Alright so given the results so far it's pretty easy to figure out how these feature banners work.

First lets look at gacha rates out of japan:

https://docs.google.com/spreadsheets/d/1vDMqDuPLSAZB914mmb0iVZGx6uSZNhY0XN4BEuSIcMg/edit?usp=sharing

(sourced from: https://docs.google.com/spreadsheets/d/1loKOK9ZyZhMzHaAY_24GzzYC3uFP7VuyTBEfAh7wWsM/edit?usp=sharing)

You can see that there's pretty clear cut % between the rarities. Don't worry about the bottom line for now.

So secondly we know from empirical data from the last couple banners that the pull rate to get any of the feature units is ~30% cumulative.

(WoL et al Results, Locke et al Results, Chiz et al Results)

If you check out the bottom rates for the feature banner you'll see that common units make up 25% of the total distribution, or 5/6 of the feature units pulled. This would mean split we'd get 12.5% each Vivi and Lani. Remaining is 5% for rare units or 1.66% each for Zidane, Amarant or Kuja.

This is extremely close to the reported results.

But you're saying hey what about the WoL(/etc) banners, why were they so good compared to Zidane if they're 4* base?

3-5 and 4-5 feature pool:

*until 5* bases are added

Well lets look at it, if you still have a 30% to pull the feature units but there's no commons you have to split that 30% over the possible options. You can see there's a ~55% chance for a 4* or 5* to be a Base 3* unit (if we remove the Base 5* chances, since there are none). This leaves us with 27.3% chance to get a 3* base, aka Lenna or Tellah at 13.7% each, and a 2.7% chance to get a 4* base, aka WoL.

So ya, using the models we should be able to figure out the pull rates of banners going forward if we know the number of units on the banner and their rarity. I'll try and clean this up a bit.