Feedback on AI's new feature: Walling

Ok, so I play against the Extreme AI a lot, and since you developers wanted feedback on this new feature as per your request in the patch notes, here’s mine.

AI walls too slowlys

The AI’s walling is too slow, making it irrelevant. Several factors are in play here, and so far, in my opinion the following ones are the most important to fix/improve:

  1. Seemingly fixed radius of 21-22 tiles. Sometimes is a bit smaller by 1 or 2 tiles. Unsure what gets into play in the decision. Anyway, this radius itself can be a bit big on certain map generations.
  2. Lack of palisades. The AI seems to want to wall using only buildings, or buildings + very few palisades. This is mental. Extremely expensive, so walling efforts are intermitent.

My proposal fixes:

  1. Define 2 phases for walling. The first phase should have a a max radius of 22 tiles around the TC but must go smaller if the use of woodlines and other resources allow. The radius should not be a radius. More like a raidus-per-side? Anyway, the AI does not seem to take woodlines (and probably other resources) into account. See the screenshot below.

Look at the woodline in the red ellipsis. Completely disregarded and the AI simply started walling (starting with the stable) right past to it. The image doesn’t show, but I can tell you that the AI continued this walling with mostly houses.

At this point in time, I have yet to see a game where the AI can complete walling. I guess I won’t see it.

Second phase of walling should account for the acquisition of secondary resources.

More on phase 1

Arabia is a pretty wide map, and all my feedback has been for games in Arabia. So I’ll try to generalize.

  1. The radius for walling should be set per-map, as Marketplace, for instance, has more woodlines and closer woodlines.
  2. Phase 1 should consider the possibility of not fully walling, but walling towards the TC. Programatically speaking, maybe a calculation on total wood needed (with palisades!) or total build time needed should be used to determine if it will be a full wall or wall-to-TC kind of walling.

AI Should Improve House Requirement Prediction

Now I bet this one programatically speaking is hard. As a human, I take a consious decision of making, say, 2 or 3 more houses than what I currently need in order to favor walling. If the AI could be more loose about how it currently decides on house construction, it would benefit wall creation. I cannot think of a good set of conditions in order to drive this decision, so go bug your team lead or senior devs about this one. :smiley:

1 Like

Maybe it should determine the radius based on the average distance to the closest 2 or so woodlines?