Time for a Master AI?

It has been a while since we shared some Barbarossa metrics. With a big influx of new players through the Steam release of Decisive Campaigns: Babarossa we now have a really good picture on the latest v1.03 (scenario version 22) balance. And I am happy to say it looks very good.

Metrics are small data points that can be shared by players (see the in-game prefs tab) with the developers. They help us to check if the balance is as it should be, how the AI is doing and more. But always in the interest of improving the game and sharing with the players how others are playing it.

The victory point collection of the German player versus the Soviet AI. Click here for big version (2MB)difficulty

These metrics show the advance of the German player (in victory points) on the vertical axis and the round number on the horizontal axis. The average player is up to a stiff challenge and even when all is going well it seldom becomes a push-over. If the German player even wins he’ll on average will need to fight in to the winter months. In the graph above the latest version is in purple.

I have been playing for quite some time with the idea to add a variant for a Soviet Master AI. This AI would combine all the best strategies. I have been monitoring quite a few and lots of them had no statistical impact on the AIs performance. Like keeping an army in Odessa or Riga. However i’d like to share 3 strategem metrics that did seem to have had some impact.

The first graph (see below) shows the different levels of confidence the AI has in its forces. It shows that with a bit less confidence (80%-green) the German player can be delayed just that bit more. Which is ironic since that setting causes the AI to give terrain and retreat more often.

The second graph shows the AI willingness to keep regular units on cities to defend them better. It shows that if it only does so when it is not retreating (blue) it has a positive impact especially in the later stages of the game; I guess in the initial stages retreating will also lead to the unit being encircled and destroyed.

The third and last graph shows the AI agressivenes expressed in a modifier of perceived German defensive strength. We can see that it seems to perform better if its less aggressive (1.0-blue).

The result of different Soviet AI strategems versis the German player. Click here for big version (5MB)sovietai

Maybe one by one these differences in German VP attained versus different Soviet strategems seem really minimal, but I think if you add them up they might make a noticeable difference of maybe upto 10VP, which would be quite a lot. I am thus plotting to add a variant that will allow you to play the AI on ‘Master Variant’ setting that will combine all those succesfull strategems together.

But what is your personal experience?

What do you think the Soviet AI could do better in order to become even stronger?

Best wishes,
Vic

This entry was posted in Artificial Intelligence, DC:Barbarossa. Bookmark the permalink.

Leave a Reply