If I have a dataset of 1000 instances, and each query is 100 instances long, and I use map@20 for the evaluation metric, is it calculating mean average precision on the top
