Wednesday, October 12, 2016

Reports Summary: Simulations Using October 9 ARPIs

In the two preceding posts, I've provided (1) a new season simulation, using teams' October 9 ARPIs as the basis for simulating future results and conference tournaments (rather than the Pre-Season ratings I've used previously) and (2) a new NCAA Tournament bracket simulation, using that new simulation.  These reports give a much more realistic look at what the entire season and the NCAA Tournament automatic qualifiers, at large selections, and seeds may look like.  In each future week's reports, I'll use that week's new ARPI ratings as the basis for the reports.

Here's a table that shows how my use of each week's new ARPI ratings should improve the "end of season likely look" each week:

Ratings Correct: Simulated Win Is Actual Win Difference from W/L is Tie or From Tie is W/L Correct or Within Half Incorrect: Simulated Win Is Actual Loss
Pre-Season ARPI Simulation 60.4% 19.1% 79.5% 20.5%
9/18 ARPI Simulation 66.0% 19.5% 85.5% 14.5%
9/25 ARPI Simulation 68.3% 18.2% 86.5% 13.5%
10/2 ARPI Simulation 69.7% 17.8% 87.4% 12.6%
10/9 ARPI Simulation 70.6% 18.0% 88.6% 11.4%

As the table indicates, when I update weekly the ARPI ratings based on which I simulate future results, the simulated results become progressively more representative of actual results.  This is particularly important in relation to my simulated conference tournaments, which have been looking increasingly unrealistic to me if I use only the Pre-Season Simulation ratings.

No comments:

Post a Comment