Predictions are evaluated using **Mean Absolute Error (MAE)** between your submitted probabilities and the true experimental outcomes.

For each gene:

* The true label is:

  * 1 if the gene’s experimental effect is ≥ 1
  * 0 if the effect is < 1
* Your prediction is a probability between 0 and 1.

The evaluation metric is:

$$ MAE = \frac{1}{\text{# of Genes}} \sum_{i=1}^{\text{# of Genes}} |\text{True-Label}_i - \text{Probability}_i| $$

This scoring rule is a **strictly proper scoring rule**, meaning that you minimize expected loss by reporting your true estimated probability. Overconfident or underconfident predictions are penalized equally.

Your grade will be computed using the following formula:

$$\max\left(0, \left(2 - \frac{\text{Your Score}}{\text{Minimum Score}}\right) * 100\right)$$

## Submission File

You must submit a CSV file with exactly two columns:

* `colID`: The gene identifier.
* `probability`: Your predicted probability that the gene improves wellbeing by at least one point.

Each row corresponds to one gene.

Example submission:

    ColID, probability
    1, 0.83
    2, 0.12
    3, 0.56
    etc.