Too close to call

Jessica Sanchez’s fans failed to pull her ahead in Dialidol, as her support petered out at about 3:30 AM eastern time last night:

They did, however, pull her much closer to even with Phil than at the beginning of the night. As such, there is no official Dialidol call, because the difference between the two is within the margin of error. Also note that I believe (not sure) that Dialidol should have left the poll open for another hour. However, at the rate measured at 4:00 AM, that would not have changed the fact that Phil would be on top.

Some people think that Phil being ahead on Dialidol is dispositive, but I’m not so sure. We saw just last week that Jessica’s votes were at least slightly undercounted compared with Joshua’s. The question is how big this effect is. We know that she’s being undercounted by at least 0.3 points. But could she be undercounted by 1.4? We don’t really know, but it could be.

Much has been made of Dialidol’s accuracy, but many of their calls are outside the margin of error. 2 out of 7 were too close to call. Yes, Dialidol ranked the winner #1 in both of those cases, but they had a 1 in 4 chance of that happening by random guess.

Meanwhile, Intrade’s estimate for Phil winning dropped from 90% to 80%, which is where I thought it ought to be. Also, you can read Votefair’s official explanation for why it’s OK that Jessica has led every poll for the year. I’m not sure I find it entirely convincing. He says that he is controlling for votes from the Philippines. But what about votes from Asian Americans? It’s not like he can IP filter those. Votefair is a self-selecting poll, and there are lots of potential problems with that.

I still lean toward Phil being the winner, but there isn’t enough information right now.

Bookmark the permalink.
  • Gmanski

    Dialidol’s results tally stopped changing (as was the case last week) at 4:00 am ET when the California telephone voting window closed.

    • Reuben

      That’s right. I do not know why the people running the site did that.

  • mcritic

    Don’t know why the author holds on to Dialidol’s vaunted accuracy, when the audience for DIalidol votes are self-selected (you have to download the app to be counted). Because of their different styles, voters tend to stay in their own camps meaning that those who vote for JS will stay resolute and vote until the wee hours of the morning. So those of us who are not part of the small sample will be the undoing of the author’s prediction.

    • Jessica

      Dialidol primarily measures busy signals, which takes into account those who have not installed the software. Go to their site for a more in-depth explanation of how Dialidol works.

    • Reuben

      I agree, by the way. I’m backing off of Dialidol.

  • Michael

    Now that we know the vote population was 132 million, doesnt the massive amount of votes increase the potential error rate due the the small smaple size on DialIdol? Cant remember my stat class in college too well. Getting old. ha!

    • Jessica

      I’m pretty rusty too, but I think given the nature of Dialidol’s primary measurement (busy signals when dialing contestant phone numbers) the total number of votes shouldn’t affect its precision — except insofar as it may be the case that a large proportion of those votes are from texting or Facebook voting, which Dialidol would have a hard time measuring. In fact the rise of Facebook-authenticated online voting may account for the dip in Dialidol’s accuracy over the past couple of seasons, since only AT&T customers are able to text in their votes (but that’s just me speculating!).

      So the answer, as it so often is, is “it depends”!

      • Michael

        I think you are right, you would need to actually compare to total number of dialing busy signals only to the total number of votes from telephone calls. The larger the population, the larget standard deviation would be in the form of raw numbers, thus it would make is a larger variance.

        However, we do not know the break-down of the vote totals and are not able to determine the number of dialed votes verses other votes.

        I also checked and it appears that Idol allows for International voting, though I am not certain through what forms. I would assume mostly Skype and Facebook. If that is the case, I would not be surprised if this is the first American Idol where the “swing” votes came from non-American countries such as the Phllipines.

        I’m just sayin’


        • Jessica

          The official rules appear to be that voting is supposed to be limited to the U.S., Puerto Rico, and the U.S. Virgin Islands (apparently not Guam or other far-flung Pacific islands) per the FAQs. Technically it’s fraud to vote from other countries, so they may filter for that. Of course they’ll never tell us if they do, or any interesting details about voting that we might want to know, so this will be yet another Idol conspiracy theory, no doubt.

  • Pinoy

    How will they know voting is from another country when magicjack uses a local US number?

    • Jessica

      I have no idea; if fraud were easy to detect and defend against there wouldn’t be very much of it.

  • sal

    I voted for jessica for 4hrs from my 8 phones. I live in hawaii and noticed dialidol chart was not updated 8pm hawaiian time. I can hardly get any votes in for 3 hrs until the fourth hr which I probably casted over a thousand votes. So no, dialidol is not up to date.