Is it true that NPRs have some guidance towards heavily investing in sensors by the way? Basically every NPR I found so far had amazing sensors of one or several degrees before mine, even when they were 3 engine tech levels behind me (or even 7 in my Wh40k game).
For the variation, I was brainstorming some Gauss distribution setups, created simply mixing multiple random generated numbers. For example there could be one like "1+ran[1]-ran[1]" as a factor to tech points, giving that on average the factor will be just one, and the grand majority of results just around that. Extremes like 0 and 2 might also be there as a factor, but just really rare. (just like with two dices of course)
Or a more jumpy result sphere by even multiplying random numbers like ran[2]*ran[2], so that on average you get 1 still, but whenever it goes above, it quickly increases towards potentially even 4 here. This could probably be better to accommodate exponentially rising tech costs (...but then again, the spawning population formula, which would give more labs on larger numbers, might already do this trick by itself, I don't know)
To incorporate challenge transformation over game time, the random factors could include some time values by themselves (..if that is even possible), like so: 1+ran[1+GameYear/50]-ran[1], so that while not strictly forbidding lower (=close zero) races to exist, the tendency shifts towards more extreme encounters. (though I would personally only like to be extreme encounters to become just stronger, not more likely)
There could be a ceiling on this positive random factor. Or instead the whole thing could take shape as negative exponent on both (like "+ran[1]^{-GameYear/100}-ran[1]^{-GameYear/100}"), which would cause the gauss curve to sharpen progressively, so that average encounters become more likely over time...
I don't know, there are many possibilities, and I think I haven't yet found what I actually wanted to present. I better wait for if this is even possible in Aurora or if you would even consider this complexity.
If you think it might be worth a look though, I can probably put together some legitimate 'chance over tech point over time' distribution graphs (in 3d
) for a couple alternatives of formulas and approaches. Would take some time to seriously think about though.