July 30 Poll Report
Good afternoon everybody. There was a debate earlier tonight, and I apologize for not getting the poll report posted (a) yesterday, when I did not get one written, or (b) earlier today, when I had one written but, being in the car all day without access to the internet, had no way to post it. Also, I could not get the debate on the radio, so I don’t know what happened. Anyway, I have a double-update to report on, although the cumulative changes from two days ago are actually not all that large. This is the poll that ran on Sunday, July 28:
Scores
|
Moulton
|
82
|
Trump
|
583
|
Bullock
|
112
|
Gravel
|
83
|
|
|
|
|
|
|
|
|
|
Predicted
|
Moulton
|
10
|
Trump
|
68
|
Bullock
|
13
|
Gravel
|
10
|
Actual
|
Moulton
|
24
|
Trump
|
29
|
Bullock
|
28
|
Gravel
|
19
|
Trump’s support is "hard", inflexible, so when we poll Trump against weaker opponents they score higher than they normally would, and Trump much lower than his Support Score indicates. This has no real impact on the accuracy of the Scores; all that happens is that the weaker candidate temporarily moves up 20 points--about one-fifth of one percent—and then those 20 points are re-distributed to the other weak candidates in future polls. If you compare this to the standard, normal levels of inaccuracy which are reported daily in the Old Fogey Polls, you can see that this is a trivial problem, but people focus on it and bitch about it because the process is new, so people haven’t learned how to read it.
I remember the same thing when I was inventing Baseball Stats every year, which actually I still do although no one pays any attention to them anymore. But anyway, I would invent, say, Runs Created, and people would complain that it didn’t deal with Sacrifice Flies in the right way or some damned thing, so people who wanted to reject the whole idea of Runs Created would focus on the failure to deal with Sacrifice Flies or whatever and continue to evaluate hitters based on the pre-sabermetric statistical triad of home runs, RBI and batting average.
But the process of petty criticism, annoying as it is—annoying as you all are—is a productive process, in that it leads in time to the refinements that eliminate some of the bugs. I sort of think of what I am doing here as being parallel to the early Runs Created method, which didn’t deal with stolen bases or caught stealing or double play balls, or parallel to the Value Approximation Method, which was a forerunner of Win Shares. Saying that I believe in what I am doing is a way of saying that it CAN be improved, not a way of saying that it can’t be improved. Traditional polling has reached a dead end; its limitations are difficult to remove. You can believe it or not, but in my own mind what I am doing is opening up the process.
In the other new poll, the poll that ran on Monday, July 29, Amy Klobuchar picked up a couple of points vs. expectations off of Bernie Sanders:
Scores
|
Steyer
|
62
|
Sanders
|
411
|
Gillibrand
|
284
|
Klobuchar
|
457
|
|
|
|
|
|
|
|
|
|
Predicted
|
Steyer
|
5
|
Sanders
|
34
|
Gillibrand
|
23
|
Klobuchar
|
38
|
Actual
|
Steyer
|
7
|
Sanders
|
33
|
Gillibrand
|
19
|
Klobuchar
|
40
|
The expectations for the poll were 91% accurate, 92% if you save the decimal points. A departure from expectations of that magnitude has no real impact on the Support Scores. A poll has impact on the Support Scores when
(a) The poll does not match expectations based on the Support Scores, or
(b) We poll the leading candidates.
(B) is a larger factor than (a). What really moves the Support Scores is when we poll the leaders. The poll yesterday was a poll of three mid-range candidates and one tail-ender, and it pretty much matched expectations, so. . .doesn’t really move the numbers. It’s like a team taking a one-point lead in the third quarter of an NBA game; you’d rather be a point ahead than a point behind, but it doesn’t really mean anything. Since my last update two days ago:
Steve Bullock is up 22 points as a result of Sunday’s poll,
Andrew Yang is up 21 points as a result of the removal of an old poll, from early June,
Seth Moulton is up 17 points as a result of Sunday’s poll,
Mike Gravel is up 14 points as a result of Sunday’s poll,
Tulsi Gabbard is up 14 points as a result of something or other,
Pete Buttigieg is down 23 points due to the cumulative effects of secondary adjustments from the two polls added and the two removed,
Kirsten Gillibrand is down 26 points as a result of Monday’s poll and the removal from the data of the poll of June 8, in which she had defeated Andrew Yang 45 to 33, and
Donald Trump is down 64 points as a result of Monday’s poll.
These are the updated Support Scores, the last Support Scores before the second round of Democratic debates:
Rank
|
First
|
Last
|
Support
|
1
|
Elizabeth
|
Warren
|
1831
|
2
|
Kamala
|
Harris
|
1018
|
3
|
Pete
|
Buttigieg
|
981
|
4
|
Joe
|
Biden
|
800
|
5
|
Donald
|
Trump
|
519
|
6
|
Amy
|
Klobuchar
|
462
|
7
|
Andrew
|
Yang
|
430
|
8
|
Bernie
|
Sanders
|
410
|
9
|
Julian
|
Castro
|
383
|
10
|
Cory
|
Booker
|
355
|
11
|
Beto
|
O'Rourke
|
343
|
12
|
Tulsi
|
Gabbard
|
286
|
13
|
Kirsten
|
Gillibrand
|
259
|
14
|
Jay
|
Inslee
|
249
|
15
|
John
|
Hickenlooper
|
233
|
16
|
Michael
|
Bennet
|
228
|
17
|
Bill
|
Weld
|
217
|
18
|
Howard
|
Schultz
|
160
|
19
|
Tim
|
Ryan
|
138
|
20
|
Steve
|
Bullock
|
136
|
21
|
John
|
Delaney
|
113
|
22
|
Seth
|
Moulton
|
99
|
23
|
Marianne
|
Williamson
|
99
|
24
|
Mike
|
Gravel
|
95
|
25
|
Bill
|
de Blasio
|
88
|
Thank you all for reading, and for following the process.