This project was developed based on Evan Miller Bayesian A/B Testing Article.
Essentially the project logic is based on OVO(one vs one) algorithm. Because A/B/C/N Bayesian Test calculation required high-cost computation($O(N^{variant count})$)
In order to comparing many Variants, I used a tricky method as followed:
choose the best armbefore calculating the winning probability.- and then
compare between the best and the other arm.
The steps were made faster than Multivariate Bayesian test($O({variant count} * N)$). But for using this project, you should assume each arm is independent.
Just visit the site!
