Defense evaluation

Questions
1

I was thinking for some time for ways to test or calculate, if specific set of stationary units, is “better” then a similar set.

The final goal is to check what is the best way to spend 5 cores in a turn, if you have multiple options, and without running a full set of (52) simulation for each of the possible combinations.

I have two ideas to simplify this task and will be happy to hear your feedback, or other suggestions:

Custom simulation:
Running one single simulation, but with many units (may be sightly modified) for example:

  • 32 Scramblers with 1000 health, and evaluate the time to score, and possible the damage received.
  • 32 EMPs with 1 health 1 damage, and evaluate the damage done

Information analyzes:
Teoretically, it should be possible to evaluate the defense by just reading it. Ofcourse this is way harder task to define then just running few hundred sims.

Some logical patterns to test:

  • how many holes are there on each main row
  • hoe many paths are there leading to next row
  • Destructor blind spots on the role
  • blind spots not even blocked by stationary units
  • Destructor exposed to frontal attacks

It is supper easy for a human to look at a defense and evaluate it, and even identify some obvious weak spots.
And it is really interesting task create an algorithm that can do something at least similar.

2

I think such an algorithm would be very difficult to make.
Back in season 1, when I didn’t have a simulator, I avaluated mooves by using the inbuilt pathing method and counting the destructors and encryptors in that path.
This worked to some extent, however the major flaw was not considering the health of stationary units. A filter breaking would often completely change the path of the unit and thus the result.
Your idea of simulating 32 scramblers in an interesting idea, but would only really work if you were trying to place filters from those 5 Cores (like you mentioned). There are just to many variables to consider when you are placing Destructors. Unfortunately I currently don’t have any ideas of my own, but it’s probably worth checking out this topic. (Especially since my simulator is quite slow :sweat_smile:)