Previously, I looked at the predictive value in the first round of the playoffs of a particular number of games working backwards from the end of the season. I wanted to run the same test in both directions:
As you can see, second-half performance has much more predictive value than first half performance. There's more useful information in a team's last 30 games than there is in their first 70. The reason? Deadline deals. Here you can see the predictive value of 20-game stretches throughout the season:
Again, we're capturing a very high predictive value around the trade deadline. Though deadline deals have a very limited impact on team performance, teams do improve when they deal.
