Comments on: Can SPM Predict the Playoffs?

By: Eddy

Eddy — Wed, 06 May 2009 01:33:57 +0000

Doing some simple arithmetic shows the Magic have a 53.8% of beating the Celtics in the Semifinals, which makes sense, given that conventional wisdom and evaluation of different metrics supported the notion the series would be a toss-up. Good stuff, Neil.

By: biggles

biggles — Tue, 05 May 2009 17:15:37 +0000

"Hey, that’s not very bad, to be honest."

You almost did too well!

My test of the hypothesis that your distributions were the "true" ones produced a p-value of 0.945 (where 0 = your distributions are crap and 1 = you were incredibly lucky or you cheated). Which, if my code is right, suggests your predictive success may be unsustainable... (Or maybe I just chose a test statistic flukily friendly to you.)

Unrealistic wish: that the Stat Geek Smackdown required you to submit probability distributions -- then it would be a much truer test of predictive skill.

By: Neil Paine

Neil Paine — Tue, 05 May 2009 17:11:53 +0000

Yeah, I don't have the data on-hand, but the underlying formula can be found here:

http://www.whowins.com/formulae/probformulae.html

Intuitively it makes sense -- given that HCA is so important, if an underdog is going to win a series, they're probably going to steal a road game and make the most of their games at home. When it comes down to a Game 6, up 3-2 at home, that's obviously your best chance to take the series; lose there, and you're back on the road for a do-or-die Game 7 with the odds heavily stacked against you.

By: Jason J

Jason J — Tue, 05 May 2009 14:57:40 +0000

Gerrit - That matches the conventional wisdom regarding playoff upsets, but I don't know if that's actually how it works out in the real world.

Top of my head that does tend to be true though. Lakers beat the the #1 seed Blazers in 6 in 1991. Bulls beat the East leading Knicks in 6 in 1993. Cavs beat the #1 seed Pistons in 6 to make the finals a couple of years ago.

By: Gerrit

Gerrit — Mon, 04 May 2009 23:31:30 +0000

Interesting that the team without HCA is more likely to win in six than seven. I'm curious if that matches real world results.

By: Neil Paine

Neil Paine — Mon, 04 May 2009 20:35:14 +0000

For 2009, WPct ~ .49869 + (.0327 * Efficiency Differential).

By: Guy F

Guy F — Mon, 04 May 2009 18:17:07 +0000

How do you convert +/- to win%?

By: Jason J

Jason J — Mon, 04 May 2009 13:18:49 +0000

Those are interesting results. I actually think the method got a little undermined by players trying to fight through injuries and simple matchups. Metrics don't account for Houston being a matchup nightmare for Portland. Metrics don't account for Kenyon Martin single covering David West or for Tyson Chandler's production being impinged by injury. Metrics can't account for Gordon finding 7th gear against Boston or to understand what exactly it means to the Celts to only have 2 legitimate bigs left in their line-up. Considering all the factors that SPM can't anticipate, those results are very solid.