OoE: Weighted randomization edge case? Attribute point boost amount seems to get dominated by scale

[EliteSoba]

I noticed this because I was setting the Attribute Point Boost Amount to 1,000 but still getting values in the 10,000s. My understanding of the math is that because the possible range is so huge, the calculated scale ends up dominating the calculations rather heavily

It's hard to describe well but I think like 30% of the time in the first calculation pass you end up with the attribute point boost being >5k above the average and a like 20% chance of getting >10k. You get a negative total about 50% of the time which leads to rescaled rerolls, which further boosts the chance of these high values. It's recursive so its nontrivial to calculate what the final probabilities end up at, but I think like ~45% (30% + 50% chance of reroll * 20% chance of >5k on reroll + ...) chance of being >5k and ~25% chance of being >10k wouldn't be unreasonable.

I think I've noticed this with a few other fields like Heart Repair Amount for OoE as well, but the attribute points stand out the most during runs. Most other fields I guess have small enough ranges and are generally better centered?

It's hard to say what my ideal solution for this would be though. In this particular case, due to the nature of the items and how strong I feel attribute points are in this randomizer, it feels more like they should be like 10x as small. In general, I'm not sure how widespread this "low (or high, too, I guess) average large range" type of issue is so the straightforward solution of "let people specify standard deviations somehow in the difficulty settings on a per-strand basis" doesn't seem worth the effort.