Saturday, May 10, 2014


This is part II of an article about the PGSC score metric to quantify Idol placement:

*Part I contains the ranks of the Idol contestants on a season-by-season basis, based on the PGSC score. Think of this as the way the contestants should be ranked, assuming neutral ground and zero pimpage. It is also the way I would rank them.
*Part II ranks the Idol contestants based on PGSC score, according to genre.

I've integrated several features used in previous models to this PGSC score model. Check it out below:

Geographic location is standard: West, Midwest, South and Northeast. There are several contestants that have moved through several locations during their lifetime, but I used their most common location--i.e. Casey Abrams is from Idyllwild, CA rather than from Illinois, and Sam Woolf is from Florida as opposed to Michigan.

Age is separated into the three brackets which I think denotes young, middle age and older as far as Idol contestants are concerned. That's 0-19, 20-23 and 24+.

Genre is a little bit arbitrary, but I chose the genre the contestant presented themselves at during the show. Adam Lambert might have done pop after the show, but for the show's purposes, he was a rocker. There are some which melded genres or couldn't be fit into a specific genre, so for those I lumped them as "hybrid". The categories are hybrid, pop, soul, country, rock, and singer-songwriter.

Instead of using the 8-degree logistic regression in the previous model, I've decided to use a logarithmic regression here, because the general trends somewhat hold even if there is more noise (the trend is more easily visualized).

This pertains to the first graph:

The first thing immediately seen in the model: singer-songwriters have a very high chance of winning, even past a lowly PGSC score of 100, but especially at the 160+ range. They blow every other genre away.

Country and pop, in that order, are the next two genres represented, and then hybrid, rock and soul alternate. My PGSC model tends to hate soul contestants due to the low freshness scores, so the at lower scores they tend to make it farther than they should (i.e. Scott Savol, Jacob Lusk, LaKisha Jones). And there are major diminishing returns in soul--the higher PGSC score the contestant gets, the curve stabilizes into a lower value than the other genres. Rockers surprisingly do not fare all that well in this metric, and hybrids are also a mixed bag.

If were to separate the genre analysis by age bracket, at ages 0-19, country (Lauren and Scotty of AI10 help with this curve), soul, pop, rock, hybrid and country represents in that order. Considering that Jena is a hybrid in this age bracket, I'm not sure if this bode well for her chances here.

At the age 20-23 age bracket, singer-songwriter represents the most, then country, then hybrid, then rock, then soul, then pop represents, in that order.

At the age 24+ age bracket, pop represents the most (but we only have three data points--and Blake Lewis's placement helps skew the curve with a higher PGSC score), and then soul, rock, singer-songwriter, country and hybrid is last.

This pertains to the second graph:

The second graph generates the logarithmic curves with respect to location. If we were to analyze by location, the South and the Midwest yield the highest probability of advancing far, in that order, and then farther off is the West, and even farther off is the Northeast. In the 0-19 bracket, west does the best, then the south, then the northeast, and finally, the midwest (they have only produced contestants in this bracket that have always finished at 8th place).

At the 20-23 age bracket, midwest does the best, by far, and then the south, and then the west, and then the northeast (the northeast has always produced contestants in this bracket that finish in 9th place).

At the 24+ bracket, the midwest also dominates, and then the south and the west are in a bit of a dead heat here, and the northeast does the worst.

In terms of genres, for country, the midwest does the best, then the south, and then we only have one data point for the northeast and the west, but they are ranked in that order.

For hybrid genres, the midwest also represents the best, then the south, the northeast and then the west (but west and northeast are close).

For pop, it's the west, then the northeast, the south and then the midwest (midwest only has one data point).

For rock, it's the midwest, west, south and then the northeast.

For singer-songwriter, it is the west, midwest, and then the south. A key issue is that the west's regression curve is based off extrapolations off Brooke White's and MK Nobilette's data points, so this isn't entirely reliable. Alex Preston will provide a data point for the northeast soon on this.

For soul, the South is most reliable. What's interesting is that midwest soul contestants do really well with lower PGSC scores, but with higher ones, they perform really awfully, even lower than the west and northeast, which are in a statistical dead heat.

Now, here's a ranking of contestants based on genre and PGSC score, for reference:

1) Kree Harrison AI12  231 COUNTRY 20-23 S
2) Phil Stacey AI6 213 COUNTRY 24+ S
3) Lauren Alaina AI10 188 COUNTRY 0-19 S
4) Janelle Arthur AI12 185 COUNTRY 20-23 S
5) Scotty McCreery AI10 176 COUNTRY 0-19 S
6) Carrie Underwood AI4 170 COUNTRY 20-23 MW
7) Bucky Covington AI5 157 COUNTRY 24+ S
8) Casey James AI9 155 COUNTRY 24+ S
9) Kristy Lee Cook AI7 143 COUNTRY 24+ W
10) Skylar Laine AI11 136 COUNTRY 0-19 S
11) Danny Gokey AI8 110 COUNTRY 24+ MW
12) Dexter Roberts AI13 98 COUNTRY 20-23 S
13) Paul Jolley AI12 80 COUNTRY 20-23 S
14) Aaron Kelly AI9 73 COUNTRY 0-19 NE
15) Kellie Pickler AI5 57 COUNTRY 0-19 S
16) Michael Sarver AI8 35 COUNTRY 24+ S


1) Haley Reinhart AI10 209 HYBRID 20-23 MW
2) Jena Irene AI13 173 HYBRID 0-19 MW
3) Jessica Meuse AI13 169 HYBRID 20-23 S
4) Matt Giraud AI8 166 HYBRID 20-23 MW
5) Crystal Bowersox AI9 154 HYBRID 20-23 MW
6) Casey Abrams AI10 139 HYBRID 20-23 W
7) Elise Testone AI11 129 HYBRID 24+ S
8) Erika Van Pelt AI11 124 HYBRID 24+ NE
9) Majesty Rose AI13 124 HYBRID 20-23 S
10)Michael Lynche AI9 123 HYBRID 24+ S
11)Anthony Fedorov AI4 122 HYBRID 0-19 NE
12) Nadia Turner AI4 121 HYBRID 24+ S
13)Devin Velez AI12 100 HYBRID 0-19 MW
14)Scott MacIntyre AI8 81 HYBRID 20-23 W
15)CJ Harris AI13 74 HYBRID 20-23 S
16)John Stevens AI3 60 HYBRID 0-19 NE
17)Heejun Han AI11 52 HYBRID 20-23 NE
18)Megan Joy AI8 42 HYBRID 20-23 W
19)Naima Adedapo AI10 32 HYBRID 24+ MW


1) Angie Miller AI12 230 POP 0-19 NE
2) Blake Lewis AI6 192 POP 24+ W
3) Diana DeGarmo AI3 166 POP 0-19 S
4)Hollie Cavanagh AI11 165 POP 0-19 S
5)Jordin Sparks AI6 147 POP 0-19 W
6)David Archuleta AI7 130 POP 0-19 W
7)Chris Richardson AI6 127 POP 20-23 S
8)Jessica Sierra AI4 118 POP 0-19 S
9)Pia Toscano AI10 113 POP 20-23 NE
10)Jessica Sanchez AI11 106 POP 0-19 W
11)Katharine McPhee AI5 106 POP 20-23 W
12)Katie Stevens AI9 106 POP 0-19 NE
13)Ace Young AI5 102 POP 24+ MW
14)Stefano Langone AI10 94 POP 20-23 W
15)Didi Benami AI9 91 POP 20-23 W
16)Haley Scarnato AI6 81 POP 24+ S
17)Jasmine Trias AI3 59 POP 0-19 W
18)Sanjaya Malakar AI6 58 POP 0-19 W
19)Lisa Tucker AI5 54 POP 0-19 W
20)Lazaro Arbos AI12 51 POP 20-23 S
21)Camile Velasco AI3 19 POP 0-19 W


1) Adam Lambert AI8 240 ROCK 24+ W
2)Michael Johns AI7 230 ROCK 24+ S
3)James Durbin AI10 214 ROCK 20-23 W
4)Chris Daughtry AI5 200 ROCK 24+ S
5)David Cook AI7 187 ROCK 24+ MW
6)Amy Adams AI3 176 ROCK 24+ W
7)Caleb Johnson AI13 168 ROCK 20-23 S
8)Bo Bice AI4 159 ROCK 24+ S
9)Allison Iraheta AI8 155 ROCK 0-19 W
10)Constantine Maroulis AI4 147 ROCK 24+ NE
11)Gina Glocksen AI6 135 ROCK 20-23 MW
12)Carly Smithson AI7 123 ROCK 24+ W
13)Colton Dixon AI11 107 ROCK 20-23 S
14)Chris Sligh AI6 88 ROCK 24+ S
15)Siobhan Magnus AI9 73 ROCK 0-19 NE


1)Kris Allen AI8 216 SING-SONG 20-23 S
2)Lee DeWyze AI9 190 SING-SONG 20-23 MW
3)Alex Preston AI13 171 SING-SONG 20-23 NE
4)Phillip Phillips AI11 167 SING-SONG 20-23 S
5)Tim Urban AI9 109 SING-SONG 20-23 S
6)Sam Woolf AI13 106 SING-SONG 0-19 S
7)Jon Peter Lewis AI3 98 SING-SONG 24+ MW
8)Brooke White AI7 92 SING-SONG 24+ W
9)Paul McDonald AI10 82 SING-SONG 24+ S
10)Jason Castro AI7 68 SING-SONG 20-23 S
11) MK Nobilette AI13 40 SING-SONG 20-23 W


1) Anoop Desai AI8 224 SOUL 20-23 S
2)LaToya London AI3 223 SOUL 24+ W
3)Jennifer Hudson AI3 180 SOUL 20-23 MW
4)Fantasia Barrino AI3 169 SOUL 0-19 S
5)Candice Glover AI12 157 SOUL 20-23 S
6) George Huff AI3 151 SOUL 20-23 S
7)Nikko Smith AI4 151 SOUL 20-23 MW
8)Paris Bennett AI5 148 SOUL 0-19 S
9)Taylor Hicks AI5 148 SOUL 24+ S
10)Elliott Yamin AI5 144 SOUL 24+ S
11)Melinda Doolittle AI6 142 SOUL 24+ S
12)Burnell Taylor AI12 139 SOUL 0-19 S
13)Chikezie AI7 129 SOUL 20-23 W
14)Syesha Mercado AI7 124 SOUL 20-23 S
15)Mandisa AI5 121 SOUL 24+ S
16)DeAndre Brackensick AI11 118 SOUL 0-19 W
17)Malaya Watson AI13 112 SOUL 0-19 MW
18)Amber Holcomb AI12 111 SOUL 0-19 S
19)Vonzell Solomon AI4 111 SOUL 20-23 S
20)Joshua Ledet AI11 107 SOUL 0-19 S
21)LaKisha Jones AI6 76 SOUL 24+ MW
22)Andrew Garcia AI9 72 SOUL 24+ W
23)Anwar Robinson AI4 58 SOUL 24+ NE
24)Lil Rounds AI8 53 SOUL 20-23 S
25)Scott Savol AI4 52 SOUL 24+ MW
26)Jacob Lusk AI10 46 SOUL 20-23 W
27)Ramiele Malubay AI7 28 SOUL 20-23 S
N/A) Curtis Finch AI12 SOUL 24+ MW

Look at the company Alex Preston shares at the singer-songwriter slot: all those guys he is clustered with won their respective seasons. As a hybrid genre type (pop/rock), Jena doesn't have precedent for producing a winner. Caleb Johnson is lumped between David Cook and Bo Bice, so he can be winner or runner-up; however, unlike Alex and Jena, he isn't in the top five with respect to genre for PGSC score, so his style has been done better on the Idol stage before.

Play with the graphs below to learn more.

No comments:

Post a Comment