SiegeAI gated behind evolution

[veden]

No description provided.

[veden]

Something like 50% evolution

[garrotte13]

At this moment if SiegeAI and Peaceful AI are turned off, then enemy invests more into MigrationAI, quickly fulfilling ALL good free areas before evolution even reaches 5-10%. With SiegeAI some points are lost due to unlucky sieging (as I wrote in other issue it is unlucky in ~70% cases - selected build place is too close to player's defenses making it some kind of direct raiding and easing the game).
Thus the question is how activity points are gained - is this process downscaled at low evo values and upscaled with higher evolution? If only squad size is evolution-scaled, then replacing SiegeAI by other stages won't help player in the beginning.
I'm talking about game starts with 200%/300% enemy difficulty or more. For example at 150%/200% biters are less active (are points received per number of enemy nests? More nests -> more points -> kicking player harder) .

[veden]

https://github.com/veden/Rampant/blob/master/libs/AIPlanning.lua#L143 to https://github.com/veden/Rampant/blob/master/libs/AIPlanning.lua#L155

Points are based on the difficulty mod setting, active nest count, evolution level, base AI state, and temperament level

[garrotte13]

Turning off both SiegeAI and MigrationAI leads to complete silence -> issue #152 about something not working in Raiding&Aggressive AI phases. Pls ask more details if needed (I didn't try to catch for points spendings during some time period so far, because it's too difficult via console - may be there is a way to read it in log?).
Therefore I decided to turn on both Siege and Migration for 30 real time minutes, then turn off for an hour. Some stupid balance. Maybe decreasing mod difficulty setting can help.
Factorio game turned out to be so easily broken by some mods combinations, when separate mods are so easy. Rampant+Warmonger+Building Time combo kick very hard together. In Rampant ver1 it could be huge raiding waves, but in actual Rampant it's unstoppable nests expansion, which can be naturally interrupted only by turrets creep or multiplayer party.

[garrotte13]

I enabled back SiegeAI and MigrationAI, because it's way too boring without them, and reduced Rampant Map setting "Difficulty Scaling" to 0.5, but must admit it makes a very small effect for MigrationAI, because biters expanded quite a lot in 2 hours. Seems like an abundance of 300% autogenerated enemy base size around provides this effect.
SiegeAI was much less intensive. It is difficult to confirm setting influence, because besides lower Difficulty Scaling I had no pollution eating nests - it is probably much-much more important for sieging nests to appear than SiegeAI phases taking place in biters non-provoking game periods.

All this matter requires a deep testing. For example to compare the same 2 hours time period started from one gamesave point for Difficulty Scaling = 1.0 and then for 0.5.
Another example: to start game with 150-200% enemy bases size setting.

Such tests took place very long ago:
https://mods.factorio.com/mod/Rampant/discussion/605348bb943f58bba9576e35 - most important one with a few different game starts tested
#73
#61
And looks like their results are not actual for Rampant version 3.x.

[garrotte13]

Played some more with 0.5 difficulty scaling (I have evolution somewhere 12..15 %) and probably it makes effect.
What I want to say with all these is that maybe instead of gating SiegeAI and Revenge settling Rampant can scale difficulty some way additionally to increasing factor of evolution? It can give other "features" to mod: more action points gained in later game, even after evolution exceeds 97% and biters "technology" progress stops - more action points -> possibility of some complex aggression behavior algorithms which costs a lot of action points (or requiring big persistent income of action points).

[garrotte13]

for example this scaling can depend on time passed since game start or biters/nests/worms destroyed or whatever.

[veden]

Action points are already gained faster at higher evolution.

Damage could be scaled infinitely for the enemies through technology being researched by the enemy force.

Part of the challenge is determining what those behaviors should be. Open to any recommendations you may have.

Part of this could be done through #135 and #134

[garrotte13]

Started new game with Rampant3.2 after fixes. Got to evolution 14% - looks good so far, with turned on PeacefulAI, SiegeAI, MigratingAI states. Purple cloud is off, but poor-coded AlienBiome was unloaded too - no balance-breaking trees/rocks and tiles pollution absorption issues.
The only change I did is difficulty scale reduced to 0.7 (usual practice till GunTurret MK2 obtained) - I think it is ok, because I play with 200%/200% enemy bases and 60%/60% absorption/pollution-triggered attack costs.
As biters are now limited in expanding it solved problems everywhere (SiegeAI, MigrationAI, revenge settling), while still allowing them to expand in considerable and reasonable amount. Wanna say that it's the best behaviour model for a big period since mod version 0.18.X as I see it now - a new gold standard. And there is no big need in gating SiegeAI after all. Probably action points generation downscaling on early game stages can be useful.

[garrotte13]

As biters are now limited in expanding it solved problems everywhere (SiegeAI, MigrationAI, revenge settling), while still allowing them to expand in considerable and reasonable amount.

Wrong. Unfortunately this is true only in game very beginning :-(