One of the things that becomes common to hear when fantasy football season is rolling around is a claim that sounds like this: “Player X had more fantasy points over the last half of the season than any other player at his position.”
Implicit in such claims is the idea that the 2nd half of the season could provide some predictive value that we couldn’t get at by looking at the season as a whole. But the difference between whether “2nd Half” claims fit into the category of trivia, or whether they are actually predictive stats, can be determined by testing. Below I’ve set up a simple module that will allow you to test the predictive ability that the 2nd half of the season has on the following year’s fantasy points per game average for wide receivers. Actually you can test the predictive value of any part of the season on the following year’s results. Just move the sliders to filter out some weeks and watch the graph and the r-squared value change.
Because I can anticipate an objection from someone who might say “Yeah, but maybe the 2nd half of the season tells us more about younger players and whether they will break out the following year” I’ve also added the ability to filter the dataset by age of the receiver. Play around with the filters and see how high you can get that r-squared value. If you think that week 17 should be thrown out by analysts when they tally the scoring totals of the previous year, just filter it out and see if it improves the r-squared. You can try any number of combinations of filters.
Getting better at fantasy football means figuring out what you know and separating it from what you don’t know or what is unknowable. Sometimes it helps to go through an exercise like this and figure out for yourself whether something has predictive power or not.