Right track, wrong track average

Net stands at −36.3 in the On Point average of 190 public polls: a robust LOWESS fit through every poll, weighted by a 15.8 day recency half life, sample size, voter screen and pollster grade, with three bisquare passes that weigh down outliers.

190 polls20 pollstersUpdated Sep 28, 202630 days −6.9Public JSON

Right track29.1

Net−36.3

Wrong track65.4

Newest 30 polls

The average over time

Size dots by weight

The line is the On Point average: a robust LOWESS fit through the polls, weighted by recency, sample, voter screen and pollster grade, refit three times so outliers count for lessDot size is each poll's weight todayHover or use the arrow keys to read any day

What moves the average today

Share of today's weight

Economist/YouGovSep 21 · 1,401 RV · C19%−38.0

Economist/YouGovSep 28 · 1,429 RV · C18%−39.0

Wall Street JournalSep 21 · 1,500 RV · Gold16%−41.0

Economist/YouGovSep 14 · 1,461 RV · C15%−35.0

Economist/YouGovSep 8 · 1,469 RV · C8.8%−36.0

Harvard-HarrisAug 30 · 2,100 RV · NR6.9%−15.0

Harvard-HarrisSep 28 · 2,200 RV · NR6.6%−17.0

Economist/YouGovAug 17 · 1,450 RV · C5.9%−32.0

Every other poll together: 3.4%

Each poll's weight is its sample size, voter screen, pollster grade and age, with a 15.8 day half life, and those weights are the kernel of a robust LOWESS fit: a local line read at the day itself, with three bisquare passes that weigh down any poll far from the rest. No single poll can carry more than 15 percent. Gold Standard pollsters, On Point Polling, Big Data Poll, The Honest Poll, Emerson, Cygnal, Quantus Insights, The New York Times and The Wall Street Journal, carry the top grade. Rasmussen Reports and ActiVote are not counted. Trafalgar and InsiderAdvantage are marked Lean R and Impact Research Lean D. The tables show every poll as published; the average counts a lean firm's poll moved the other way by its measured lean against other polls of the same races, and the chart shows that move as a hollow dot for the poll as published joined to the filled dot the average counts: Trafalgar 3.8 points right over 35 polls, InsiderAdvantage 4.6 points right over 42 polls, Impact Research 0.0 points left over 0 polls.

All polls

190 of 190


On Point average	Sep 28, 2026	190 polls	29.1	65.4	−36.3	100%
Economist/YouGovC	Sep 28, 26	1,429RV	27	66	−39.0	18.4%
Harvard-HarrisNR	Sep 28, 26	2,200RV	35	52	−17.0	6.6%
Economist/YouGovC	Sep 21, 26	1,401RV	27	65	−38.0	18.5%
Wall Street JournalGold	Sep 21, 26	1,500RV	27	68	−41.0	16.4%
Economist/YouGovC	Sep 14, 26	1,461RV	29	64	−35.0	15.1%
Economist/YouGovC	Sep 8, 26	1,469RV	29	65	−36.0	8.8%
Economist/YouGovC	Aug 31, 26	n/aRV	27	62	−35.0	0.7%
Harvard-HarrisNR	Aug 30, 26	2,100RV	37	52	−15.0	6.9%
Economist/YouGovC	Aug 17, 26	1,450RV	31	63	−32.0	5.9%
Economist/YouGovC	Aug 3, 26	1,473RV	32	63	−31.0	2.4%
Harvard-HarrisNR	Jul 12, 26	1,776RV	35	52	−17.0	0.2%
Big Data PollGold	May 27, 26	2,784LV	31	60	−29.0	0.1%

Current average

All polls

Right track

29.1%

Wrong track

65.4%

Net

−36.3

The numbers

Polls in the average190

Pollsters20

Newest pollSep 28, 26

Change, 30 days−6.9

Methodology

The On Point average is a robust LOWESS fit through every public poll, rebuilt for every day: each poll is weighted by recency with a 15.8 day half life, sample size, voter screen and pollster grade, a local line is fit through time and read at the day itself, and three bisquare robustness passes weigh down polls far from the rest. Rasmussen Reports and ActiVote are left out, and the measured lean of Trafalgar and InsiderAdvantage, to the right, and Impact Research, to the left, is taken off their polls in the average while the tables show them as published. On Point Polling polls are fielded by On Point Polling. How the average works.

Voter sample

−36.3
190 polls in the average