I am digging through the data and I got to the part where I was going to use the tournament dataset to test my results. Well, all of the data_type == ‘test’ records don’t have a target value. What is the reason for this subset? The ‘live’ set is what we turn in, right?
Sorry if this is dumb and I am missing something.
You “turn in” predictions for everything in the “tournament file” – validation, test, and live. Live is what you are scored on (live hasn’t happened yet). test is data for which Numerai has the targets but you don’t – for internal backtesting.
Cool. Thanks for the info. That is going to be a large list.
Well, at least you’re in good company, I suffered from the same confusion just a few days ago, but @wigglemuse cleared me up.