Top 4 redux prediction (final)

Name Song DialIdol VoteFair
Candice Glover When I Was Your Man/
You’ve Changed
73 3.816 22 0.565
Kree Harrison See You Again/
Stormy Weather
57 4.837 26 0.485
Angie Miller Diamonds/
Someone To Watch Over Me
49 2.944 44 0.479
Amber Holcomb Just Give Me A Reason/
My Funny Valentine
41.5 7.175 9 0.472

Final update 2:50 PM EST. No ranking changes.

Updated 10:46 AM EST. Angie gained a bit, and Kree fell, so they swapped positions.


The methodology for the finals model is described here. The model is 87% accurate on ranking within a margin of error of +/- 3%. Probabilities being what they are, somebody with a not-safe probability of just 0.25 will be in the bottom 2 one out of four times. Please do not comment that the numbers are wrong. They are probabilities, not certainties or even claims. Do not gamble based on these numbers.

Names in green are predicted safe. Names in red are the most likely bottom 2. Names in yellow are undecided. Any person not in green is not considered safe by the model. The most probable bottom 2 is Candice and Angie Kree. However, anybody on the list being in the bottom 2 would not be shocking. [updated: Angie no longer 2nd most probable, Kree is. It’s still basically a tie.]

Sorry to do this folks, but there’s no firm prediction possible based on the preliminary numbers. Amber, Kree, and Angie are running what amounts to a dead heat as far as the model can tell, and so it can’t make any safe calls. Nobody is far-and-away winning this race.

However, for the second week in a row, the model thinks Candice is the most at risk of elimination. She was in the bottom 2 last week, so I can certainly see it happening. Amber would be the one my gut would pick, but the model chooses her as the least likely, mainly on the strength of her Dialidol number, which is significant. If you’ve read the methodology post, you know that Dialidol becomes a stronger indicator as the year goes on, whereas WNTS and Votefair become less predictive with each passing week.

As perverse as it might seem, this prediction is independent of the Top 3 Tracker, which takes a longer view of the season. In a sense, the T3T is not sensitive to week-to-week shocks. As such, it would say that Amber is gone. However, based just on tonight’s numbers, Candice is.

Note that I haven’t a clue how I would take into account the two weeks’ votes being pooled. These numbers are based purely on tonight; since Candice was the highest probability in both weeks, that’s unambiguous that she would be the most likely.

Two things to ponder. 1. Amber was first on Dialidol last week and this week, and was nonetheless in the bottom 2 last week. 2. Angie was last third on Dialidol last week, but we know she was in the top 2. If Dialidol is very different from previous years, of course, the model will be way off. However, the model was correct in calling Angie safe last week, so the other factors do matter. Take it for what it’s worth.

Bookmark the permalink.
  • bose_sucks

    Did you take into account they said the scores from last week and this week would be combined? Candice and Amber were in the bottom 2 last week.

  • bose_sucks

    Now that you can cast 50 votes on, Dial Idol has lost its usefulness, just look at their predictions this season, no one was in the red at any time.

  • Dan

    I agree- you have to ditch Dial Idol or at least reduce the contribution to the prediction. There aren’t enough people contributing to those numbers like they used to — just look back on previous seasons. So far the most accurate predictor this season has been the FunOfficePools stats. Haven’t been wrong any week in the top 10.

  • Mudflapman

    Dialidol hasn’t been accurate for a while because of other avenues to vote. If you must include it, do not use the DI Score, instead, use the RAW Numbers

    As you can see, it has been more accurate (since last season)