I have arranged the dataset for your ease. Now at the end you can find 22 columns (Players_1, Players_2, .......... Players_22). These columns are to be predicted one by one using chain classifier. For multilabel classification we have to implement various Models and at the end we will have a table having name of model and its accuracy score.
the last 22 column of player names are to be predicted and the previous column are features. so we will predict first column from these features, then we will predict second player column from these features and first predicted player column, then we will predict third player from these features and 1,2 predicted player columns and so on will predict 22 column for this multi lable classification.