On Thu, 6 Jun 2013, Mark wrote:
> The NCAA is beginning to discuss the shortcomings PWR. THe folks on this
> list are generally agreed that KRACH is a better statistical model than PWR,
> but does KRACH address the issue of home vs away games?
There is a simple extension to Bradley-Terry (which is what KRACH is)
which does this. You assign a multiplier to model how much better a
team is at home, and then you fit that multiplier by requiring the
actual number of wins (or points in this case) for home teams to be
the same as the number predicted by the model. We (I guess Ken Butler
and I, although Adam Wodon was the one who said we needed a catchy
name) dubbed it KASA ("KRACH Adjusted for Site Advantage"), and the
machinery to calculate it exists on CHN, although it's not heavily
promoted yet.
John T. Whelan
Cornell '91