Poblano's new numbers and campaign resources

Those of you who are regulars here already know poblano, DKos's resident numbers guru, who recently revealed himself to be (in real life) sports statistician Nate Silver. You probably also know that his own blog, fivethirtyeight.com, has recently attracted national attention, and deservedly so.

Today, Nate made a major update to his methodology that re-evaluates each state's win probability in light of national trends. This allows the picture in each state, even unpolled states, to change as rapidly as the national polls do. The result is (a) a huge improvement in the win probability of Barack Obama; (b) a national popular vote projection that is much closer with national polls; and (c) a projected win percentage that is much closer to that predicted by "financial" type futures markets.

All well and good. But how should an intelligent candidate (and perhaps you're thinking of an intelligent candidate right now) actually use these numbers?

Earlier today, before Nate's changes, I wrote him an email suggesting how his numbers can be used to allocate a candidate's resources (time and money) to the states where it would do the most good. The same argument applies after the change, so I'll repeat it here.

The idea is this: you don't want to spend time in states where either candidate has a lock, or a near-certainty to win; instead, you want to spend most of your time on states where the win probability is close to 50%, and the closer it is to 50%, the more resources you want to allocate to that state.

Additionally, you want to spend more time and money trying to win a big state (all other things being equal) than a small state.

The resource allocation scheme takes both of these into account, as follows. First, determine a state's "swinginess" by a raw number that indicates how close to 50 the expected win percentage is. You do that with this equation:

Swinginess = 50 - abs(Win% - 50)

... where abs() is the absolute value function, and Win% is the probability that Obama will win the state.

Second, multiply the state's "swinginess" by the number of electoral votes it has. That gives you the resource weight of each state. Express that as a fraction of the total resource weights of all states to determine how much of your time and money to spend in a given state. If you do that, following Nate's new numbers, here is the result:

 
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   

State	EV	538prob	Swinginess	Resource weight	Resource %
FL	27	37.5	75	2025	0.102905
OH	20	61.9	76.2	1524	0.077445
MI	17	57.6	84.8	1441.6	0.073258
VA	13	52.4	95.2	1237.6	0.062891
NC	15	37.4	74.8	1122	0.057017
PA	21	74.4	51.2	1075.2	0.054639
MO	11	48.6	97.2	1069.2	0.054334
IN	11	48.1	96.2	1058.2	0.053775
TX	34	14.8	29.6	1006.4	0.051142
CA	55	94.8	10.4	572	0.029067
GA	15	17	34	510	0.025917
NJ	15	83.6	32.8	492	0.025002
CO	9	73.1	53.8	484.2	0.024606
WI	10	77.7	44.6	446	0.022664
NV	5	59.9	80.2	401	0.020378
AZ	10	16.6	33.2	332	0.016871
NH	4	58.5	83	332	0.016871
SC	8	20.7	41.4	331.2	0.016831
LA	9	17.6	35.2	316.8	0.016099
NY	31	95.1	9.8	303.8	0.015438
NM	5	69.7	60.6	303	0.015398
ND	3	42.2	84.4	253.2	0.012867
WV	5	24.8	49.6	248	0.012603
IA	7	83	34	238	0.012094
MT	3	34	68	204	0.010367
AK	3	33.5	67	201	0.010214
MN	10	90.2	19.6	196	0.00996
SD	3	31.6	63.2	189.6	0.009635
IL	21	95.7	8.6	180.6	0.009178
CT	7	87.7	24.6	172.2	0.008751
KS	6	14.3	28.6	171.6	0.00872
OR	7	89.3	21.4	149.8	0.007612
MD	10	92.8	14.4	144	0.007318
MA	12	94.3	11.4	136.8	0.006952
MS	6	10.8	21.6	129.6	0.006586
TN	11	5.1	10.2	112.2	0.005702
WA	11	95.4	9.2	101.2	0.005143
NE	5	9.6	19.2	96	0.004878
ME	4	90.9	18.2	72.8	0.003699
RI	4	92.8	14.4	57.6	0.002927
AR	6	4.8	9.6	57.6	0.002927
DE	3	91	18	54	0.002744
OK	7	2.5	5	35	0.001779
WY	3	4.4	8.8	26.4	0.001342
ID	4	2.3	4.6	18.4	0.000935
AL	9	0.9	1.8	16.2	0.000823
KY	8	0.8	1.6	12.8	0.00065
HI	4	98.8	2.4	9.6	0.000488
VT	3	98.5	3	9	0.000457
UT	5	0.2	0.4	2	0.000102
DC	3	100	0	0	0

First things first: Obama should spend a lot of time in Florida, on the attack. There are a lot of electoral votes there and the state is closer than many think. An upset there would portend a national landslide. Other "red" states ripe for attack are North Carolina, Missouri, Indiana, and (surprise, surprise) Texas and Georgia.

Second, in addition to playing offense Obama needs to play defense, especially in the Rust Belt states of Ohio, Michigan, and Pennsylvania. Another surprise here is that California needs some attention too, not because there's much chance to lose it, but because it's so big that if it did swing to the GOP, the result would be catastrophic.

Virginia is the big question mark in poblano's projection: very close to 50% probability of winning, and it will clearly get a lot of attention from both candidates. But not too much: if either candidate spends much more than 6% of his time there, it could have been more profitably used elsewhere.