early on there was quite a lot of criticisms directed at it from early reviews that the AI was behaving poorly.
For example, when invading France one of the best things the French AI could do in order to slow down the Axis AI advance was to simply sit tight along defensive positions and force the Axis player to slog through with attacks.
The reaction was that the AI was "doing nothing", because it was not attacking back, and was then a poor AI. So now the AI attacks every once in a while to get the feeling that it is doing something, but not all attacks will necessarily be good ones.
It's a difficult piece of perception to get just right as early on when you first play the game, if the AI understands that the best play is to simply defend, the AI comes across as poor because it is not doing much other than appearing to be just 'sitting there'.
But later on, after a few games, the realization that the better move is to indeed not always attack becomes more apparent, but then the AI looks poor because it is making bad attacks.
Maybe in the future there will be a way for us to model early AI versus later AI once a player has played through X number of games so it will look like the AI is learning on the same curve as a human player.