American Idol Season 9 commentary

Finale commentary

As in last year's finale, text-message votes probably tipped the balance (causing Lee DeWyze to win over Crystal Bowersox.  (For reasons explained below, the voice phone lines were saturated, which means those numbers were not significantly different between Crystal and Lee.)  As explained below, whichever singer has wealthier fans has an advantage, because those fans can afford to cast more text-message votes.

Lee had an advantage because of his appeal to girls and young women (who tend to be the biggest intended demographic), and because Crystal's tattoos and single-mother-status offend conservative families (who vote in large numbers for religious & political reasons).  And as stated here the week before the finale: "The greater enthusiasm/dedication of fans of Lee DeWyze may cause him to win."

Special Tip for the Finale's Official Voting!

Here's the voting tip:  While your favorite contestant's lines are busy, use a land line to call the other contestant's line, and when you get through, do not hang up.  Instead, wait until you get disconnected!  For example, if you prefer Crystal Bowersox, call one of the numbers for Lee DeWyze and stay connected when you finally get through, and wait to get disconnected.

Here's why it's fair, how it works, and some clarifications:

• During the time you are hogging a voice line of the contestant you like less, you are blocking that contestant's fans from reaching that (virtual) voting destination.  Fewer votes getting through means fewer votes for the competition.
• Without this strategic voting, both contestants will get about the same number of votes on their voice lines—because each voice line has the same capacity limit.  This equal-capacity problem is explained below in the saturated voice lines section.
• This hog-the-competition's-line voting strategy is available to the fans of both contestants, so it's not biased in favor of either Crystal or Lee.  Whichever contestant has more-dedicated, better-networked fans will get the bigger advantage from using this voting strategy.
• It only works if you are using a land line.  If you are using a cell phone, do not use this strategy, because it is more likely to hurt your favorite (for reasons explained below).
• The strategy does not work for text-message voting!  It's only for phone voting from land lines.
• Don't forget to sometimes—especially after the first hour—check all the phone lines for your favorite contestant.  If they aren't busy (that is, if you are often getting through), then keep those lines busy (instead of blocking the competitor's line).
• When you call to vote for your favorite, hang up as soon as you know you got through.  Don't block other people from getting through to your favorite.  (The tendency for fans to listen to their favorite's voice may account for why, this year, the show didn't use the singer's voice to answer the line.)
• If you normally use text messaging to cast your votes, stick with text-message voting!  That's because text-message votes do not have any significant bandwidth/capacity limits, so every text-message vote counts.  (In contrast, each time you dial a voice line, it only counts if you get through.)
• Yes, this approach is actually fair.  Whichever contestant has more fans using this strategy, will get more voice votes compared to the other contestant.
• The show doesn't use fair voting methods, so this strategy actually improves the fairness.
• Finally, this strategy only works in the finale, where there are just two choices, and where there are so many people calling on the voice lines.  (Earlier in the season the voice lines are not continuously saturated, and the results are unpredictable if there are more than two choices.)

Power text-message voting

Based on evidence from previous results, lots of people are doing power texting.  Is power text-message voting fair?  No.  Especially because it gives an unfair advantage to wealthier people who are better able to afford text messaging (and unlimited text-messaging plans).  See below (How to replace the judges' save with a voter-based save) for how this problem can be solved.

Text-messaging traffic has no noticeable bandwidth limitations (how many text messages can be handled per minute), and that traffic cannot be measured the way busy signals on land lines can measure phone-vote counts.  So until the results are revealed, we don't know which contestants are getting high volumes of text-message votes.  In contrast, as explained below, and as measured using busy signals, the voice-line vote counts are not meaningfully different.  The result is the increased importance of "power" voting with text messages.  The show has not yet figured out how to deal with this bias.  (For perspective, remember that during season 2, when Ruben Studdard won instead of Clay Aiken, vote counts were skewed because of votes dialed from computer modems.)

Saturated voice lines

The official vote counts on the voice-based phone lines are about the same for each contestant.  You can think of it this way: if an amusement park has two popular rides, and each ride handles 100 people per minute, the ride with the longer waiting line does not serve any more rides than the one with the shorter waiting line.  Until the official vote-counters measure the diversity of voters getting through—which is like taking into account how many repeat riders are in each line at the amusement park—there will be yet more surprise eliminations.

Outrunning a bear

The American Idol show makes the common mistake of thinking that the choice with the fewest votes is least popular.  It's not unlike the joke about outrunning a bear: you don't have to be the fastest runner, you just have to run faster than the slowest person.  This means the fans of the least-popular contestant just need to provide slightly more votes than the fans of the next-to-last contestant.  When there are more contestants (say, more than four or five), that's easier to do.

The usual voting tip: use a land line

It's useful to consider that each official vote from a land line is worth, let's say, two official votes from a cell phone.  Why?  Because during a cell-phone call you are blocking other voters who are trying to get a dial tone from the same cell-phone tower.  And the people you are blocking are (statistically) likely to be voting for the same contestant.  (Don't take the two-to-one ratio number too literally; it's just a way to convey the importance of using a land-line phone.)  If you use a phone that goes through a cable-TV service, that's not a land-line phone; cable-based phones are also bottlenecked, in this case by cable-TV equipment.

Top 3 results

The elimination of Casey James was not a surprise.

Top 4 results

The elimination of Michael Lynche was not a surprise.

Top 5 results

How could Siobhan Magnus—who is the most popular contestant in this past week's VoteFair poll—get eliminated this early?  (And how could almost-least-popular Aaron Kelly end up in the top three instead of the bottom three?)  Here are the special reasons that apply this week:

• The Facebook fan page for Siobhan Magnus contained Aaron Kelly's official-voting phone numbers.  (Wikipedia has a link to that historic page.)  The wrong text-message number would not be noticed.  And this year, the wrong voice-phone number was not noticed because the singers' voices were not used to answer the calls.  ["Keeping it real": A previous comment here was mistaken because I was out dancing on voting nights, so I didn't know about this change.]  This also explains why Aaron Kelly was not in the bottom three.
• This week's “country-music” orientation attracted country-music fans who voted for, and probably told their friends to vote for, Aaron Kelly, who is fond of singing country music.  This surge of heavy voting by country-music fans has been a problem in the past; that's why the show no longer calls it country-music week, and presumably why they delay it for when there are fewer contestants (so that non-country-music fans aren't doing as much vote-splitting).  But putting Shania Twain and her songs on the show had the same effect.

Top 6 results

Finally, long overdue, Tim Urban was eliminated.  See below (Tim Urban's delayed elimination) for the voting-based reasons he stayed on the show for so long.

Top 7 results

This week's double elimination of Andrew Garcia and Katie Stevens was not a surprise regarding Andrew Garcia and was only a slight surprise regarding Katie Stevens.  Almost no one was safe, so there wasn't as much strategic voting (because that requires knowing how others will be voting), and that eliminates one source of unpredictable results.  As previously explained, Katie Stevens was vulnerable because voters who prefer a female winner are heavily voting for the dramatically better female singers, namely Crystal Bowersox and Siobhan Magnus.  (In contrast, the gap between the worst and best male singers is not as large.)

How to replace the judges' save with a voter-based save

The voters, not the judges, should be the ones who save a contestant from an early elimination.  How can this be done?  It is only possible if 1-2-3 preference information is collected from at least some of the voters.

Taking into account the objectives of the American Idol show, the best approach might be to allow text-message voters to submit a 1-2-3 ballot once from each cell-phone number.  For example, texting “4 1 2” to a special (cell-phone) number can indicate contestant 4 as the voter's first choice, contestant 1 as the voter's second choice, contestant 2 as the third choice, and (in this example) the other contestants would be tied as least-preferred.

These ballots would be counted to produce “pairwise counts,” of the kind shown in each VoteFair-results page.  If these pairwise counts reveal that the contestant with the fewest traditional (first-choice-only) “votes” is actually preferred (based on the pairwise counts) over the second-from-the-bottom contestant and the third-from-the-bottom contestant (and possibly the fourth-from-the-bottom contestant, depending on how many contestants are left), then that contestant would be saved automatically (without the involvement of the judges) and the contestant with the next-fewest (traditional) “votes” would be eliminated instead.

(For advanced readers of this commentary, an even more important consideration would be to protect the second-most-representative contestant, who is the contestant who would be most popular if the voters who support the most popular contestant have their influence reduced by an appropriate amount.)

AT&T might initially object to running software that supports 1-2-3 voting and pairwise counting for American Idol use—until they realize they can make lots of money selling the same service to many customers (especially collaborative businesses making wiser business decisions in situations that involve more than two choices).

Why Michael Lynche got the fewest first-choice votes

The surprise of Michael Lynche getting the fewest ‘votes’ serves as yet another example that the contestant with the fewest votes is not necessarily the least popular.  For readers following this commentary, his last-position vote count was not a complete surprise because an earlier commentary here pointed out that earlier VoteFair poll results in the Top-10 week (based on rankings from 801 voters) suggested that Michael Lynche was a likely surprise elimination.  (In those results, notice his short traditional-vote count.)  Although he didn't end up in the bottom three that week, vote-splitting easily explains why he got the fewest first-choice votes this week, even though he is about fifth (out of nine) in popularity.

The rule of the “judges save” was created to compensate for times when an obviously popular singer ends up with the fewest votes.  What the show's producers, and the official vote counters (who insist their counting is “fairer than fair”) still fail to recognize is that the contestant with the fewest first-choice votes is not necessarily the least popular.

Michael Lynche, Katie Stevens, Aaron Kelly, Andrew Garcia, and Tim Urban are close to one another in first-choice popularity, so the order in which they are eliminated is unpredictable, which makes the elimination sequence seem crazy.

Currently the big gender imbalance—namely 6 males and 3 females—makes the elimination of a male more likely.  That's why this week's bottom three were males.  If Katie Stevens, as the least-popular female, had a bad performance, that would have interrupted this bias.

Was it a mistake to use the “judges save” on Michael Lynche?  No, because the show's ultimate goal is to attract attention for the commercials, and the drama of the “judges save” increased the drama.  Also, this increases the drama for the following week when two contestants are eliminated.  Plus, the “judges save” is no longer available to compensate for unfair results.

Runoff voting isn't fair if it assumes the fewest first-choice votes indicates the lowest popularity

The same vote-counting mistake being made on American Idol voting is creeping into government elections in places like San Francisco (CA), Burlington (VT), and Aspen (CO) where a voting method named “instant runoff voting” (abbreviated IRV) is used.  The IRV method is based on the mistaken belief that the candidate with the fewest first-choice votes is the least popular.  The method has produced clearly unfair results, especially in Burlington and Aspen, so the method has been repealed in those places.  The instant-runoff-voting method correctly uses 1-2-3 ballots, but it also uses an unfair counting method based on thinking that the choice with the fewest votes is least popular.  It's crazy for American Idol to use the same unfair counting method that also fails to produce fair results in government elections.

Tim Urban's delayed elimination

Nearly everyone is asking: “Why is Tim Urban still in the competition?”  Here's why:

• Apparently he is getting votes from well-networked (white) Christian church groups.  (To be balanced about this, Michael Lynche is probably getting support from African-American Christian church groups.)  This isn't something new; it's been going on for the past few seasons, and became obvious in last season's finale.
• Very likely, many phone votes (including those mentioned above) are coming from people who do not watch the show.  Notice that any viewer can email, text, or call friends/church-members/classmates/etc. telling them when to start calling and which phone number to call.  Apparently this is more common among well-networked older folks, and less common among younger folks.
• When younger folks are well-networked, they disagree about who to vote for, so their votes are split among several different contestants.  As explained in Wikipedia's vote-splitting article, a large group that concentrates their votes on a single contestant can outvote a collectively larger group that splits its votes.  This is why 1-2-3 ballots (instead of single-mark ballots) are essential to fair voting results (and should be allowed as part of American Idol's text-message voting).  This is why money has so much influence in politics; people with money focus, instead of split, their financial support.
• Younger voters tend to use cell phones (and cable-based phones) instead of land lines.  This means that younger voters are blocking one another at cell-phone towers (and cable-TV facilities) while older voters are not getting as many busy signals because they use land lines.
• Finally, (prior to this commentary being posted) Tim Urban, along with Crystal Bowersox, Casey James, and Michael Lynche, were less self-conscious and less afraid compared to the other, less-popular singers.  This widespread excessive stagefright/self-consciousness/etc. was clearly evident in this year's disappointing Top-24 week, and continued until the Top 9 performances.  As stated here prior to the Top 9 performances, perhaps someone behind the scenes was getting angry at them when they made mistakes, or perhaps the judges accidentally chose singers who are afraid of big audiences, or perhaps the contestants needed (and then recently got?) more coaching in how to focus on their love of music to take their mind off the cameras, the microphone, the judges, and (unintentionally) distracting members of the audience.  Regardless of the reason, the less-popular singers were crippling their performances because of their self-conscious thinking.  In contrast, Tim Urban knows he is on borrowed time, and he's choosing to have fun while he's there, so (comparatively) he was more relaxed and less self-conscious than the others who are bumping along at the bottom.

Clarification: Poll results are not Tuesday-night specific

The poll results here are not Tuesday-night specific.  In other words, the predictions here are not updated after each Tuesday-night performance, so the predictions become inaccurate if singers do significantly better or worse than usual.  In particular, Andrew Garcia had a much-better-than-usual performance in the Top 10 week, so the prediction here that he was most likely to be eliminated became out-of-date.  Also keep in mind that this VoteFair poll includes many votes from earlier in the week (including right after the elimination).

Top 9 results

Didi Benami was a surprise elimination this week.  To see why this surprise happened, look at the Top 10 poll results and look at the lengths of the blue bars in the “Traditional vote count (for comparison)” column of the VoteFair results page.  The less-popular singers, namely Michael Lynche, Aaron Kelly, Didi Benami, Katie Stevens, Andrew Garcia, and Tim Urban are all bumping along at the bottom in terms of first-choice votes.  This means that a small difference in performance is all it took to determine who was eliminated.

If voters were allowed to indicate a second choice, third choice, and so on, the elimination sequence would not be so unpredictable.

Top 10 results

Finally Paige Miles was eliminated.  As fewer choices remain, it is more difficult for a group of people to keep someone from having the fewest first-choice votes.  Of course her awful singing this week helped to tip the balance.

Top 11 results

The loss of Lacey Brown is a slight surprise.  Yet this is the first week of voting for the males and females together, so AI voters are processing new information.  The results—only one male in the bottom three—indicates that voting for the female singers is concentrated on the top few females, especially Siobhan Magnus and Crystal Bowersox.  That leaves the less-popular females most vulnerable.

It was no surprise that Paige Miles and Tim Urban were in the bottom three because they are the least popular.

Top 12 results

As predicted here last week (see below), Lilly Scott and Katelyn Epperly were at risk of a possible early “surprise” elimination.  Why?  To repeat the point below, “These singers are liked by lots of people, but they are not the first choice of enough voters—compared to some of the less-overall-popular singers.”  Of course inconsistent performances and poor song choices also were involved, yet these two were still more popular than several other female singers.

Among the male singers, Alex Lambert was a surprise elimination.  To repeat the important point, “He is liked by lots of people, but he is not the first choice of enough voters.”

When comparing VoteFair ranking results to “plurality” results (which the show uses), consider that VoteFair ranking indicates how much a singer's concerts will be appreciated and how often their recordings will be played and requested.  In contrast, plurality ranking indicates how many music recordings and concert tickets the singer will sell—without regard for whether the recordings are played more than once, and whether the concert is disappointing.  Naturally the latter is what the show's music-industry producers care about.  In politics, the former is more important—for fair representation.

Top 16 week

Based on recent VoteFair results, it looks like each of these singers is at risk of a possible early “surprise” elimination next week:

• Lilly Scott
• Katelyn Epperly
• Andrew Garcia
• Michael Lynche

These singers are liked by lots of people, but they are not the first choice of enough voters (compared to some of the less-overall-popular singers).

Top 20 week

There were no big surprises this week.  Nervousness is still hurting some singers (until they get eliminated, ironically).

Top 24 week

The first week's results were surprising because the vocal performances were so weak that personality and appearances took on added importance.  Also, some people were probably voting strategically (rather than sincerely), and that produces unexpected results if their assumptions are incorrect regarding how others will be voting.

 

---

Like these American Idol polls?  Spread the word!  Also you are invited to voice your opinion — about VoteFair ranking (not about contestants) — on the Testimonials page.