Performance of a team late in the season is probably a better indicator of how they will perform in an end of season playoff tournament, ultimately, whether they will win the championship. Let team playoff eligibility reflect this, perhaps using a weighted averaging formula. But what should the weights in the formula be, and on what features?
We propose a feedback approach for calculating playoff eligibility. Playoff spots are filled probabilistically (a lottery) with weights determined by "this year's formula", which attempts to select the teams most likely to win the championship. After the championship is decided, the formula is updated for next year so the features of the team which won have higher probability of being selected on the future.
The probabilistic nature of filling the spots allows for "exploration" of the formula space to prevent it from getting stuck in a local maximum. Occasionally, a low probability team will be selected and go on to win, prompting a significant update to the formula.
(Alternatively, if you want every game of the regular season to count the same, then eliminate all postseason play.)
No comments :
Post a Comment