My thinking was could you set up some tests (say 5 days , 3 episodes per day) that was manageable to confirm your process and calculations. Once you get it working for a few days and a few episodes, it should work for a larger number. It's just that testing with 270+ episodes over 191 days seems extreme for testing.
You could separately calculate the averages or whatever you need in advance( to predict what the true number is), and then run your "processor" against the test data and see (the result). Was the result = the expected? If no, can you account for the difference? Perhaps you need to adjust your "processor/algorithm", then run it again.