Use the regression as the out-of-sample predictive model for 2023 based on factor loadings from 2018-2022 data. For each stock, for each month in 2023, cross-multiply the computed loadings based on data through Dec2022 by the factor return data for Jan-May 2023, plus intercept plus rF, i.e. apply the 30 regression equations to compute out-of-sample predicted returns for the 30 stocks for Jan-May 2023 based on the responses to the factor realizations. Compute alphas for each stock for each month of 2023 by subtracting the predicted from the actual return. Average the alphas. Think, read up on regression equations, don't call me