Silver had a good night on the Presidential front; on the Senate races, however, not so spectacular. Pollster.com, however, was right on the money on both.
All summer and fall, he rated Montana and North Dakota as distant long shots for the Dems, even when the polling average showed the Dem candidate ahead.
At the end, here was Montana
Lean Republican Montana Dem Rep Margin
Polling average 47.6 46.2 Tester +1.4
Adjusted polling average 47.5 45.6 Tester +1.9
State fundamentals 42.2 50.7 Rehberg +8.5
Now-cast 45.8 47.3 Rehberg +1.5
Projected vote share 48.4 49.9 Rehberg +1.5
Chance of winning 34% 66%
This was profoundly off for the State fundamentals to be +8.5 R in a statewide race (with a Democratic governor and 2 Senators) - or for the fundamentals to control over the polling to such a degree at the end. Flipping the odds from Tester being 2% ahead in the polling to the equivalent of 1.5% behind was just wrong. Pollster.com also "misunderestimated" Testor's lead, but at least they got the direction right.
At the end, here was North Dakota:
North Dakota Dem Rep Margin
Polling average 44.3 48.2 Berg +3.9
Adjusted polling average 45.2 48.1 Berg +2.9
State fundamentals 41.7 50.8 Berg +9.1
Now-cast 43.6 49.3 Berg +5.7
Projected vote share 47.2 52.8 Berg +5.6
Chance of winning 8% 92%
Polls 538
Mason-Dixon 10/28 45.0 47.0 Berg +2.0
Rasmussen 10/18 45.0 50.0 Berg +5.0
Forum/Essman 10/15 40.0 50.0 Berg +10.0
Mason-Dixon 10/5 47.0 47.0 Tie
Rasmussen 7/11 40.0 49.0 Berg +9.0
8% !?
North Dakota had the same problem on State fundamentals, but there was an additional problem. The polling was sparse, and he excluded some high quality partisan polls - especially the final 2 polls by the Mellman Group which has a good record, especially on Dem turnout issues. In contrast, Pollster.com included the partisan polls and got it almost perfect .3% edge for Heidtkamp.
In terms of final results v. final polling position, Pollster.com matched Silver in the Presidential race and surpassed him in the Senate races.
___________
Another point of perhaps more importance to Kossacks going forward - pay attention to Mellman polls, especially with hard to poll Democratic groups like Spanish predominate Hispanics and Native Americans. Remember how Mellman was the only pollster calling Harry Reid's race right two years ago?
Obviously, a campaign is going to release only favorable polls so the Mellman polls. So many of them may be a couple of points too favorable to the Dem. But if a Mellman poll shows a Dem candidate a couple of points ahead, go ahead and assume it's at least very competitive and open your pocketbook or consider volunteering.
Going forward, I would hope that Kossacks such as Steve Singiser will start examining the performance of Democratic pollsters beyond PPP and let us know the ones whose polls we can take to the bank.