Replies

Below is a model that uses the most recent poll instead of the expert assessment, when there is one. I "expire" the poll if it is greater than 60 days old and revert back to the expert assessment until a new poll is taken. If the trend is moving downward, I suspect it is because a poll expired and the expert asssessment is lagging public opinion.

This week, the polls cover 76% of the 112 races being tracked. 85 polls are being used out of 99 that were captured, for a poll utilization of 86%.

Using the expert ratings above and recent poll data, the results of 32,000 simulated elections are listed in the table below. The definitions of the columns are:

"GOP House Seats P10” – the value where the probability of the actual number being lower is 10%.
”GOP House Seats EV” – the expected value (mean) or probability-weighted average of the simulation.
”GOP House Seats P90” – the value where the probability of the actual number being lower is 90%. Or, it is the value where the probability of the actual number being higher is 10%.
”Probability of 218” – The probability associated with value 218 (the House majority goal).
”GOP Gain” – The difference between the GOP EV and the current number of seats (179), rounded normally.

Week	GOP House Seats P10	GOP House Seats EV	GOP House Seats P90	Probability of 218	GOP Gain
31-Jul-10	207	211.93	217	8.43%	33
07-Aug-10	205	210.66	216	4.51%	32
14-Aug-10	205	210.56	216	4.44%	32
21-Aug-10	208	212.88	218	11.83%	34
28-Aug-10	211	216.78	220	42.58%	38
04-Sep-10	213	217.53	222	50.40%	39
11-Sep-10	214	219.05	224	66.39%	40
18-Sep-10	214	218.54	223	62.30%	40
25-Sep-10	214	218.46	223	61.17%	39
02-Oct-10	217	220.83	225	83.86%	42
09-Oct-10	219	223.48	228	96.15%	44
16-Oct-10	222	225.93	230	99.33%	47
23-Oct-10	224	228.38	233	99.94%	49
30-Oct-10	227	231.29	236	100.00%	52
01-Nov-10	228	233.08	238	100.00%	54

This makes my prediction 49-59 (54).

The maps.

-PJ