Top 7 projection and Twitter sentiment

Name WNTS Rating Popularity Order Safe Probability
Clark 80 47.0 5 98.0%
Tyanna 84 13.0 6 97.2%
Joey 50 10.0 4 69.5%
Quentin 60 7.0 3 61.0%
JAX 45 13.0 1 58.9%
Rayvon 43 2.0 7 58.1%
Nick 48 9.0 2 54.1%

Final update: Joey sticks at down 7 points from last week, but she’s still 4th most popular. Nevertheless, she falls out of safe territory. Nick falls a little, and overtakes Rayvon as the least safe (but not by much).

Update 2: Rayvon falls as Jax rises. Still no confident picks, but Rayvon is now the least safe.

Update: new numbers. Also, the fit in the Top 7 is somewhat better using data the Top 10 to the Top 6. In that case, gender becomes less significant, as does performance order. Jax rises one slot, as does Quentin, whose numbers were lagging a bit. Joey’s popularity is still down 7 points from last week, and Tyanna’s 6 points up.

Green = projected safe, yellow = too close to call

Hear me out.

First of all, nobody is even close to being called in the bottom 2 outside the margin of error. The model is not projecting Jax is in the bottom 2. Nope.


Historically speaking, Jax is not in good territory. First, she is a woman, and AI voters just don’t seem to like those. Her performance of Poker Face was not particularly well received (I thought it was marginally better than the rest of the internet did). And while she’s fairly popular on Votefair, she isn’t overwhelmingly so. She trails Joey, for instance.

What tips the scales somewhat against Jax is that she went first, which ain’t the best place to be. Now, you could argue, and it’s not necessarily wrong, that performance order isn’t as important now that you can vote right away on your smart phone. Ok, sure. But, I know that I wait til the end of the show, and that means other people might too. And they might forget to vote for Jax. (Full Disclosure: Since the Top 12, I have only voted for Joey.)

Twitter isn’t overwhelmingly positive on Jax either, to be honest:

TwitterSentiment2015-04-09Nick has more negatives than Jax. But Jax has almost as many as the controversial Joey Cook.

In any case,  nearly equally likely to be in the bottom 2 (yet again) is Rayvon. He’s got a great Twitter organization pulling for him, and both weeks they have made #SaveRayvon a trending hashtag. That’s something of a pickle for the Idol producers: what if the Idol save pits Rayvon as a weekly gladiator against another contestant, and like Maximus he consistently comes out ahead? Now that would be entertaining.

Things aren’t great for Quentin, although he is still better than fair to be safe. Clark, Tyanna, and Joey are all projected safe for now.

Bookmark the permalink.
  • Kent R

    From looking at last weeks predictions vs results… It appears your model could be strengthened by finding a way to factor in the relative disparagy between WNTS rating & popularity. For instance- last week Rayvon’s 68 WNTS score was within 3 points of 2nd place (Joey’s 71)… yet his popularity was 0%. Such a big gap between WNTS score and popularity likely indicates the extend to which Rayvon is failing to connect to viewers… and connecting with viewers likely impacts votes as much as song order or maybe even WNTS score. Just an idea- but it would be interesting to see if “connecting to viewers” could somehow be quantified through comparing those 2 numbers.

    Also- I’m not sure if twitter sentiment factors into the formula… put a potential “false read” could result from an increase in the amount of discussion regarding song choice. For example- Clark & Nick seemed to fall back a fair amount in this indication of late (there is a very good chance this is influenced to some extent by fans of theirs that are unhappy with their song choices- which garners a lot of discussion. Where the false read come in is the more a fan is pulling for a contestant (who’s song choices they are unhappy with)- the stronger their sentiment (read as negative) of letting the contestant know in hopes of them picking smarter songs in the future. But these supporters would still be voting for the contestant- if not more so to make sure they are safe. Sort of round about explanation- but you probably get the underlying point regarding “sentiment bots.”