DialIdol reliability

[GrannyMae]

imho, they sucked this year. there is a reason why their site boasts the 87% from last year, but fails to mention this year.

i think the producers figured out a way to fool them, or it was so close this year that their methodology has been exposed as totally flawed.

[domer2]

Been saying it for a while...there is a basic error in the program this year. For users that do not have carriers with busy signals (instead a taped message saying "all circuits are busy"), those are counted as votes instead of busy signals.

Additionally, I doubt the program is very sophisticated. One could probably just trick it into thinking its dialing the correct numbers when in fact its dialing your home phone and therefore gets a busy signal.

Combine the two methods and you could manipulate it any way you wanted to.

[LotsOfOuts69]

WTH is the problem? Dial Idol never once clamied to get all the contestants in the exact orders. All it every claimed to do was show who has a chance of going home and who is completely safe.

It was a statistical dead heat this week. Dial Idol has been 100% dead on all year for all that it claimed to do. For those who expect it to list exact orders you are using it wrong. WTF do you expect?

If you complaining that its not very useful in helping you bet or get an edge somewhere, it never was, nor should you be using it in that way.

[TheVegan]

[ QUOTE ]
WTH is the problem? Dial Idol never once clamied to get all the contestants in the exact orders. All it every claimed to do was show who has a chance of going home and who is completely safe.

It was a statistical dead heat this week. Dial Idol has been 100% dead on all year for all that it claimed to do. For those who expect it to list exact orders you are using it wrong. WTF do you expect?

If you complaining that its not very useful in helping you bet or get an edge somewhere, it never was, nor should you be using it in that way.

[/ QUOTE ]

Actually dial idol has helped me a lot in betting on who is getting eliminated. Hell because they showed how close the vote was last night I took the 7-1 on Melinda getting eliminated.

[domer2]

[ QUOTE ]
WTH is the problem? Dial Idol never once clamied to get all the contestants in the exact orders. All it every claimed to do was show who has a chance of going home and who is completely safe.

It was a statistical dead heat this week. Dial Idol has been 100% dead on all year for all that it claimed to do. For those who expect it to list exact orders you are using it wrong. WTF do you expect?

If you complaining that its not very useful in helping you bet or get an edge somewhere, it never was, nor should you be using it in that way.

[/ QUOTE ]

Last year you could set your watch to it, even with MOE. 180 degrees different this year.

[NoChance]

I mentioned this in another thread. Last year the average MOE was 1.5 - 1.6. This year they changed it to 2.5 - 2.6. I am guessing they did this to create bigger ranges and be able to claim they are always correct with their predictions. Look back and you often see 6 or 7 people in danger of going home this year. That was due to tinkering with the MOE.

[StellarWind]

[ QUOTE ]
i think the producers figured out a way to fool them, or it was so close this year that their methodology has been exposed as totally flawed.

[/ QUOTE ]
I wasn't around last year so I don't know how it worked in its glory era. I can tell you some of what's wrong right now:

1. Very few reporting users in the Western time zones. This is very strange but it's clearly true.

2. Rise of text messaging. There has been a big marketing push for SMS in the U.S. wireless industry this year. Entire demographic blocks prefer texting to phone calls for many purposes and these voters are very different than your "average" AI viewer.

3. Rise of nontraditional telephone carriers such as cable companies, broadband fixed wireless providers, and Internet telephone companies such as Vonage. Millions of household phones are now outside the traditional Local Exchange Carriers. How much do you want to bet that many of these new telephone systems don't interface properly with the DI feature that counts busy signals? DialIdol users are probably early adopters of technology who gravitate toward these new services.

4. Counting votes after 11:00 when AI has closed the lines. There is a steady stream of votes from 11:00 - 12:00 EDT when AFAIK no one is voting anyplace. Probably some Eastern modems continue to "vote" far into the night.

5. AI voting is probably managed by an Intelligent Networking application. Votes can be counted locally in each region instead of being hauled long-distance to a central server at great expense. Busy rates may be high in some areas and low in others based on the amount of network capacity that has been provisioned for AI in that area. Certainly my experience has been that busy rates are a mere fraction of what DI claims. After ten minutes my busy rate is virtually zero.

Why does this matter? The problem is DI users aren't a smooth geographic sample. Suppose the Florida region suffers from low capacity and high busy rates. Maybe it's hard to get through to vote for anyone. Now enter the luck of small samples: perhaps Melinda has six voter-users in Florida and the others have none. That's a major skew because these six modems will report over-and-over that Melinda is busy and distort the national sample. Never mind that she may actually be running last in the state. Regardless of how hard it is to vote for Jordin in Florida it won't matter because no one happens to be reporting it.

I know there have been complaints in past years about excessive busy signals. I think the big change this year may be that AI and AT&T did something about it. Possibly they used to use a centralized system and now they don't. Maybe they had a distributed system but they upgraded it in some fashion. Whatever they did I'm sure it was about reducing busy signals and saving money and not just an attack on DI. Hundreds of millions of toll-free calls per season is a huge expense in long-distance charges and call processing equipment. Replacing a centralized system with an IN application running in AT&T switches across the country might pay for itself in about three weeks.