We have "id" column in our datasets and it clearly has some predictive power.

I suppose that a lot of participants (only my hypothesis) are exploiting it.

@Zindi can You confirm that using "id" is allowed?