Please use this thread to discuss the accuracy (or lack thereof) of the various sites that attempt to predict who will win and who will be eliminated. This also includes polls etc. I'd like to consolidate all of that talk into one thread because right now it's in a bunch of different places and some people enjoy using those sites for speculation while others don't.

Thanks for the new thread! I love this stuff.

Moved from spoilers and spec thread:


It's been a while since I've investigated the DialIdol tabulation protocols, but I think they don't use total call numbers in their scoring. Like harvest said, the calls logged reflect only DialIdol users, which is a very skewed sample. But measuring the busy signal gives you a better idea of how the contestants are doing relative to each other. When I looked at it, I was confused by how the first week/second week data fit together, and it almost looked like they went by the first week, where Penn did really well in busy signal percentage. I'm guessing his numbers dropped a lot in week 2 because that's when he started shooting his mouth off about Kristi and everything else.

But even for AI, which is a much easier show for DI to track than DWTS, there's not much accuracy in the early going. Up until this week, I think the only person they ever declared safe was David A. Now that the field has narrowed, they're calling a lot more people safe, and they've been completely accurate. I'm thinking that will happen with DWTS as the sason progresses.

My issue with votenumber1 is that it includes votes from people in 40 different countries. I think that's why Cristian does much better there than in other polls -- since the majority of his fans are outside the US. My guess is that votenumber1 will have a hard time predicting his elimination.

Thanks for that info, Artemis. You know a lot more about this than I do. In the past, I could click on previous weeks to review the date, but I can't figure out how to do it this time, so I assumed they'd combined the first two week's data under "This Week."

If they went on Week 1 data, which I can't remember and don't know how to access anymore, then finally there's a way to explain their "finding" that Penn was safe. Albeit a wrong finding. But looks like DialIdol could always be wrong because they don't have anyway to gauge online voting, do they?

But you've got a point that number1vote also has a flaw - well, probably more than one. The forty countries thing plus if there are contestants with lots of older fans, then those people would be less likely to surf the 'net and vote on the number1vote site, yet they might have learned how to call a lot due to AI, or maybe even do Internet voting, if not searching/surfing, if that makes any sense.

I think Dial Idol also has been factoring in the judges scores just for DWTS but that still doesn't explain why they thought Penn was safe since his judges scores were awful, too.

Dial Idol is scaring me since it has Adam and Steve in 2nd and 3rd with Cristian bringing up the rear. They also have him with the lowest busy signal percentage by far. Though I actually tried to vote for Cristian and couldn't get through. So who knows. Yikes!

Dialidol confuses because Shannon/Steve are bottom 2 in total votes but they are higher up according to busy signals and "safe".

According to the voting site, Kristi,Jason, Cristian and Shannon are top 4. Mario, Marissa, Steve and Marlee are the bottom 4.

ETA: It is now Kristi, Adam (2nd), Shannon, Cristian.

DialIdol isn't very useful, but someone shared a trick during SYTYCD. Ignore busy signals and the DialIdol ranking. That system doesn't work for smaller amounts of votes, particularly with the limits on voting in DwtS. Check actual votes by interval. The people in the lead for the first 15 minutes of calling are the people who are actually in the lead. I can't promise it will work for DwtS because we didn't know about it on SYTCYD for long enough to test against coincidence, but it did reliably pick the top contestants for Top 10 and up.

I am not sure if the first 15 minutes call accurate for DWTS since a lot of people call as soon as the show begins. I think dialidol is just too small of a sample size.

The other thing I don't understand about DialIdol, but granted I don't know much how their software works, is that their rankings come up after the show ends on the east coast. Okay, but at that point the show isn't over in the rest of the country. However, usually the rankings and numbers never change. The numbers that are posted about 10:30 EST are the same numbers that are there the next day--despite the fact that a good part of the country hasn't even voted yet by the time DialIdol gives their rankings.

I wish I knew more about how DialIdol works -- but maybe, msrayrudd, DialIdol is located in the Eastern Time Zone and makes all its calls for people/clients during the show or right after it ends on the East Coast. And then it's done with all it's gonna do ??

The Vote Number 1 site is predicting Marissa, Steve and Marlee (shocking, really) as their bottom 3. I think Marlee has a bigger fan base than people think.

And Marlee and Mario were tied at one point yesterday - and now, there's only 1/10th of a point difference between them. Hopefully it won't be Marlee - she doesn't deserve to be sent home! I started worrying about her last night and gave her some email votes.

I do not think Marlee is going as many seemed to vote for her and Mario last night.

I don't trust DialIdol but it has been accurate these past two weeks when you look at the Raw Numbers page. It correctly predicted Monica and Penn's elimination last week; Steve tonight.

Still not convinced. Just wanted to point that out.

Worth noting on their raw numbers, but their overall DialIdol score page has been way off. They had both Penn and now Steve as safe.

I am nervous now because Shannon/Derek are on the bottom end of the dialidol raw numbers. But then they are higher up on the Vote Number 1 site. But that dialidol score is off. Both sites were right though on the bottom 2 because I do think Marissa was in the bottom 2.

I think that the DialIdol raw numbers had it right with Mario and Steve in the B2 and Shannon as the third one at the bottom.

I wonder though about Shannon because she along with Cristian are a lot more popular on the other site (while Mario and Steve were in the bottom of that site). I guess I just need to vote more for my favorite.

ETA: On the recent vote number 1 poll, Kristi is still in the lead followed by Jason, Cristian, Adam and Shannon. When did Adam get to be so popular? Is it his fans, that he is this season's Jerry Springer or Julianne? I am guessing a combo of all three.

I am curious how the wanting to get Marissa and/or Adam off the show will result in impacting Mario, Cristian and Shannon. I think those three have unknown fan bases but I think they deserve to last longer on the show.

If you look at dialidol's raw numbers, Adam is higher vote totals than Priscilla, Shannon, Steve and Mario. I also would not attribute Adam's popularity just on Julianne but also on Adam's fans who probably want to see Julianne. Though I would discount her fan base voting for her.

Nice to find this site. Thanks, Pem.

Below are the comments posted on the Dialidol site for this last week. I think we need to wait to see if this site can become more accurate in the way their poll rankings are determined, before we take it too seriously.


Below are comments from the week before that.


Here is the direct link to read the commentary page. http://stars.dialidol.com/asp/news/news.asp

Now, the Votenumber1 site says: We are the top site for online voting for IDOL & DANCING! I hear what some of you have said about 40 countries being able to vote, but I'm not sure I would dismiss this out of hand, because of that. They seem to be the more accurate of the two, despite the 40 countries thing.

For those of you interested in data of this nature, here is a site that talks about the demographics of the votenumber1 site and the dialidol site.

http://www.quantcast.com/votenumber1.com

http://www.quantcast.com/search/dialidol.com

I think the jury is still out--but it will be interesting to observe these two sites, along with the TV Guide Poll and the AOL Poll.

TV Guide's latest poll results. They are asking the same question we are posing which is that it is a race for second place. I actually do not think Marissa is going this week and I do not think Shannon benefitted from sympathy votes. I think Shannon and Cristian and just picking up more fans.

I agree with you harvestbasket,in reference to Derek/Shannon. Shannon/Derek fanbase growing because she is doing really good and getting better every week.-fans can see her dedication & determination-She and Derek seem to have such a great rapport and look so adorable together!

FWIW, the most recent dial idol raw numbers. Marlee clearly has fan support, more probably feeling she was overlooked. Fan base wise, I guess she is giving Kristi a challenge. Now, I do not know how to figure Adam getting more votes than Shannon. Busy signal wise, she was on top. I am surprised though at the bottom.

I don't put much stock in DialIdol but the AOL poll was pretty accurate last season. I'm not sure about it's accuracy so far this season. I'm surprised to see Marlee in the bottom three of the AOL poll. I threw her half my votes tonight because I'm not sure how strong her fanbase is holding up. I'm looking forward to the return of the bottom two tomorrow so we can gauge how the audience is voting...

Can we really be using DialIdol as any sort of predictor when they themselves admit it's based on like ten people dialing?

I thought the dialidol raw vote numbers were a bit more accurate but I agree it is a small sample size. I thought the AOL poll was off because I remember Cameron leading all last year.

Here's that vote number 1 site. I am surprised how popular Shannon, Cristian and Adam are on this site. Not sure on its accuracy.

Most people I know voting for Kristi do so online and not so much by phone. Cristian probably won't be eliminated because of his score. I'm guessing Priscilla is in the bottom two but not sure who will join her.


I don't think it's an accurate predictor but the raw numbers have shown who would be eliminated the last two weeks. It will be harder this week since the scores were higher and not so far apart.

I am just not understanding how the Dialidol polling works at all. Between raw scores, busy percentages, the dialidol scores, and the dialidol scores (DWTS), they are downright confusing. Can anyone explain?

As for the TV Guide Poll (which I can never consistently find), it is pretty simple, and easy to understand. On the AOL poll site, I don't seem to be able to find a way to date the results--nor does it give any percentages. So the only site I'm really trusting right now is the Votenumber1 site. (And it's interesting to see Kristi and Mark's lead go from 60% early on to almost 27% now. Wonder what's happening there.)

I do not know how to read into that votenumber1 site since it allows for international votes. This may explain the higher vote totals for Cristian, though it does seem like his fan base is growing. I still think Kristi's lead is substantial because Adam is like second or third on that site.

Deleted for now.

Looks like the Vote Number 1 site was wrong. Tee hee.

I doubt they were wrong about Kristi being on top, but anyone can assume that.

There is just no way to predict DWTS viewers voting patterns.

How were they wrong? I thought they had them listed as an outside chance, along with Marlee and Fabian. Am I wrong? Although, they did predict it would be Marissa.

They didn't list Adam at all.

Okay, so we're back to not having a poll we can trust for the lower numbers, I guess.

ETA: I just checked back to some poll results that I saved from Votenumber1.

This was from earlier today:
Kristi and Mark 27.3%
Shannon and Derek 17.44%
Adam and Julianne 16.1%
Cristian and Cheryl 11.35%
Jason and Edyta 9.34%
Marlee and Fabian 5.21%
Priscilla and Louis 4.85%
Mario and Karina 4.64%
Marissa and Tony 3.77%

So, why, oh why, is Marissa safe--and Adam gone?

That's why I never trust the polls. I find them to be inaccurate most of the time. It's hard to know if the people actually voting in those polls are actually voting online, by phone, text, etc. for their favorite couple to begin with. And the fact that international voters are allowed to vote in that Votenumber1 site makes it even harder for me to figure out.

The AOL poll had Adam as second to the bottom and Priscilla at the bottom. If you add in the scores from the judges, that may have been the most accurate.

Internet polls are not representative of the audience of the show, because online activity tends to skew younger (and male), whereas the average (or was it median?) age of a DWTS viewer is 56 (and female). So, naturally, internet polls will over-represent the Adams and under-represent the Priscillas.

That’s why I like DialIdol. It’s not an internet poll – it is entirely based on phone calls. The problem is that there are so few people using it for DWTS, the sample isn’t large enough to overcome the huge margin of error.

Donamadrina – have you checked out the Wikipedia entry on DialIdol? That gives a pretty good explanation, and it helps to understand why DI doesn’t really work well for DWTS if you look at its history. DI started out as a software you could download to have your computer modem act as a phone and call in your votes for American Idol so that you wouldn’t have to sit there for an hour pressing redial on your phone, especially since the voting could be very frustrating when you’re getting constant busy signals. It makes sense for AI, since there are no vote limits and you’re trying to get in as many calls as you can in an hour. It’s not as useful for DWTS because DWTS allows you only a small number of votes each week, so it doesn’t take a whole hour out of your night to get them in. Also, DWTS gives you the option of voting online.

It was only after a lot of people started using DialIdol that the developer realized that when you took the aggregate data from a large pool of users (how many calls they made versus how many busy signals they received), you could tell which contestants were safe and which were in danger by looking at the percentage of busy signals they got. The key there is “large pool of users.” Right now, I think the only people who use DialIdol for DWTS are the avid American Idol viewers who also vote for DWTS. It’s just not a large pool. It’s no surprise that they called the Gute wrong last week, since they were working off a total of 35 calls – 30 votes and 5 busy signals.

DialIdol is only going to work as a prognosticator for DWTS if DWTS viewers start downloading the software so it can track the patterns. Basically, if you expect it to give you answers, you have to put your money where your mouth is and use it to make your calls. Of course, I myself have been too chicken to do this because I hate downloading any software from the internet, but I also hate to criticize DI if I’m not willing to do my part to make it work. So I’m still trying to decide how badly I want to be able to know who is in danger.

I do not know how much stock to put into any polls, even that AOL poll but Adam was toward the bottom on that poll. I do think the votenumber1 site is probably accurate about Kristi and Priscilla though. But based on the raw vote totals, it looks like it will be Marlee and Kristi at the end.

Today the AOL poll says they correctly predicted Adam leaving, but everytime I checked that poll yesterday Priscilla was the one with the least votes. Did Priscilla get a late bump from the AOL voters?

These polls are obviously not accurate. I never trust any internet polls for anything. That doesn't mean they aren't fun to participate in or to look at, but they should never be seen as rock solid fact.

Thanks, Artemis, for taking the time to provide that rundown of how everything works on the Dialidol. Very informative. It also leaves me wondering how anybody can trust it at all.

Soooooooooo, for us junior statisticians, what do we do? I guess look at all the polls and make our best guesses. That will be easy if the top numbers all match, or are close, and the bottom numbers all match, or are close. But when we see Adam high in one poll, and in the middle in three others, what do you do? Vote him out, of course. :-)

I've noticed that you do have to check the last minute results.

Another non-scientific poll. TV Guide's top 8 poll. I wish I was confident that Marissa really was this unpopular but I know not to get my hopes up.

This came out today.

http://www.tvguide.com/PhotoGallery/Dancin...gs__1172/1.aspx

I was thinking about the accuracy of Vote Numbe r1 and DialIdol. Maybe they're off because it only predicts the audience vote. Does DialIdol add the judge scores? That is 50% of their overall rank.

With DialIdol, the bottom three are Mario, Shannon, and Priscilla. If this is accurate Mario won't be in the bottom two because he is 4,6 points ahead of the other two. I'm guessing Shannon has a bigger fan base so Priscilla is out.

I was keeping track of AOL all week, and in the last few days things changed: Krisit, Mario, Cristian, Shannon (and in that order). Shannon slipped from 3rd to 4th. Interesting, seeing as though she did slip. However, the votenumber1 poll had her 2nd all week. I have not tracked Dialidol, although I looked, because frankly, I still don't understand how it gets all the contradictory numbers.

On the various message boards, people think that it will be Priscilla, Cristian, or Marissa going home tonight. Well, we shall see!

On the AOL poll, I noticed Kristi and Mario are consistently 1 and 2. Cristian, Shannon seem to switch places depending on the week. I think Mario is not going to be in the bottom 2 even if that votenumber 1 site has him with a smaller percentage because of the point difference.

I think Priscilla will be one of the bottom 2 again and not sure on the third. It will show the fan base of Cristian, Shannon and Marlee this week.

It's so interesting to me that Jason and Edyta are so low in the votenumber1 poll, yet on the Results show they are consistently called first every week, along with Kristi and Mark, as safe.

http://www.votenumber1.com/votedancingwthestars.html

donamadrina-The order of being called safe means nothing on the show. Emmitt was consistently called later in the show. Jason and Kristi are the top scorers and seen as frontrunners so it is foolish to call them safe later in the show.

I don't put much stock into the order they are called safe either, I agree that Kristi and Jason are the top scorers so calling them safe later on the show wouldn't make much sense. Also, the votenumber1poll has Cristian as third in that poll as a favorite which doesn't make much sense either since he was in the bottom 2 on Tuesday.

Cristian score was on the bottom end of the score. I figure Kristi is getting probably at least half other popular vote so take a percentage of the remaining, I can see how he might end up in the bottom. But with that votenumber1 site, it has been more accurate but it allows for international votes and I am sure they are voting for Cristian.