When you submit solutions, you will create a CSV with two columns: ID and Predicted. Since your output will be distance matrices, we will create n^2 points for every n x n distance matrix. In other words, for every distance matrix, make new points where the ID is {casp_id}_{i}_{j} for points i, j = 0,...,n - 1, and the Predicted is the (i, j) element of the predicted distance matrix for that protein.
The evaluation metric for this competition is the root mean squared error across these distance matrix values.