The predictions are evaluated on R^2 score.

The R^2 score is defined as 

$$ R^2 = 1 - \frac{SS_{res}}{SS_{tot}}$$

where SS_{res} is the mean squared error of your predictions and SS_{tot} is the variance of the predictions. The scores can be calculated using r2_score in sklearn.metrics

The highest technically possible score (perfect prediction, which is realistically not possible) in this competition is 1.0