Closed foolishgrunt closed 2 years ago
Actually I just saw #6234, and realized that this is probably caused by the same issue.
I believe 3.19.11-patch1 fixed this too, however, this is the one issue I linked to the misguided PR 6244 that I didn't actually repro and test. If you could be so kind...
Previous version should be released to everyone, can you check that it's solved?
Just updated to 3.19.12, and the issue appears to be fixed - the atoll is the next expansion priority. The terrain yields also appear sane again.
Sorry for the necrobump - I dont know if this is better than creating a new issue or not.
There may be something else here at play with the issue or border expansion. I'm on version 3.19.15 and just noticed a tundra being prioritized over an atoll.
Without the save, this is guesswork. Like, the algorithm ignores resource visibility due to tech...
Sorry about that.
Ouch. chooseNewTileToOwn
is expensive, yet gets called by the top right Table twice in short succession just to see whether expansion has stopped?
OK, Automation.rankTileForExpansion
has a line if (tile.isWater) score += 25
- actual yields give 1 point each. So, Tundra gets score (low=good) 199 (200 for distance, -1 for yield), and Atoll gets 222 (200 dist + 25 for isWater - 3 yield)...
Dunno what the designer thought, so dunno best approach. Make water a +1 penalty?
... that whole function is new from #4510 and hasn't changed much since.
I think we don't need to prioritize water tiles for expansion at all
This looks like a case of "1:1 bug compatible" with Civ V...
I'm not sure its a bug; water tiles cannot be improved and thus making them less likely to be expanded to kind of makes sense.
My mistake, I understood this backwards 🙃
Changing the +25 water penalty to a +1 would give an atoll priority over tundra (which seems commonsense enough), but I believe would tie with grassland/plains - and I don't know the system would resolve that tie.
Something else that the current math causes: prioritizing mountain over both coast and ocean.
Isn't it taking potential improvements into account anyway? Or was that a different tile quality guessing function in the mapgen/startpos space?
Tl.dr -> unify all tile evaluation functions, and while learning what flags/params you would need to conserve current functionality you'd learn enough to foresee other as-yet undiscovered unintended quirks... ?
Automatic border expansion is greatly improved since the time I created issue 3059. However, today I noticed that automatic border expansion was prioritizing a basic coast tile over a reef - I'm not aware of any circumstance under which this should be the case. (It also prioritized an ocean tile and a then-unimproved plains tile over the same reef, which also seems wrong to me.)