Hi @Koleshjr and @Brainaic, I will refer you to the following blog post for you to better understand: https://zindi.africa/learn/introducing-multi-metric-evaluation-or-one-metric-to-rule-them-all.

The MAE value from the benchmark is used for normalised.

@Joel Apologies for bugging you, but their is still a bit of confusion.

From the link you shared, it states:

This means you are still aiming for a normalised leaderboard score close to 1 for regression tasks, while continuing to minimise your local validation error in the usual way.

Does this mean that the MAE scores in the LB that are close to 1 are performing better? What about MAE scores that are above one.

Also for transparency please share the xmin and xmax values used to normalise MAE as there is no starter nb for this comp