Tuesday, April 13, 2010

What if the Top 100 Children's Novels Poll had....

While tabulating all the data from Fuse#8's Top 100 Children's Novels Poll I was fasinated by the way some titles accrued points by earning lots of 10th, 8th and 9th place votes while others earned their spots by earning less votes of higher value. There is an argument to be made that the more votes a title recieves, regardless of the value of the vote, signifies a certain universal acknowledgment of excellence. Alternately those books which did not earn as many votes but were placed closer to first on the top tens in which they were selected signifies a deeper affection felt by those who chose to recognize them at all. For example: Wilson Rawls' Where the Red Fern Grows was placed 46th on the list with only 11 votes. Other titles with an equal number of votes ranged between 81st and 60th. What allowed Where the Red Fern Grows to surpass all of the other titles with 11 votes and many titles with votes in the high teens? Well, of Where the Red Fern Grows' 11 votes an almost half (5) were 1st place votes! Those who placed Where the Red Fern Grows on their top 10s placed it at the top or near the top in all but one instance. The vote tally was: 5 firsts, 1 second, 2 thirds, 2 fourths and 1 tenth place vote. 91% of the title's votes were in the top 5 and 73% were in the top 3. Where the Red Fern Grows earned 90 points from its 11 votes. (5*10)+(1*9)+(2*8)+(2*7)+(1*1) = 90
I use a measure I termed 'Average Vote Rank' to assign a measure to a books "belovedness" regardless to how many votes it received. A book which only receives first place votes will have an average vote rank of 1, and a book with only tenth place votes (regardless of how many 10th place votes) will have an average vote rank of 10. Where the Red Fern Grows has an Average Vote Rank of 2.82 which is the lowest of any book in the top 100. Charlotte's Web the #1 book has an Average Vote Rank of 3.56 (the 7th best average rank on the list).

Rebecca Stead's When You Reach Me (#39) serves as a great counterexample. Here is a title with 24 votes but no first or second place votes, not surprising when you consider its newness (in ten years I'm betting it's recognized along side of Holes as the very best and most loved of the modern Newberys). When You Reach Me was the first book to appear in the countdown with over 20 votes but because so many of the votes (50% of WYRM's votes where 10th-7th place votes) had little value, the title only earned 107 points. Wizard of Oz also earn 107 points but did so with nine less votes. When You Reach Me's average vote rank is 6.54 while Wizard of Oz's average vote rank is 3.87.

Average Vote Rank is a valuable metric but because a book with one feverish supporter can place higher than a book with over 100 only slightly less feverish supporters the value is not useful in creating an entire ranking.


I played around with a number of different methods and ended up creating 5 alternative rankings plus a ranking in which I average all 5 alternates and the official/actual ranking.

Below I will explain each ranking methodology and highlight some of the titles which benefited and those which lost the most ground with each method.

The first modified ranking is also the simplest. In Total Votes Ranking I simply gave each title one point for each vote regardless of the value of the vote. With this method titles 1-11 were unaffected, The Hobbit and Bridge to Terabithia swapped spots and Hatchet jumped all the way from 26st to 18th. Where the Red Fern Grows suffered greatly dropping down to 61st but When You Reach Me gained some ground and just squeaked into the top 25.

Total Votes Ranking

Rank Points Title
1 101 Charlotte's Web
2 87 Harry Potter and the Philosopher's Stone
3 87 A Wrinkle in Time
4 72 The Lion, the Witch and the Wardrobe
5 70 From the Mixed Up Files of Mrs. Basil E. Frankweiler
6 68 Holes
7 59 The Giver
8 58 The Secret Garden
9 55 Anne of Green Gables
10 54 The Phantom Tollbooth
11 41 The Westing Game
12 38 Bridge to Terabithia
13 33 The Hobbit
14 32 Harriet the Spy
15 31 Charlie and the Chocolate Factory
16 31 Because of Winn-Dixie
17 31 Harry Potter and the Prisoner of Azkaban
18 28 Hatchet
19 27 Percy Jackson and the Olympians: The Lightning Thief
20 26 James and the Giant Peach
21 26 The Tale of Despereaux
22 26 Tuck Everlasting
23 26 Matilda
24 26 Maniac Magee
25 24 When You Reach Me


The total votes method igonores the value of vote placement completely. In my second reranking method I go to the opposite end of the spectrum. In this ranking I ask the question "What if Fuse#8 only allowed voters to select 3 titles instead of 10?" Here I ignored any votes that weren't 1st, 2nd or 3rd place votes. I then used the traditional values; 1st place votes earned a title 10 points, 2nd place - 9 points and 3rd place - 8 points. When I re-tabulated the points I go the following new top 25 (26 actually as Winnie-the-Pooh and James and the Giant Peach tied for 25th place).

Top 3 only Ranking

Rank Points Title
1 577 Charlotte's Web
2 386 A Wrinkle in Time
3 242 Anne of Green Gables
4 225 The Giver
5 213 The Lion, the Witch and the Wardrobe
6 209 Holes
7 208 Harry Potter and the Philosopher's Stone
8 202 The Secret Garden
9 190 From the Mixed Up Files of Mrs. Basil E. Frankweiler
10 163 The Westing Game
11 137 The Hobbit
12 115 The Phantom Tollbooth
13 112 Harry Potter and the Prisoner of Azkaban
14 105 Bridge to Terabithia
16 101 Harry Potter and the Deathly Hallows
15 101 Maniac Magee
17 96 Harriet the Spy
18 85 Alice's Adventures in Wonderland/Through the Looking Glass
19 84 Matilda
20 83 Because of Winn-Dixie
21 79 Little Women
22 77 The Wonderful Wizard of Oz
23 75 Where the Red Fern Grows
24 73 Little House in the Big Woods
25 71 Winnie-the-Pooh
25 71 Charlie and the Chocolate Factory


As you can see Anne of Green Gables benefits tremendously from this re-ranking. You will also notice that Tuck Everlasting, The Tale of Despereaux and Percy Jackson have fallen out of the top 25. They are replaced with Where the Red Fern Grows, The Wonderful Wizard of Oz and Alice's Adventures in Wonderland. It's interesting that two recent kid favorites (Despereaux and Percy Jackson) are push out in favor of some all time classics.



The next method is almost identical to the top 3 votes method except that here I imagine Fuse#8 requiring voters to submit a top 5. So I only ignore votes for 10th-6th place, thus the Top 5 votes method.

Here is the top 25 generated from this method

Top 5 only Ranking
Rank Points Title
1 668 Charlotte's Web
2 491 A Wrinkle in Time
3 443 Harry Potter and the Philosopher's Stone
4 357 The Lion, the Witch and the Wardrobe
5 321 Anne of Green Gables
6 315 From the Mixed Up Files of Mrs. Basil E. Frankweiler
7 304 The Giver
8 300 The Secret Garden
9 294 Holes
10 202 The Westing Game
11 200 The Phantom Tollbooth
12 175 The Hobbit
13 165 Harry Potter and the Prisoner of Azkaban
14 145 Bridge to Terabithia
15 136 Because of Winn-Dixie
16 135 Harriet the Spy
17 132 Maniac Magee
18 121 Harry Potter and the Deathly Hallows
19 113 Little Women
20 110 Matilda
21 105 Alice's Adventures in Wonderland/Through the Looking Glass
22 104 Little House in the Big Woods
23 98 The Tale of Despereaux
24 98 Mrs. Frisby and the Rats of NIMH
25 98 The Wonderful Wizard of Oz

Not too different from the Top 3 only ranking except that Mrs. Frisby and the Rats of NIMH now makes the top 25 instead of Where the Red Fern Grows.


Not totally satisfied with ignoring votes but also wanting to give greater weight to books the voters placed indicated they thought more highly of, I began researching different voting methodologies. Thanks to Nate Silver's wonderful politics blog, I discovered a method of vote counting that gives significantly more weight to votes closer to the top pick. In this method I look at all the votes cast but instead of giving 10 points to first place votes, 9 place to second, 8 to third, etc, I assigned 1 point to first place votes, 0.5 points to 2nd place votes, .33 to third place votes etc. In this method a second place vote is worth 1/2 a first place vote. See the chart below for a full points breakdown:

Vote points as fraction points as decimal

1st 1/1 1.0
2nd 1/2 0.5
3rd 1/3 0.333
4th 1/4 0.25
5th 1/5 0.20
6th 1/6 0.167
7th 1/7 0.143
8th 1/8 0.125
9th 1/9 0.111
1oth 1/10 0.1

The advantage of this method is that each point value is in relation to a first place vote's value. A title would need 8 eighth place votes to equal the same as a single first place vote (0.125 x 8 = 1.0) Mathematicians might call this the N-1 or inverse function point value system. When scores are re-tabulated with this method no votes are ignored but 1st place votes are given premium importance. Notice there is a huge difference between 1st and 2nd place vote values but very little difference between 9th and 10th place values. Below is the Top 25 using this scoring method.

N-1 point system Ranking

Rank Points Title
1 50.9742084 Charlotte's Web
2 36.24579455 A Wrinkle in Time
3 29.98006477 Harry Potter and the Philosopher's Stone
4 24.90144611 The Lion, the Witch and the Wardrobe
5 23.23439811 The Giver
6 23.14139644 Anne of Green Gables
7 21.9169221 From the Mixed Up Files of Mrs. Basil E. Frankweiler
8 21.57674688 Holes
9 21.54171433 The Secret Garden
10 14.91163544 The Westing Game
11 13.80227022 The Phantom Tollbooth
12 12.92379355 The Hobbit
13 10.39182644 Bridge to Terabithia
14 9.93138233 Because of Winn-Dixie
15 9.44330211 Maniac Magee
16 9.40441322 Harry Potter and the Prisoner of Azkaban
17 8.99268344 Matilda
18 8.88727055 Harriet the Spy
19 8.84079322 Harry Potter and the Deathly Hallows
20 7.75771533 Tuck Everlasting
21 7.72801744 Charlie and the Chocolate Factory
22 7.715857 The Wonderful Wizard of Oz
23 7.62996911 A Little Princess
24 7.56474722 Little House in the Big Woods
25 7.23316044 Hatchet

As you can see the top 10 is not changed much in this method either, though Westing Game does sneak in front of The Phantom Tollbooth. Westing game had 13 less votes than Tollbooth but 4 more 1st place votes, in the
N-1
calculation The Westing Game's 1st place votes were worth more than the sum total of Tollbooth's lesser valued votes. In the original/standard ranking Tollbooth earned 43 more points than The Westing Game but Westing Games average vote ranking (5.0) was better than Tollbooth's (5.4) the
N-1
method therefore accounts for both the number of votes and the average vote ranking. I really like this method!

The final ranking system I employed combined two of the above systems. In this method I used only 1st - 5th place votes and then scored the votes with the
N-1
method. Below are the results of the rescoring.

N-1 point system Top 5 Only Ranking

Rank Points Title
1 47.61 Charlotte's Web
2 32.28 A Wrinkle in Time
3 25.95 Harry Potter and the Philosopher's Stone
4 21.15 The Lion, the Witch and the Wardrobe
5 21.07 Anne of Green Gables
6 20.24 The Giver
7 18.71 The Secret Garden
8 17.99 From the Mixed Up Files of Mrs. Basil E. Frankweiler
9 17.43 Holes
10 12.83 The Westing Game
11 11.45 The Hobbit
12 10.1 The Phantom Tollbooth
13 8.21 Maniac Magee
14 8.19 Harry Potter and the Deathly Hallows
15 8.16 Harry Potter and the Prisoner of Azkaban
16 8.01 Because of Winn-Dixie
17 7.88 Bridge to Terabithia
18 7.25 The Wonderful Wizard of Oz
19 7.23 Matilda
20 7 Harriet the Spy
21 6.66 Where the Red Fern Grows
22 6.33 A Little Princess
23 6.21 Little House in the Big Woods
24 6.02 Little Women
25 5.88 Tuck Everlasting

Again not too much variation in this method but what you don't see is that with this scoring Percy Jackson falls all the way to number 41! Not to anger Percy Jackson fans, but can we all agree its not one of the top 25 children's novels? Also of note is the way Harry Potters 3 & 7 find themselves side by side in with this scoring. In the original results Azkaban was 14 and Deathly Hollows was 24.


Finally I averaged all the above rankings as well as the original/standard rankings to create one additional ranking method that encompasses all the methods. The resulting top 25 is below. The title's original rank is in parentheses. Titles in green improved, titles in red dropped and titles in black remained constant.

Avg Rank Title
1 Charlotte's Web (1)
2 A Wrinkle in Time (2)
3 Harry Potter and the Philosopher's Stone (3)
4 The Lion, the Witch and the Wardrobe (4)
5 The Giver (7)
6 Anne of Green Gables (9)
7 From the Mixed Up Files of Mrs. Basil E. Frankweiler (5)
8 Holes (6)
9 The Secret Garden (8)
10 The Westing Game (11)
11 The Phantom Tollbooth (10)
12 The Hobbit (12)
13 Bridge to Terabithia (13)
14 Harry Potter and the Prisoner of Azkaban (14)
15 Because of Winn-Dixie (15)
16 Maniac Magee (17)
17 Harriet the Spy (16)
18 Matilda (18)
19 Harry Potter and the Deathly Hallows (24)
20 Little House in the Big Woods (23)
21 Tuck Everlasting (20)
22 Little Women (25)
23 Charlie and the Chocolate Factory (19)
24 Alice's Adventures in Wonderland/Through the Looking Glass (27)
25 A Little Princess (28)


When I examined all the titles and how their rankings changed with the different methods I noticed that some titles remained fairly constant regardless of the method and some titles moved up and down the rankings quite dramatically.

Only two titles maintained their spots regardless of the system, Charlotte's Web and A Wrinkle in Time. The title with the greatest variance is When You Reach Me, which climbed all the way to 25th when only counting votes and dropped as low as 77th when only scoring with top 3 votes. When You Reach Me's average was 46th place. Other highly variable titles included Island of the Blue Dolphin, The High King, On the Banks of Plum Creek and All-of-a-Kind Family. The average variance for the titles in the top 100 is 20 spots. The average variance for top 25 titles is 6.24 spots. The average variance for titles in the top 10 is a mere 2.5 spots.

For 12 titles the lowest
[closest to #1]
ranking the title achieved was its actually ranking.
Meaning these titles either stayed at their original spot or dropped spots with every modification.
These titles were: Charlotte's Web, A Wrinkle in Time, The Lion, the Witch and the Wardrobe, The Phantom Tollbooth, Holes, Harry Potter #1, From the Mixed Up, Dark Rising, Harry Potter #2, Tuck Everlasting, Despereaux, and Ramona Quimby, Age 8.

Twelve titles also achieved their highest [closest to #100] ranking with the original ranking. Meaning these titles either stayed at their original spot or improved with every modification. These titles were:
Charlotte's Web, A Wrinkle in Time, The Westing Game, The Giver, Anne of Green Gables, Are You There God? It's Me Margret, The Graveyard Book, Edward Tulane, Children of Green Knowe, Sideway Stories, The Egypt Game, and The Witches

Below is an embedded google spreadsheet that shows all 100 titles and all the different rankings.

Enjoy!





3 comments:

Miriam said...

Oh, Eric, I should've figured you for a 538 reader. ::grins::

As always, thank you!

Jed said...

I've been speculating on what would have happened if the voting had been for top-20. Specifically, I wonder whether a book like Stuart Little would have made the top-100 or higher with a lot of votes in the 11-20 range. (People not wanting to vote for two Whites and the incredible consensus for Charlotte's Web kept it off, it seems; but an extra ten might have made it a lot easier to vote for multiples by one author). No way to know, but interesting to guess!

SchrefflerFamily said...

I know I nearly had a top 20 before I managed to pare it down to 10. And even there, I was not duplicating authors just to get it down that low!