Now with the final 2015 Gubernatorial race in the books, the model actual evened itself out in the end. The model only favored Democrats by 0.9% across the board. Also, with a small sample size, getting 2/3 races called with an average win probability of ~73.3% is spot on. If you have an average win probability of 73% (as I have explained before), you should miss calls in 27/100 races. Therefore, if projecting 50 states +DC next year you have a 73% winner probability on average, you should get (on average) ~13 states wrong. Though the win probability in all reality would be much higher than that due to the number of likely safe states.
This will likely be my final projection for tomorrow's Gubernatorial race. If you go by polls alone, the margin would be 14% in favor of Edwards. The fundamentals of the race pull this one back quite a bit. 30% of the aggregate is composed of an extrapolation of the Jungle Primary results based upon historic trends. This only takes into account a race where things can be considered "typical." This race, however, has been anything but typical. Therefore, the polling aggregate (composes, of course, of 70% of the model) pulls the fundamental projection from 57.43%-42.57% to the margin I am projecting for tomorrow.
Because of the large disparity between fundamentals and polling (not to mention a large standard deviation within the polling margins), the model spits out a 66% win probability for the Democrat, John Bel Edwards.
Any financial help to offset the costs of running this site is always appreciated.