Most competitive races (aggregate prediction)

(1/1)

Badlands17:
If a race has more than a 95% chance of going one way or the other as determined by the prediction data, it won't be given a rank because most maps that predict a contrary result are maps not intended to be a realistic outcome and thus equivalent to statistical noise (those races are AL, AR, DE, GA, ID, IL, IA, KS, MA, MI, MS-A, MT, NE, NJ, OK, RI, SC, SD, TN, TX, WV, WY-A, and WY-B). MS-B can only be given an estimate because of different people counting MS-A and MS-B.

1. AK - Begich vs. Stevens (52.6%)
2. MN - Coleman vs. Franken (62.5%)
3. OR - Smith vs. Merkley (68.7%)
4. MS-B - Wicker vs. Musgrove (~70%)
5. LA - Landrieu vs. Kennedy (79.3%)
6. NH - Shaheen vs. Sununu (86.7%)*
7. ME - Collins vs. Allen (86.7%)*
8. CO - M. Udall vs. Schaffer (87.7%)
9. NM - T. Udall vs. Pearce (88.8%)
10. NC - Dole vs. Hagan (89.1%)*
11. VA - Warner vs. Gilmore (89.1%)*
12. KY - McConnell vs. Lunsford (91.2%)

* The tiebreaker used in the event of the same number of predictions being made for an outcome is the aggregate confidence, which was strong for VA but only lean for NC. In the case of NH and ME, where the confidence for both was lean, I used the number of predictions rating it a tossup, which was higher for NH.

MarkWarner08:
Interesting. Welcome to the forum, Badlands.

How did you determine those percentages? Is this based on a regression analysis?

Navigation

[0] Message Index