In a few days I start in a datathon in which they have tell us the formula of RRMSE is (1), but I tried to search some implementation, and while searching for it I saw in CrossValidated a person saying that rRMSE is RMSE/mean, then I founded a github code in R (Fgmutils) with the same formula, and the same one in a python package (spotpy).
Then I tried to search through google images and I saw another couple versions of RRMSE, so wasn't sure if it is a mistake of the organizers or is another possible way to compute this metric.
(1): $\sqrt{\frac{1}{N}\sum _{i=1}^N\frac{\left(e_i-s_i\right)^2}{e_i}}$
(2): $\frac{\sqrt{\frac{1}{N}\sum_{i=1}^{N}(e_{i}-s_{i})^2}}{\bar{e}}$
In both formulas $e_i$ is the actual value and $s_i$ the predicted value. I was thinking if it is possible to extract $e_i$ in (1) and that way transform it in (2) but I'm not that good with maths.