After analysing the German artificial intelligence (AI) I have done a lot of coding and scripting on both sides.
As you can see in the video above both sides are functioning more or less as they should and I am moving over to fine-tuning things now. The latest version of the AI comes with different plans for the German and Soviet AI. Every time the game is played the German AI will be assigned a random plan from a list of about 10 different possibilities. The Soviet AI does something similar. The difference being the German AI really picks different schwerpunkten each time, while the Soviet AI picks a different psychology each time.
So for example one game the German AI might focus on Leningrad and the next game it will try to make a deep envelopment of Moscow from the south. While the Soviet AI might be more or less aggressive and be more or less likely to create ‘fortress cities’. The idea here is to avoid replays to be the same experience. Coupled with the dynamic decisions and events of the game itself I am quite sure we are giving the player a very interesting experience here.
Feedback on the beta forums has been good, but for some players the AI on regular setting was to difficult and for others to easy. To combat this and to properly test all those possible different plans the AI will be using: the newest version of Barbarossa has been equiped with an optional metrics sharing button in the preferences menu. I strongly advise all testers (and in the future: players) to switch this button on in order for us to get some core data on game balance. What the metrics does is sending your game progress abstracted and anonymous in less than a dozen statistics to our server every 5 rounds. It should be enough info to make proper statistical analysis on which plans work the best and at what AI level what percentage of the players get beat. Exactly what we need to know to further balance the game.
Best wishes,
Vic
