DialIdol had some problems with the first 20-30 minutes of last nights vote for the girls only. That data was not collected (the votes were still submitted to SYTYCD though). Also despite what the predictions page implies - there is NO definitive prediction. Comfort is NOT definitively predicted to be voted off. The reason it shows like that is because she never got a signal busy signal and her score is thus - zero. Margin of error thus calculates out to zero even though that clearly is not right. I'll fix this - maybe. :) The weather is just too darn nice in the summer!

The forums were hit by a spammer this weekend. They used some funky method to pretend to post as me - as best I can tell they were NOT me. Unfortunately the way they were doing it involved new user registration so to combat the spam I turned new user registration off. It will remain that way until I find some time to fix the problem.

DialIdol definitively predicted that David Cook would win American Idol - and he did - congratulations David Cook!

For the season DialIdol finishes with 30 correct predictions (noted by the smiley) and one wrong prediction (Kristy Lee Cook - noted by the frown) - a stunning 97% accuracy rate!

This has been DialIdol's most successful season and I personally enjoyed this season as much as I have since I've been doing DialIdol. I'd like to thank all of you who visit my little part of this tangled web we weave -without you there would be no DialIdol.com!

Let's be clear - DialIdol is definitively predicting that David Cook will win American Idol. No ifs, ands or buts about it. No crazy margin of error claims or anything. The DialIdol scores are significantly different enough to wash all that out the window.

This comes as a bit of a surprise to me given that David Archuleta "knocked out" David Cook during the performance show so before you decide to skip the results show tonight you might want to consider that I myself will be watching with the usual uncertain anticipation that has become habit for me!

DialIdol.com got slammed pretty hard last night by lots of fans trying to see DialIdol's predictions. Page load time was in the neighborhood of 1-2 minutes from 9pm - 1am. Sorry about this. Sometimes it seems that it is easier to predict who will win American Idol than it is to predict how much traffic DialIdol will get. :)

Week w/3 (to 2) Idols Comments

Syesha Mercado was absolutely predicted to be voted off and she absolutely was so that counts as being accurate. That leaves us with the matchup of David Vs. David which is as interesting of a final to me as we have had since Carrie Vs. Bo from season four. All of this should make for a good show next Tuesday - and Wednesday!

Week w/4 (to 3) Idols Comments

Jason Castro was absolutely predicted to be voted off and he absolutely was so that counts as being accurate.

Some thoughts about this weeks predictions.

• First of all lets make one thing clear - Jason Castro is definitively predicted to be voted off tonight. No ifs ands or buts about it.

• After last nights performance show it is a wonder to me how Jason Castro got any busy signals.

  • For starters Jason opened the show by yawning on national TV. It was quick but I definitely saw it.

  • His performances were the worst I have ever seen at this stage of the game.

• I didn't dive deep into analysis of the data this week because quite frankly it seems to make sense. Nonetheless I did notice that Jason did poorly during East coast voting and then improved during the West coast voting. It seems to me that if it were up to the West coast Jason would be sticking around another week.

• Did you notice at the beginning of the show (right about when Jason yawned) Ryan noted that three of the four remaining contestants had finished number one in voting at least once during the season? I'm sure we can all agree that the Davids finished number one more than once or twice which leaves us to wonder whether Syesha or Jason was the other to finish first? If you ask me this just lends further proof that perhaps DialIdol was dead on accurate last week with Syesha getting the most votes! (You may think she was in the bottom two but in reality they never, ever said so!) Sure there were a few weeks that Jason did well on DialIdol - but I have a hard time believing he did better than David Archuleta during those weeks when David Archuleta's strength was more dominant than it is now.

Week w/5 (to 4) Idols Comments

Brooke White was predicted to be in danger of being voted off and she was so that counts as being accurate.

Ryan Seacrest never stated that Syesha Mercado and Brooke constituted the bottom two therefore the accuracy of Syesha's prediction cannot be measured (by me at least).

Some thoughts about this weeks predictions.

• I was surprised to see Syesha in the #1 slot. In analyzing the East/Central time zone data I found that her strength seemed balanced (not overly influenced by a few dialers) but it seemed David Cook had slightly more balanced strength. When you consider his West coast strength last night I do believe David Cook probably finished ahead of Syesha but none of the data suggests that either are in danger.

• The East coast data also suggests Jason Castro did worse than the predictions imply. His West coast showing gave him the bump he needed to move out of last place but I have a feeling both Jason and Brooke are even closer than the final predictions imply.

• David Archuleta - in danger? Huh? All of the data suggests it's that close but I wouldn't expect him to be going home. Considering that Syesha performed last, did well and the fact that Simon may have scared people into voting for her by saying she might be in danger right before the phone lines opened - I think that David being in danger was the biggest surprise but I have a feeling you all might think differently. :)

• I did not get a chance to dive any deeper into the West coast data than looking at DIHardGraphs which indicate David Cook had more strength than Syesha and Jason Castro had more strength than Brooke White and all of them had (slightly) more strength than David Archuleta.

• The phone lines acted weird at least once last night at about 9:23pm (Eastern) - everyone spiked to higher busy percentages than the pattern had been indicating.

Week w/6 (to 5) Idols Comments

As you know, there wasn't much of a prediction this week with David Cook being the only definitive prediction. Nevertheless, Carly, who sat at the bottom of DialIdol's predictions, was predicted to be in danger of being voted off - and she was, so DialIdol is back on track this week.

Or is it?

Again this week American Idol only decided to commit to a bottom two(/three) – never disclosing who actually got the fewest number of votes and instead stated “Carly you are leaving us tonight.” I personally think the 30 million viewers who watch every week deserve less ambiguity – but that's just me.
 

• I looked over the DialIdol call data from the last two weeks. Here are my observations.
  • During the week with eight Idols (when Michael Johns got voted off) I found at least three different dialers who fit this criteria.
    • They were dialing for Michael Johns and someone else. They busy percent for everyone else was was low but the busy percent for Michael Johns was 100%.
      • This probably was not accurate data and helped to inflate Michael Johns DialIdol score. This may account for the disparity between the results and the predictions.
  • During the week with seven Idols (Kristy Lee Cook got voted off) I found nothing that stuck out as being odd. What I did find was data that more strongly suggested that of DialIdol's bottom four, Brooke had the lowest prediction and Kristy had the highest.
    • This data, combined with the fact that Ryan Seacrest never specifically stated that Kristy Lee Cook got the fewest number of votes has me wondering if she really did?
      • Ryan Seacrest said this “After the nationwide vote the person staying with us is Brooke. Kristy Lee Cook leaves us tonight on American Idol.”
        • So did she get the fewest votes? Or did the producers decide that keeping the real lowest vote getter made better financial sense? Or did Kristy Lee herself simply decide to leave?
          • I am not trying to imply that Kristy Lee didn't get the fewest number of votes. What I am trying to do is point out that American Idol's vague results were even more vague this week. I for one would like someone from American Idol to go on record as stating that Kristy Lee Cook got the fewest votes last week and for this ambiguity to never happen again.
• Due to heavy traffic immediately after American Idol airs, I ordered another server. I actually ordered it during the vote last week and had it online by 11:30pm (Eastern) that night. I can't tell you how happy I am with ThePlanet.com - if you have a need for dedicated servers - keep them in mind - they rock.
• Go Cavs!

Week w/7 (to 6) Idols Comments

Kristy Lee Cook was absolutely predicted to be safe – she was sent home which was absolutely wrong.

I did find Ryan's “choice” of words to be somewhat interesting...

“After the nationwide vote the person staying with us is Brooke. Kristy Lee Cook leaves us tonight on American Idol.”

Week w/8 (to 7) Idols Comments

Michael Johns was predicted to be in danger of being voted off and he was so that counts as being accurate.

Carly Smithson and Syesha Mercado were predicted to be in danger of being in the bottom three and were.

For those keeping score – Tivo and Zabasearch weren't any more accurate. Tivo predicted Syesha to be eliminated and she was not. ZabaSearch predicted a bottom three consisting of Carly, Brooke and Syesha – which was no more accurate than DialIdol. They also said no one would be sent home which was wrong. To top it off, ZabaSearch changed their prediction to reflect the results after they were announced.

Of course it wasn't DialIdol's brightest moment – but DialIdol wasn't wrong either. Who knows – with season highs in traffic for a Tuesday, maybe DialIdol visitors saved Carly and Syesha who flip-flopped for the bottom slot all night long?

• I haven't poured over the data like I usually do in the morning for two reasons - the first is that the results show doesn't air until tomorrow - giving me more time. The second is that I think the predictions seem to fit for the most part.
  • With two hundredths of one percent separating them, I wouldn't be surprised if Michael Johns and David Cook trade places tomorrow night in the bottom three and because Brooke is only three tenths of one percent from both of them, even she could return to the bottom three.
  • It is pretty much a dead heat for last place between Carly and Syesha. Carly did better during the East coast vote despite starting the night off very weak. Syesha got a tiny bump from the left coast. My suspicion is that Syesha is in more trouble than Carly but if you've been reading this page lately - you know what happens when I start spouting off my own opinion. :)
• Dancing with the Stars results seemed to match DialIdol's predictions a little better - but not quite as good as American Idol this year. On a personal note - Sheryl Crow's performance of Out of Our Heads last night during Dancing with the Stars made me drop everything and listen. It was by far and away the best singing/song I heard last night and also by far and away the most inspiring of yesterdays inspirational songs.
• Traffic was pretty hefty last night - so much so that the server's got slow at times and I had to respond. Hopefully this is a good sign for Idol Gives Back.

• Right now DialIdol is configured for the standard two hour voting window. If American Idol changes the length of that voting window as they have been known to do for special votes like this then DialIdol may need to be closed/reopened to reflect the new voting window. It is very likely that they will not announce the length of the voting window until during the show and that means DialIdol won't be corrected until during the show. Worst case would be DialIdol doesn't dial the full window for you.

Week w/10 (to 9) Idols Comments

• Ramiele Malubay was predicted to be in danger of being voted off and she was so that counts as being accurate.

• Kristy Lee Cook was predicted to be in the bottom three and she was.

• Brooke White was predicted to be in the bottom three and she was, in fact she was four tenths of one percent away from being on one of the bottom three rows.

• On a lighter note, I guess I was wrong about the two other “predictions” - they did not name a bottom, middle and top three as they usually do and they also sent someone home – contrary to my speculation. I guess this underscores why the busy signals make the predictions around here – not me.

• Some observations from the data.
  • Ramiele had the lowest busy percentage during 82% of the East coast voting window last night. In fact her East coast trend was so weak that at the end of East coast voting she was definitively predicted to be voted off. This is very telling and not a good sign for her at all.
  • Ramiele got the same bump from the West coast this week that she got last week. Last week this bump seemed to keep her out of the bottom three - I don't know if it will keep her safe this week though.
  • Contrary to last week - Jason Castro started strong and then tapered off and contrary to Ramiele - his West coast trend is quite weak.
  • Only two tenths of a percent separate Jason, Kristy and Carly and only two tenths of a percent separate Carly from Brooke so the bottom three is going to be very close - just like last week.
• I'm merely speculating here and have nothing to base this on but with Idol Gives Back being next week and April fool's being yesterday, wouldn't it be a nifty little trick if no one was sent home tonight? This would be "shocking" which American Idol loves to be - and it would add drama to Idol Gives Back, which might lead to more viewers and thus more donations. One thing working against this is that last year on Idol Gives Back they sent no one home and I believe they explained 'that on a night like that - no one should go'. As usual, the only one way to know for sure is to - watch!
• On the other hand if Idol holds true to form we'll get a bottom, middle and top three tonight - the most detailed results we seem to get. Should be quite nerve wrecking, err, interesting - <gulp>. :)
• As far as Dancing with the Stars goes, no doubt about that one... Steve getting voted off was wrong. Quite frankly I think the problem is a lack of dialers (there were only like 10 people dialing) and as a result starting next week, for DWTS only, margin of error will apply in both directions instead of just one. Incidentally if I had that rule in place this week, Steve would have been in danger of being voted off.

• Some observations from the data.
  • Only four one thousandths of one percent separate rows 6-8!
  • Jason Castro fought back from last place all night and finally flipped into second last late into East coast voting. This could be a sign of a strong base - or something else.
  • Ramiele got a nice bump from the West coast - whether this truly reflects their voting pattern or not remains to be seen. I find this to be irregular. Due to this bump and the fact that she went first I think Ramiele may be in more trouble than DialIdol suggests - and as it is now she is in a lot of trouble!
  • It's early in the morning but if my math is correct - Carly Smithson would have been shown in the yellow had I not made the adjustment to the way DialIdol range is calculated.
• As far as Dancing with the Stars goes, no doubt about that one... It was a split. Monica getting voted off was right but Penn was wrong and would have been even if the DialIdol range change wasn't in effect. I'll look at this more closely this weekend but right now I am chalking it up to the unusual voting format for the week.

• In an effort to reduce the number of "ties" I have made an adjustment to the way margin of error effects the predicted DialIdol range which should result in less "yellow" predictions and more green/red ones.
  • The change being implemented is that when figuring out any one contestants range, margin of error will only apply to that contestant - the margins of errors for other contestants will no longer be factored in.
    • For example last week (w/11 Idols) Syesha's lowest adjusted score would have shown up as green because there is no way her lowest possible score (6.843) minus her margin of error (2.852) = 3.991 would have reached Kristy Lee Cook's score of 3.439.
    • In the past Syesha's lowest possible score of 3.991 was compared against Kristy's highest possible score of 3.439 + her margin of error of 2.903 = 6.342 and since Syesha's lowest possible score (3.439) was lower than Kristys highest possible score (6.342) - Syesha was considered in danger.
  • This change only applies to predictions made after today - not historical ones.
  • I do not believe this is fair to DialIdol but I'm going to roll with it until the busy signals pickup.
    • This new method would really only have made DialIdol wrong about the guys' week 20 prediction this year. Everything else was more or less the same - just more green and a smidge of red.
      • I created a modeling program that allows me to rescore any given week with different algorithms. Using the method in place now, that week 20 prediction for the guys would have been accurate even with this new MOE rule. This actually applies to several of the predictions from last year as well.
• As far as DWTS goes, I repeat - there is no prediction for the girls from last week because there weren't enough people dialing. Tonight that will change but don't expect this weeks girls DWTS to be even close to accurate since DialIdol is missing half of the girls data.
• Speaking of DWTS - lets hope it works tonight! There is a lot of room for error tonight because of the voting format for this elimination round. I've tested the code but that was last week which is a lifetime ago in DialIdol code years. Lol.
• An observation - as far as DialIdol predictions go - David Archuleta has won every single one of them to date. The only other person to win that many won American Idol season five - his name was Taylor Hicks.

• Looking at last nights voting data I made the following quick observations.
  • The very first interval Amanda scored unusually lower than the rest of the pack and she spent the better part of the next hour fixing that until she leveled off.
  • Looking at data from specific users I saw no support for Syesha in the top tier of people with busy signals though I did see one each for Amanda and Kristy. Syesha may have polled lower than usual yet again in DialIdol. I would be surprised if she ended up in the bottom three.
  • Ramiele and Chikezie switched slots as a result of West coast voting. Usually East coast voting means more so that could be something to watch for tonight.
    • Ramiele also got a big bump upwards at 1:09am (Eastern) or 9 minutes into West coast voting. This might have something to do with it.
  • All in all the predictions seem to be in order and regular.
• I get the feeling that Dancing with the Stars was no match for American Idol last night. Very few people voted for DWTS - so few that it didn't meet the requirements of one measly vote per contestant per interval - and thus DialIdol never produced a prediction for the girls.

• DialIdol has added support for Dancing with the Stars.
  • The default view of the Predictions page for Dancing with the Stars (http://stars.dialidol.com/asp/predictions/predictions.asp) will now incorporate the Judges scores into the DialIdol score per DWTS scoring procedures.
    • This same view will be the view seen in the DialIdol application.
    • This view is not available in DIHardGraphs (yet).
  • The traditional DialIdol score without the Judges scores is available in the Stat drop down.
  • The Judges scores are also available in the drop down (when they become available).
  • The week w/12 stars will pose a serious challenge for DialIdol as the voting procedures are confusing. It appears that on Monday you will vote for the guys only - then on Tuesday you will vote for the girls only. These scores are all then carried over to the following Monday when you vote for all the stars in one night - but they will be scored separately (guys/girls) and one guy will be eliminated on Tuesday (3/25) and so will one girl.
    • DialIdol was not really designed for this sort of thing so there could be some challenges. I have tested and will continue to test to make sure the bugs are worked out but don't be surprised if we have to chalk this week up to bad luck.
      • One of the known challenges that will NOT be addressed is that during the vote on 3/24 all the contestants will be lumped together but will only be competing against their group, in this situation the SAFE/UNSAFE dialing options in the DialIdol application will pick the star to dial for based on the scores of all 12 stars - not the individualized groups.
    • This coming week you will only see the group voting that night available in the DialIdol application. The following week you will (hopefully - fingers crossed) see all contestants and everything will work swimmingly.
    • On Tuesday (3/18) there will be overlap of both DWTS and AI - both the show and the voting windows will overlap.
  • From now on DialIdol will only accept data and auto-dial for the half hour after the show - it will no longer accept data or auto-dial during the show. (Unlike American Idol, DWTS phone lines open when the show begins and remain open for one half hour after the show ends.)
    • This is being done primarily because voting during the finale last year cost DialIdol the correct prediction.
    • It is also being done because DWTS limits the number of votes you can submit - all of which DI should be able to cast in that 1/2 hour.
    • It is also being done because most of the lines are busy during this period of time and thus there is a better chance for more accurate predictions.
• There is nothing new to report with DialIdol's coverage of American Idol - nothing has changed.
  • OK - since DialIdol for American Idol shares the same code as DialIdol for DWTS there is the opportunity for some whacked out behavior this week - but it all adds to the fun - right?

• I can't find any significant irregularities in the voting patterns - as far as that goes things look copacetic.
  • Last week's problem in with Tennessee getting an unusually high number of busy signals didn't seem to appear last night.
• Syesha looks to be in a lot of trouble, as far as DialIdol goes she performed very poorly most of the night including many intervals when she had no busy signals at all. That said - Syesha makes me a little nervous because I think she polls lower on DialIdol for some reason than American Idol results seem to indicate.
• David Hernandez is next in line for trouble and since he started off so poorly during the first half hour of voting I have to suggest that he is in serious danger too.
• Ramiele is right there with David Hernandez but I didn't see the same weakness in her voting pattern that I saw in Syesha's and David H's.
• Though everyone else except David Archuleta is within the margin of getting kicked off - I would be surprised if it were any of the rest.
• All signs point to Syesha going home tonight but there is a reason we watch - and that is because anything can happen!

• I looked at explanations for Danny Noriega's curious climb last week and found nothing irregular except the fact that it looks irregular.
• I tested seven different algorithms against this years voting data and though one method got the week w/16 Idols a little closer to the actual result - the method used last week was still the best overall so I'm sticking with it.
• DialIdol will be expanding support for Dancing with the Stars this year to incorporate the judges score into DialIdol's prediction. With this incorporated into the prediction DialIdol can now be graded on predictions for Dancing with the Stars!
  • I mention this because the Dancing with the Stars page on DialIdol.com (www.dialstars.com or stars.dialidol.com) now has a new drop down on the predictions page in support of this.
    • The default view of that page remains the DialIdol score without the judges score. that will change next weekend.
    • The DialIdol application will not yet use this scoring method.
    • There will be more news coming in the next week on DWTS!

• DialIdol v3.5 is being released today. Changes in this version include:
  • Skype support is being introduced in this version. Click here to read more about DialIdol & Skype.
  • There is a new seasonal drop down so you can view your historical stats.

• DialIdol is moving to its in-season home these next few days so the site may act a bit hokey.
• I've archived all but the last months posts from the forums.
• I think I've got the whole SPAM thing under control so I've re-enabled the ability to send people emails (if you've authorized such an action) from the forums. Let me know if anything seems out of sorts.
• I'll be reintroducing some new stuff in the next week as we march towards the first vote on 2/19!

• I'm liking the kindler gentler auditions with a bigger emphasis on the people who advance to Hollywood. I've seen some good talent too but are they ringers?
• I'm posting the alleged top 24 which is already plastered all over the Internet. This stuff usually leaks for lotsa reasons. It might explain why they do everything live when it really counts.
• I'm working on DialIdol offline and got a few things that will show up in February.
  • Skype compatability will be beta tested in the next version of DialIdol. Since I've never used Skype to vote and since they currently seem to block that number I am calling it a beta-feature. Not fully supported.
  • I'm going to add some web polls to the home page.
  • The never ending quest for a prefect predictions page continues and this season v3 hits the web!

• Happy New Year!
• I'm dusting DialIdol off for the new year... Fixing some old bugs and adding some stuff.
  • Online now
    • Word verification to the Contact page and to the forums registration page - this will help reduce SPAM for me.
    • Forum registration can finally succeed now. Previously emails weren't being received so clicking on the link was impossible and registration had been failing for all new users.
  • Coming in February
    • Enhanced Predictions page table
      • Simpler to read, easier on the eyes and better communicates DialIdol's message.
      • Fixes a bug with Firefox where it consumed the entire CPU and generally trashed the browser when more than four contestants were displayed. Unfortunately the fix introduces a new bug for Firefox where the tooltips don't display right in Firefox. There is nothing I can do about this - its a bug in Firefox or Flash.
    • All historical data is online including the season DialIdol was initially developed which was mid-way through American Idol's fourth season.
    • I've fixed a bug preventing the display of GeoPredictions weeks where the guys/girls were split.

• Updates from 2007
• Updates from 2006
• Updates from 2005