Don't overextend attack

[CTimmerman]

The first step to attacking multiple opponents (#8996) is not overextending on one. See this patch saving player one:

[CTimmerman]

Now checking isGoaded and entitiesMustAttack (which should be enough on its own imo):

[tool4ever]

The main problem remains: you're trying to modify a pretty extensive calculation result with way simpler heuristics.

This will certainly work to the AIs advantage in some constructed cases but at the same time just introduce at least as much new potential for misplays, e.g. when your damage_guess ignores modifiers (having Infect would already be enough) remaining attackers might end up suiciding for no real benefit.

For a smarter selection you could do your threshold approach while inside the relevant parts of doAssault.
Or/And the other opponents should be considered too.
I'm actually not sure if just not assaulting when your teamsize is smaller wouldn't be a better (certainly safer) default anyway?

On top you keep simply resolving comments and only force-pushing which doesn't help at all in understanding what changes you make to address any concerns...

[CTimmerman]

The issue this patch fixes is about using too many attackers. It appears that the current doAssault adds all of them.

As commented, the current patch checks for needing to attack (using isGoaded and static check, but not the singular rememberedtoattack check in the existing code; i'd make shouldAttack a single function checking all potential causes) and leaves 2 extra attackers beyond the 20 life overkill (more than enough for shared infect, but an edge case would be a bunch of mites sorted last. To avoid edge case code i'd just Monte Carlo or neural net the options.).

A smaller team sounds unable to swing for the win so should not return true in doAssault, unless there's a big enough trampler in there, which should make the rest not needed for the attack. I'll see about modifying doAssault and leave the messy history.