Hey, @lz01 and @Elio, the main benefit of this competition for Dar Si Hmad is not a perfect water yield predictor but for the organization to learn more about the underlying structure of the problem, and how the various data sources may be used effectively to make _good_ predictions.
As with all of the competitions, winners will be determined based on scores on the private leaderboard. One of the reasons the train/test split is structured with longer time dropouts (rather than random selection) is to make it harder to win by relying on future data. However, the competition is evaluated based on what we can measure, which is the error on the held out dataset.