Scenario Editor AI making accurate unit testing impossible

:play_button: GAME INFORMATION

  • GAME BUILD #: 177723
  • GAME PLATFORM: Steam
  • OPERATING SYSTEM: Windows 11

:play_button: ISSUE EXPERIENCED

I want to test unit interactions quickly using the scenario editor. However, I can never get accurate results because the AI always interferes via some sort of obnoxious retargeting that causes units to mill around for several seconds instead of actually engaging their initial targets. This persists despite setting the personality (under the player tab) to “none”, using various AI scripts like immortal, bright spark, and illuminati, using a custom “immobile” ai script, and setting various triggers to try and override the computer’s behavior.

I simply want the units to behave for player 2 as they do for player 1 given zero player input; units engage the first target they see and then move to the next after it is defeated. This was the default behavior of the scenario editor for quite some time but at some point this bizarre AI has been implemented that sabotages every fight. I can’t see how anyone would ever want this behavior to occur in any scenario.

:play_button: REPRODUCTION STEPS

  1. On the start screen click Editors → Create Scenario
  2. Click Units. Line up 10 knights for player 1 and 10 knights for player 2 facing each other. Give player 2 a couple hussars somewhere else on the map so it doesn’t use the knights to scout.
  3. Click Menu → Test. Notice how the knights for player 2 spend an absurd amount of time milling around before attacking and thus lose every fight. This is the behavior I want to stop

:play_button: EXPECTED RESULT

The fight should be extremely close given it involves the same units from an even position. Red is instead always steamrolled due to the computer’s egregious mismanagement.

:play_button: IMAGE

I second this recommendation. This bug is, in my opinion, the single most annoying thing in the game.

Note: as a temporary workaround you can have both units be controlled by AI. It’s not perfect due to AI variability, but it can at least give a fair comparison.