http://cland.lsce.ipsl.fr/index.php/workshops/forecasting-crop-yields/33-data-challenge-in-french Some files have an extra index column, you can easily see this in the summary statistics Kaggle provides. year_harvest should go to max 53, but instead it goes to the max=number of rows. So use read_table: df_train = pd.read_table('../input/TrainingDataSet_Maize.csv', index_col=0) df_test_to_submit = pd.read_table('../input/TestDataSet_Maize_blind.csv')