• DialIdol v3.4.0.0 has been released.
  • This version obeys the "Time Between Calls" setting when it dials the number known to be busy. It is a minor feature enhancement but given we are in the off-season right now and gearing up for Dancing with the Stars I figured I'd get everyone on the same version.
• It appears Dancing with the Stars is going to combine the results of the six couples that perform Monday with the six couples that perform Tuesday - voting one couple off based on the votes of both shows. This seems a bit unfair to me from a competition perspective because the audience size - and thus the voting pool may be different from one night to the next but whatever. Since this is a one time only deal and its the first week of the season when people are very disinterested I have decided to not do anything programmatic to obey this... So during both votes you will actually be able to call a number for the people not performing. This can have adverse impacts on the predictions but since they are not expected to be accurate we'll live with it.
• I may build the judges scoring into the predictions for Dancing with the Stars since it is popular and frequent. Stay tuned - though it may not happen until season 6.
• Go Tribe!

• DialIdol v3.3.0.0 has been released.
  • This version will allow you to submit call data to DialIdol even if your data is all busy. To achieve this DialIdol will now dial a number known to be busy during the show if a certain number of calls in a row (50 for now but its configurable) are NOT busy. If that attempt results in a busy signal data will continue to be submitted.

• So if you've visited the Predictions page for SYTYCD this year you saw some pretty whacked out DialIdol ranks. This has been fixed. It was not a bug in code but rather a very basic configuration setting that I introduced during season 6 of AI gone wrong. The delay in fixing this bug was because I am trying to enjoy my summer without interference from DialIdol and because I knew tracking the problem down would take as long as it took - 4 hours (spread over two weeks). Anyway - it is fixed.
• There will be no support for Canadian Idol anymore. Low traffic combined with it being very difficult to enter the phone numbers and me wanting to enjoy my summer are the reasons. Sorry.
•  Thanks in part to all of you, I now have this...

• Oops... I haven't watch SYTYCD yet this season and didn't realize it was time to vote this week. Hopefully we'll be ready to go next week. I'm not testing it nearly as much because this is my off-season and so that may mean more bugs.
• Go Cavs!

• DialIdol is moving to it's summer home today and will be up and down. Especially the forums - I will be doing some maintenance on them as well.
• Save Jericho!
• Go Cavs!

• It turns out the predictions for Dancing with the Stars were very accurate this season including the definitive final prediction that was dead on accurate.
• Just to be clear - DialIdol is making a definitive prediction for the finale of American Idol Season 6. Margin of Error doesn't matter one little bit today. Jordin Sparks is predicted to be the next American Idol. So You Think You Can Dance will be covered too so look forward to that. It follows the same format as American Idol.
• Save Jericho!
• Go Cavs!

• If you are a Jericho fan like I am who is NUTS over CBS' decision to cancel the show in the midst of a massive cliffhanger - check out the web sites below. Even if you aren't a Jericho fan you will laugh at how we loyal Jericho fans have sent 6721 pounds of NUTS (and counting) to CBS (yes we literally sent NUTS and there is a deeper meaning than you know).
  • http://www.jericholives.com (the rallying site)
  • http://www.nutsonline.com/jericho (a stats page showing details on the NUTS shipments)
  • http://www.youtube.com/watch?v=zMVo4v8Avzw (video of one of the deliveries)
• Go Cavs!

• Scores are so close this week that DialIdol is unable to produce any substantive prediction. (Sorry!) You can look more closely at the data to get a better sense of how your favorite is doing and draw your own conclusions.
  • Check out the GeoPredictions page that breaks down the vote by state and even area code!
  • Check out the DIHardGraphs page that shows patterns that developed throughout last nights voting.
    • See how the West coast votes differently than the East coast.
    • See how Melinda did extremely well on the East coast but not so well on the West coast.
    • See how Blake and Jordin battled it out for last place the entire night.
• Looking at that data myself...
  • I see that Melinda is probably in good shape (but still within range of going home) because she did well during the East coast voting which is critical. She was not popular on the West coast though which could be the X factor that sends her home.
  • Jordin and Blake are separated by 6/10 of a percent. This is incredibly close!
  • Jordin grew her lead during the most critical hour of voting - the first hour and also did well during the West coast vote.
  • Blake edged Jordin during Arizona/Mountain voting which isn't a critical time period.
  • Blake lost his home state of Washington to Melinda. In fact he wasn't even competitive!
  • Analyzing the data seems to point to Blake going home but it really is so close any of them could go home - we'll just have to watch and see!
• Don't forget that during this week last year the results proved that all three Idols (Elliot, Taylor and Katharine) had 33% of the vote and were all separated by less than one percent!
• Go Cavs!

• Rumor is that there will be three lines per contestant tonight. Since it is still just a rumor I haven't added the third line yet. When it is confirmed I will add the third line - if you don't see three lines per contestant but the show announced three lines per contestant close and reopen DialIdol. Or just wait until about 8:20pm Eastern before opening it period. I should have all the info in by then.
• Go Cavs!

• The voting window is usually two hours for non-finale, non-charity type weeks. This week DialIdol is assuming that American Idol will return to a two hour voting window. If the show announces a four hour window you must close/reopen DialIdol to reflect the new four hour window. If necessary you will want to do this around 8:40pm Eastern.
• DialIdol is assuming there will be two phone lines per contestant this week but there is a small chance there could be three or more - DialIdol is prepared for this.
• DialIdol's DNS issues should be completely worked out. I switched registrars and name servers and now have complete control over the environment.
• This page is getting pretty long.
• Go Cavs!

• My hometown newspaper, The Plain Dealer, did an exceptional article on DialIdol complete with a full page story on the front of the "PDQ" section. Of particular interest is a quote from FOX spokeswoman Jennifer Sprague where she finally confirms that "Using a modem dialer is not the same as 'power dialing'". Which means DialIdol is not subject to their power dialer rule - thus confirming that all DialIdol votes count!
  • Reading the article online is not the same experience as viewing the article in The Plain Dealer so if you have access to this paper then I highly recommend getting it. Finding the "PDQ" section is a bit difficult but once you do you can't miss the article - it consumes the entire front page complete with a 1/2 page picture of my head!
  • Here is a link to the full quote from FOX which includes a surprising-to-some revelation that they have never manipulated the vote due to power dialers.
• After analyzing the data there were several bugs that effected the scoring this week and for the first time ever I am manually overriding the automatically generated data with data that better reflects what would have happened without the bugs. The override is not significant though.
  • The primary bug caused the time zone divider to be different amongst contestants. This should be the same for all contestants so I made it the same. The result better matches a respected anonymous person's calculations though it uses a different method. It also "jives" better with what GeoPredictions shows us.
    • This bug caused the scores between the two lines of a contestant to be different (seen only in DIHardGraphs). Though they are still different in DIHardGraphs, they are now the same on the predictions page (even though you can't see it).
    • As a result of correcting the data, DIHardGraphs will not properly reflect scores for Idol Gives Back week however when voting resumes the scores for this week will be correct. So 1/2 of week 6 data in DIHardGraphs will be a little off, 1/2 will (hopefully) be a little right. :)
    • This correction effectively means that those score intervals that were all zeros did not impact scoring because the only thing they did was increase the divider for specific contestants.
  • A second bug existed that was fixed during the vote last week that caused multiple intervals to be posted per interval. These extra intervals were all zeros. These intervals have been deleted from the database so they won't show up in DIHardGraphs anymore. Other than not showing up in DIHardGraphs deleting the data has no effect.

The OLD predictions are shown below.

• The GeoPredictions page no longer shows ties.
• Voting will resume Tuesday night with no worries about how the predictions will be impacted.
• I've got a new server up and running that will be available to service the application Tuesday night so hopefully capacity issues won't be a problem.
  • A second server will post twice the interval data and might fake you into thinking that the bug I mentioned above is back - its not.
• My DNS issues haven't been completely worked out yet but hopefully they will be by Tuesday.

• I swear I put code in to prevent these ties from happening! The proof exists in higher scores (which are higher partially because of more busy signals and partially because of the changes I made) and lower margin of error.
• If you were running the DialIdol client to vote last night you may have experienced many problems including the application taking a long time to load, possibly not loading at all, hanging while submitting data, etc. These problems were all due to these primary factors.
  • High call volume.
  • Two phone numbers per contestant meant twelve lines and exercised additional code that cost more CPU processing time which meant that DialIdol's capacity was less than last week and probably less than at any other point this season.
  • Normally the web site you visit with your web browser is served by one server and the client application is served by a separate server with the intention of utilizing both for both purposes under high traffic scenarios. However this weekend there were some DNS problems that caused DialIdol.com to be down. In order to prevent those problems from interfering with last nights vote I tweaked the DNS time to live (TTL) for DialIdol.com which prevented me from adding the other server into the mix to spread the load over DialIdol's two servers.
    • Since I am trying to fix two bugs that counteract each other this is harder to fix but not as big of a deal as you'd think.
  • I will be adding a third server to the mix for the remainder of the season to prevent this from happening again.
  • The fewer total phone lines, the more capacity DialIdol can handle so based on past seasons things should only get better from now.
• Prior to 10:18pm EST last night the extreme stress the server was under produced an interesting bug that resulted in several intervals being posted "twice". Once with the appropriate data and once with no data at all. This can be seen a little in DIHardGraphs. I coded a fix at 10:18pm EST and it didn't happen again.
  • This will have a small effect on the predictions. In fact since the scores are so close between Melinda, Blake and Lakisha that had this bug not existed their row in the chart may be different. Though Chris would still be at the bottom and Jordin still at the top. I don't focus on the row position because the DialIdol rank is more representative of DialIdol's predictions however some of you do so keep this in mind.
• Another bug surfaced that made the scores for each contestants line be mismatched which isn't supposed to be. This can only be seen in DIHardGraphs. Not sure what caused it but the server not being able to keep up with the demand is an easy cause.
• Yes the GeoPredictions page is showing ties everywhere - expect a fix this weekend at the earliest. Hovering over a state/area code will show you who is winning/losing.
• All of this stuff doesn't change the fact that Chris Richardson has the same pattern that Sanjaya, Haley and Gina had the past three weeks and we all know what happened to them.

• So I have spent a lot of time pondering the predicting formula since I last wrote... Here are some thoughts.
  • Dancing with the Stars has such low scores for two reasons.
    • Voting "kicked in" earlier than usual - during the show (you are allowed to vote DURING the show). The lines aren't busy at all at this time so the scores were naturally low which brought down scores.
    • The busy percent weight that was implemented yesterday as a trial ended up weighting scores even lower than usual since the busy percentages were low.
  • As a result I have decided to back out the busy percent weight which will then limit the formula to adding only two more things for this week.
    • The first is to average the busy percents of each submission during an interval to determine the busy percent for that contestant instead of adding all the busy/votes together and getting the percentage from that. This will prevent people who are exclusively voting for one contestant to over influence others who are voting for more than one.
    • The second is more important and is the primary driver for changes at this stage of the game. The change is being implemented to put an end to the statistical ties that have become problematic recently. To achieve this the DialIdol score will now be divided by the sum of the interval weights as opposed to the number of intervals.
      • This alone will not change the "row order" or proportions of each Idols score compared to each other but it will make the scores higher and MOE lower which will result in less ties. It will also finer tune "DialIdol rank".
• As I mentioned - DialIdol is prepared for two (or more) lines per contestant and for a four hour voting window - both of which seem likely to happen tonight. If this happens please close/reopen DialIdol if you only see one of the contestants phone lines. The phone numbers are dynamic - the contestant list and voting window are not.
  • If there are two lines per contestant the GeoPredictions page will show ties everywhere - this will be fixed next weekend at the earliest.

• There were two significant outages this weekend that resulted in DialIdol.com being down. The problem was with the DNS servers that map the name DialIdol.com to the IP address your computer needs to know to render the page. I have little control over this environment (though I am trying to change that) and am not entirely sure the problem has been completely resolved but we'll see.
• Given that American Idol expects 60 million votes this week I suspect they will be surprising us with either more phone lines per contestant, an extended voting window or both. DialIdol is prepared for these circumstances.
  • Right now a bug exists in the GeoPredictions display with more than one phone number per contestant. The color will reflect a tie because all the lines for a contestant are reported on and all the lines have the same score.
• I have made great progress on improvements to DialIdol's predicting formula but am not 100% sure I will use it this week. The formula itself improves accuracy such that 5/6 of the people to leave the top 12 were in the bottom slot. It also makes the bottom 3s look a lot better - but not perfect. However I am not entirely sure every line of code has been tested in that implementation. A lot will depend on whether they use two lines per contestant this week, how voting goes for Dancing with the Stars tomorrow and whether I have time to test and review the code more before the vote Tuesday.
  • It may seem insane to do this on the heels of getting last week right but there is still the problem of predicting everyone to be voted off except one person to worry about!
• I ended up increasing the consecutive call rule to 400 last week, not the previously reported 300. This had a major impact on the predictions and this week I may bump it even higher - especially if the voting window is extended.
• Other changes to the formula that are not currently documented anywhere else.
  • Previously during a single interval all the busies and votes were tallied to get the busy percentage. This was having adverse effects on Phil & Haley's scores. The way this will work going forward is that DialIdol will average the busy percents of all submitted DialIdolers. What it protects against is one person or a few people over-influencing the vote. More information later.
    • For example previously if two people submitted data as follows in an interval:
      • User 1 : Phil votes : 1 Phil Busies : 9
      • User 2 : Phil votes : 0 Phil Busies : 4
    • Under the old method the busy percent would be 1 + 0 / (9 + 4 + 1 + 0) = 7%
    • Under the new method the busy percent is : (1 / (1 + 9)) + (0 / (0 + 4)) / 2 = 5%
  • More weight will be given to the intervals that have higher busy percentages. The score will be calculated as previous - then multiplied by the average busy percent. If at 9:12pm the lines are 90% busy the score will be : 90% * 78% * 90% = 63%.
  • DialIdol will not divide by the number of intervals anymore but instead by the sum of all the weights of the time zones for each interval.
    • This will result in higher scores, lower margins or error, less ties, and a more (but not completely) mathematically valid MOE number.
• Go Cavs - thank goodness there are no games scheduled during a vote!

• I have increased the consecutive call rule to 300. That is 300 consecutive calls of the same result will trigger a submission stop. It used to be 200. This is effective this week.
• I've got good news and bad news.
  • Good news
    • I have decided to make fairly significant changes to the predictions formula that when applied to past weeks improve the accuracy of the predictions.
  • Bad news
    • Due to a business trip I had to go on for my day job I was unable to sufficiently test the rather significant code changes that needed to be implemented to support the new formula so it won't be available this week.
  • I may be able to post a theoretical predictions using this new formula late Wednesday afternoon - around 6pm'ish in the forums so look for it. DialIdol will not be scored against it though.
  • This is all subject to change as further testing may prove it doesn't work well enough or can't be posted this season.
• Anticipated changes for next week.
  • The formula will implement the following major changes.
    • Previously during a single interval all the busies and votes were tallied to get the busy percentage. This was having adverse effects on Phil & Haley's scores. The way this will work going forward is that DialIdol will average the busy percents of all submitted DialIdolers. What it protects against is one person or a few people over-influencing the vote. More information later.
    • More weight will be given to the intervals that have higher busy percentages. The score will be calculated as previous - then multiplied by the busy percent. If at 9:12pm the lines are 90% busy the score will be : 90% * 78% * 90% = 63%.
    • DialIdol will not divide by the number of intervals anymore but instead by the sum of all the weights of the time zones for each interval.
      • This will result in higher scores, lower margins or error, less ties, and a more (but not completely) mathematically valid MOE number.

• Happy Easter, if applicable!
• Check out DialIdol's power rankings in Rolling Stone magazine on newsstands now.
  • I actually haven't seen a copy yet because the issue is not yet on my newsstand! :(
• The tiny bit of study I did on the predictions this week indicate Phil Stacey has a strong support base in the Mountain timezone which, combined with lower call volume compared to other DialIdol dialed contestants may be elevating him higher in recent weeks. I haven't concluded this as a strong theory yet and don't have any code in to fix it.

• American Idol is 67 minutes long this week so voting begins at 9:07pm EST.
• Doh! Dancing with the Stars allows you to vote during the show - DialIdol will now allow you to vote during the show too.
• I am releasing DialIdol v3.2.2.0 today which is a minor improvement.
  • Previously DialIdol calculated the call timeout automatically by adding 3 seconds to the average length of time it took to detect a busy signal. The 3 second value is now configurable in Config/Options. It can go no lower than 3 seconds and no higher than 10 seconds. Whatever this value is set to (still 3 seconds by default) will be added to the average length of time it takes to detect a busy signal to determine the call timeout. If less than three busy signals have been detected the call timeout is not auto calculated, this new value is irrelevant and the call timeout in Config/Options is used instead.
  • If you think your calls were timing out before the vote was registered (as one user did and as I observed during Dancing with the Stars) then increase this new timeout.
• A bug in the scoring struck DialIdol's predictions this past week. Basically the new rule that set a minimum of 10 calls changed to 0 calls at 2:15am EST because IIS (the web server) recycled DialIdol.com and I forgot to configure the minimum when IIS starts. The rule worked fine before 2:15 but worked worse than ever afterwards.
  • Stephanie Edwards benefited from this significantly and probably would have been in a lower row had the rule been working.
  • Chris Richardson wouldn't have been effected too much.

• The GeoPredictions page will now display states that are tied and states that have no scores in unique colors.
• The DIHardGraphs page now has 2.5 times the interval data and allows you to resize the graph so you can look more closely at the graph.
  • This data always existed - it just wasn't displayed. Even now every few intervals is not displayed. An interval occurs every three minutes.
• After parsing through the data I was able to notice...
  • How Brandon's score rocketed higher by one single phone call being placed every few minutes. If the scores from these intervals were excluded from scoring (due to low call volume) Brandon would have been in last place. Excluding this data from scoring when call volume is low is the right thing to do and so from now on a (webmaster) configurable minimum amount of calls must be reached for each contestant in order for all of them to score in any given interval. Right now the minimum is 10 calls.
    • You can see this on the DIHardGraphs page if you check out Brandon's interval statistics and pay close attention to the tail end of the West coast vote (the 2am hour).
  • Sanjaya had some weird stuff going on and none of it is easily fixable. He did have extremely low call volume and had this been higher his score would probably have been more accurate. 
• I am releasing v3.2.1.0 today which fixes a hang condition that existed when the clock struck midnight. This condition may have also manifested itself under different circumstances though those would be few and far between.
  • If you continue to have these or other problems please continue to post your problems in the forums or contact DialIdol.
• DialIdol will be available for Dancing with the Stars mostly to help you VOTE. Due to this shows format the predictions are NOT expected to be accurate at all and no matter what this show will NOT effect DialIdol's accuracy.

• The colors of the key in GeoPredictions now match the color of the states.
• The Predictions page got a makeover.
  • Detailed tooltips now appear when you hover the mouse over each Idol explaining the prediction for that contestant.
    • Thanks to this dude for posting his awesome free tooltip code on the Internet.
  • For each contestant DialIdol will now make one of three predictions.
    • Predicted to be safe
    • Predicted to be in danger of being voted off
    • Predicted to be voted off
  • The color of the font AND the color of the bar graph will now BOTH represent a color that corresponds to one of the three predictions DialIdol will make.
    • As usual blue means they are predicted to be safe.
    • This new orangish gold color means they are predicted to be in danger of being voted off.
    • Red means they are predicted to be voted off.
  • The flash bars should now load in Opera browsers.
• I've changed the way DialIdol grades itself in terms of the accuracy of Predictions. I've also created a FAQ on this.
  • This new grading method changes DialIdols correct percentage from 100% down to 92% for season 6 of American Idol through week 16.
    • This is because Alaina Alexander was absolutely predicted to be voted off but didn't get voted off during the week w/24 Idols.
• DialIdol v3.2.0.0 continues to remain the current version of DialIdol as I am not releasing anything new.

• I am releasing v3.2.0.0 today, This version should:
  • Eliminate the modem in use by fax dialog box that was added in v3.1.0.0 - except when you run "Is DialIdol Working?" - it will still show up there.
  • Fixes bugs that surfaced when you clicked an ad in the ad window.
  • Automatically adds pauses between calls if more than 2 or 5 errors are detected which should fix the problem of many errors accumulating that some people were experiencing.
• I've tweaked the way margin of error is calculated to reduce its size. This doesn't effect DialIdols accuracy claim of 100% though it does make this seasons predictions look a little better because less people are predicted to be voted off for weeks 24 & 20. That is not why I did it though, I did it because going forward I want smaller margins of error and less contestants to be predicted off.
  • This tweak more closely reflects last years method. Technically there was no change to any formula - just factoring in the consideration that DialIdol is running on one server now - not multiple.
• In case you are wondering about traffic to DialIdol.com here are some stats from this past week alone.
  • There were more than 1 million page views on more than 15 million hits.
  • The Predictions page was viewed over 500,000 times.
  • Thursday (3/1) was the most active day with just under 500,000 page views.
  • Roughly 100,000 people visited DialIdol.
  • Almost 600 people dialed with DialIdol.

• I am releasing v3.1.0.0 today. This should fix some compatibility issues DialIdol has been having. It also enables you to create a log file (via Config/Options) that may shed some light any potential future issues.
• I added a bunch of area codes to the GeoPredictions page to satisfy all requests as of today.
• The DIHardGraphs page is now displaying the correct data.

• I am releasing v3.0.1.0 today. This should fix some of the bugs associated with the "Is DialIdol Working" feature. The bug that it fixes would not have prevented DialIdol from actually dialing for real - just via this "Is DialIdol Working" feature. There still appears to be some additional issues that I'll be working through.

• Well here we are - time to vote again! Personally, I significantly prefer the live performances and drama that goes with them to the auditions portion of the season and this preference has nothing to do with DialIdol.
• The new DialIdol and DialIdol.com have been posted. It took more time, effort, stress, planning, blood and sweat (fortunately no tears - but it came close a few times!) than last year but its all worth it now as I triumphantly bring you up-to-date.
  • The new look is much improved - don't you think?
  • I am thinking that you're really going to love the new GeoPredictions feature. It's against all odds that this feature actually made it onto the site and looks as decent as it does. I am really excited about it on a personal level because (obviously) I love statistics and this feature gives us a whole new breakdown of statistics and brings voting even closer to home.
    • Unfortunately not every area code has been charted on the map for display. The data for every area code is being collected and as the season progresses I'll find time to punch in the area codes that need displaying on the map. If you're actually reading this and you want your area code on the map - send me an email and I'll expedite it.
  • DIHardGraphs is another neat feature that will allow you to chart a breakdown of statistics every few minutes throughout the vote.
  • The predictions page got overhauled and is hopefully both easier on the eyes and easier on the brain.
• The DialIdol application was 80% rewritten, especially the modem interface pieces. Though it hasn't been tested on a mammoth number of PCs yet, I like this code much better and expect it to solve a lot of peoples problems.
  • The new Auto-Start dialing feature is pretty cool, I plan to use it exclusively.
  • DialIdol now has the ability to track your statistics from week to week so you can look back and see how you voted or if you accidentally close DialIdol and reopen it your stats are still there.
• I've changed the way people contact me from directly sending me an email to using a contact form. I was getting too much SPAM and this had to be done. Now when you use the form to contact me that email will rise to my attention instead of potentially getting lost in a web of SPAM. The old email addresses will still be used to communicate with you and your SPAM filter still had to filter those addresses (*.dialidol.com) out.
• So that's that. This first week will be interesting. I expect there to be some bugs in the system that need working out so please be patient with me and remember DialIdol is still free and still just a hobby for me!

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