The best poll in 2012 consistently shows Trump leading or tied

If this poll is right again, the pollsters need to consider abandoning their current likely voter screens

The LA Times/Daybreak poll has consistently shown Trump ahead or tied. Looking at the poll's tracking chart, one can see that Hillary has trailed Trump throughout, except for the period during her convention.

Of course this poll, considering that it is showing good results for Trump, is criticized by the New York Times's Nate Cohn. He argues that the combination of weighting small demographic categories more heavily, and the misfortune of including a young black man who is supporting Trump, results in double-digit support for Trump among black voters, thus throwing off the entire poll in favor of Trump.

The poll's creators would argue that while a smaller demographic group's may have an outsized impact on one day's results, the impact becomes negligible throughout the entire week's sample.

Is it true? Does that multiplication of one young black man's vote render the poll useless this time around? Probably not. The poll's generators point out that the point of the poll is to determine the level of certainty of voters' plans to actually vote and the intensity of the support for a candidate, as opposed to simply asking preference, like most polls.

The poll asks the participants, who are the same people each week, what the percent chance is that they will vote for a candidate. The poll opts for that method rather than asking if the person is definitely going to vote, might vote or will not vote, like most surveys. I would argue that the percentage approach allows for incorporation of more voter variables, and therefore may yield greater accuracy.

Daily Oct. Surprise tracker 

See the full results here

The following unique characteristics reveal why this poll is reaching a different result and may be capturing the hidden Trump vote. This poll uses only Internet polling, so the bias that depresses the Trump vote when talking to a live pollster is eliminated. The poll includes voters who did not vote in previous elections, again, a factor that may be picking up voters for Trump that other polls are missing.

Participants have a week to answer the poll, so those who work at odd hours or are unavailable during normal calling hours can be reached. Last, even those without Internet access can participate in the poll because the pollster provides that randomly selected individual with a computer and Internet access for the duration of the poll. This gives this poll an advantage over other Internet pollsters like SurveyMonkey and CVoter, who each show a lead for Hillary.

The poll seems to have bridged some gaps that traditional pollsters have not handled sufficiently. But in the end, as with all polls, it really comes down to the assumptions pollsters make about turnout. As many have heard, a pollster provided his data to four other pollsters, and four out of five of those pollsters analyzing the same data arrived at a different result, ranging from +4 Clinton to +1 for Trump. Most pollsters in America are using similar assumptions and Trump is down nationally by 3.6 and in the battleground states by 1.5. If those assumptions are off a little bit, Trump may well be ahead even today, as happened in the Brexit polling.

Ultimately we won't know until election day if this poll was accurate in 2016. We do know, however, that this poll was an outlier in 2012 right to the end, but ultimately got the final vote almost exactly right, estimating a 3.8% win for Obama where he actually received a 3.9 % win. It's hard to argue with that result. For those who want to dismiss this poll's results, just remember that many did the same in 2012 and ended up very wrong.

Follow my new account on Twitter for daily updates on the election!