Probabilistic senate prediction competition final results

I almost forgot to post this thread.

In the six months leading up the election I created a thread for each month for people to predict the results of ‘swingable’ senate races.

Here are those threads…

May thread
June thread
July thread
August thread
September thread
October thread

The whole thing was wildly unpopular and I was the only participant in more than month, but there were seven participants total so I may as well post the results.

And the winner is… me.

Name Month MSE
Lance_Turbo October 0.1576
Hari_Seldon May 0.1808
Lance_Turbo May 0.1813
Lance_Turbo June 0.2005
JKellyMap July 0.2092
Lance_Turbo August 0.2161
JKellyMap June 0.2189
Lance_Turbo July 0.2190
Hari_Seldon June 0.2198
Dr.Strangelove May 0.2248
Lance_Turbo September 0.2281
JKellyMap August 0.2299
Calavera May 0.2308
septimus May 0.2479
Oredigger77 May 0.5385

Hari_Seldon, with psychohistory doing the heavy lifting, took 2nd, and remarkably my forecast from way back in May finished 3rd.

Thanks to all who participated.


Thanks for running this! Obviously the Senate election process has been quite the rollercoaster ride.

Not last place. Woo!

That said, I’m not sure this enhances my opinion of prediction markets. Aside from Oredigger77, all of our error values were quite similar. And not monotonically decreasing in time (my stupid algorithm in May narrowly beat yours in September). It looks like one gets about 20% error no matter what method and data you use, within reason.

Senate polling was pretty bad this year in my opinion.

North Carolina and Maine stand out for being way off, but polling also showed that a lot of races would be competitive that didn’t end up being even close (IA, MT, KS, SC, and even KY).