It would be cool to have a section on how you/the team tune and evaluate the AI and economy, when that post rolls around. Flicking through the economic_plan and job_weight files reveals some odd numbers, ones that must come from observations/feedback, somehow.As Enfield said, we're looking at AI. When we can give you solid details on it, we will![]()
Naturally, I imagine you've got dedicated tools, but screwing with Forge/CG balancing mod ideas necessitated me creating a set of "income map filters" to track (in this case) CG output of all AIs net income in observer mode: (White < 0 CGs a month, Red +0-10 CGs p/month, orange +11-20 a month, green is 21+ p/month),
Interestingly, I find that every single time, AI score more-or-less maps to CG output on ensign difficulty, the top half all have positive CGs, whilst the bottom half have none / < 5 and the bottom quartile are lacking in food (death spiral) with the last lacking in energy too (double death spiral).
This relationship breaks down as the number of free jobs increases, though shrinking planet sizes caps this issue, and I'm able to get ... not quite stellar, but servicable AIs in 2300 as a result by reducing the min moon size to 6 and min planet size to 9 whilst reducing the number of districts yielded by planet size from 1 to 0.80 ~ 0.70 (so 20%-30% less districts on average, and making smaller worlds statistically more likely, which caps off the max number of free jobs the AI has to melt-down over when allocating pops to jobs, which seems to be the root cause of economic death.
- Likewise, reducing required pop growth to 1 but only letting them grow when free jobs exist via game rules [i.e. always full jobs skips the pop-job-scarcity issue], leads to surprisingly economically strong AIs - again suggesting that Jobs allocation is overlooking some fundamental factor, which likely feeds back into income deficits and the AI overbuilding/underbuilding certain districts, exacerbating its situation).
Last edited:
- 10
- 1