Prediction Analysis Decade 2010s

Throughout the last 6 weeks, 100 Warm Tunas has been aggregating your votes posted to popular social networks to provide the internet's most accurate prediction of triple j's Hottest 100 of the Decade.

Summary

  • We collected 3,181 entries.
  • We tallied 29,123 votes across these entries.
  • triple j counted approximately 1,800,000 votes.
  • Therefore, we collected a sample of 1.62%.
  • 11.2% of votes were collected via Instagram direct message.
  • We did not successfully predict #1
  • We predicted 6 out of the top 10 songs.
  • We predicted 25 out of the top 30 songs.
  • We predicted 79 out of the 100 songs played in the countdown.

Evidently the amount of data collected was less than the Hottest 100 of 2019 however, our prediction still performed better when looking at the overall count of predicted songs in the top 100.

Technical Analysis

Similar to the Hottest 100 of 2019, we were not able to predict the outright winner of the countdown ,however we did successfully predict 25 of the songs to rank in the top 30.

Lets take a look at the top 10 of the official countdown and match it up with the predicted places in 100 Warm Tunas:

Artist Title ABC Rank Tunas Rank Difference
Tame Impala The Less I Know The Better 1 5 4
Gotye Somebody That I Used To Know {Ft. Kimbra} 2 6 4
Arctic Monkeys Do I Wanna Know? 3 3 0
Violent Soho Covered In Chrome 4 1 3
RUFUS DU SOL Innerbloom 5 4 1
Gang Of Youths Magnolia 6 2 4
Foster the People Pumped Up Kicks 7 13 6
Flume Never Be Like You {Ft. Kai} 8 15 7
Angus & Julia Stone Big Jet Plane 9 21 12
Matt Corby Brother 10 11 1

Lets pull apart this table a bit more and grab some statistics about how we did with our prediction overall:

Predicted Out Of Top N Percentage
6 10 60.0%
14 20 70.0%
25 30 83.3%
32 40 80.0%
39 50 78.0%
50 60 83.3%
58 70 82.9%
64 80 80.0%
72 90 80.0%
79 100 79.0%
155 200 77.5%

From the above data, we can see that:

  • The average error for the top ten ranks was 4.2 positions.
  • Warm Tunas predicted 6 out of the top 10 songs.
  • Warm Tunas predicted 14 out of the top 20 songs.
  • Warm Tunas predicted 25 out of the top 30 songs.
  • Warm Tunas predicted 79 out of the 100 songs played in the countdown.

To compare Warm Tuna's predictions vs actual rankings, a scatter plot has been provided below. We can see as we get closer to rank 1, the 100 Warm Tunas prediction gets better and converges upon the actual rankings played out on the day:

Vote counts over time

Throughout the voting period, we observed a few key changes to the top 10, observing Magnolia and Do I Wanna Know competing for 2nd place in the prediction, as well as The Less I Know The Better and Innerbloom competing for 4th place in the prediction. This is potentially the most movement we have seen in a prediction since starting the project.

Vote counts by social network

The following table provides a break-down of vote counts per social network. There are some strong biases depending on which sources are included when producing rankings for the prediction:

Song Title Insta. Feed Insta. Story Twitter Reddit Insta. DM Sum Total
Covered In Chrome - Violent Soho 314 58 20 27 67 486 479
Magnolia - Gang Of Youths 212 56 29 30 81 408 394
Do I Wanna Know? - Arctic Monkeys 183 56 43 23 57 362 357
Innerbloom - RUFUS DU SOL 185 43 31 31 61 351 344
The Less I Know The Better - Tame Impala 120 53 38 34 76 321 315
Somebody That I Used To Know {Ft. Kimbra} - Gotye 161 45 38 13 36 293 287
It's Nice To Be Alive - Ball Park Music 162 39 26 18 37 282 270
Breezeblocks - Alt-J 164 29 15 15 39 262 254
Midnight City - M83 141 30 22 23 27 243 240
Runaway {ft. Pusha T} - Kanye West 121 22 34 21 40 238 237
Brother - Matt Corby 129 48 20 3 34 234 233
Royals - Lorde 116 38 31 14 23 222 216
Pumped Up Kicks - Foster the People 108 45 22 14 32 221 214
Let It Happen - Tame Impala 87 27 26 45 36 221 213
Never Be Like You {Ft. Kai} - Flume 125 45 16 9 27 222 213
Clair De Lune {Ft. Christine Hoberg} - Flight Facilities 119 31 18 7 23 198 189
Let Me Down Easy - Gang Of Youths 86 33 21 17 30 187 183
King Kunta - Kendrick Lamar 94 25 20 15 24 178 175
Tokyo (Vampires & Wolves) - The Wombats 86 30 18 17 32 183 174
Video Games - Lana Del Rey 100 28 23 10 15 176 171

For example, if we only count votes from Instagram Stories, we are able to predict the 10 songs to place in the top 10 (ignoring order). By counting just Instagram Direct Messages and Twitter submissions, we are able to predict 1st place.

Finally, by excluding data collected from Instagram Feed, but using all other sources (Instagram DM, Instagram Story, Twitter, Reddit), we are able to provide a reasonably accurate prediction of the top 10:

Artist Title ABC Rank Tunas Rank Difference
Tame Impala The Less I Know The Better 1 1 0
Gotye Somebody That I Used To Know {Ft. Kimbra} 2 6 4
Arctic Monkeys Do I Wanna Know? 3 3 0
Violent Soho Covered In Chrome 4 4 0
RUFUS DU SOL Innerbloom 5 5 0
Gang Of Youths Magnolia 6 2 4
Foster the People Pumped Up Kicks 7 10 3
Flume Never Be Like You {Ft. Kai} 8 16 8
Angus & Julia Stone Big Jet Plane 9 23 14
Matt Corby Brother 10 11 1

This trend continues throughout the result set:

Predicted Out Of Top N Percentage
7 10 70.0%
14 20 70.0%
27 30 90.0%
33 40 82.5%
39 50 78.0%
47 60 78.3%
54 70 77.1%
61 80 76.2%
73 90 81.1%
80 100 80.0%
154 200 77.0%

The accuracy of the top 10 is at 70%, the top 30 at 90%, and the top 100 at 80%. Whilst the accuracy across the top 200 stays almost constant these are some clear improvements over the results when Instagram feed images are included.

This trend is really interesting because counting just these sources means only 12,247 votes across 1,352 entries are counted toward the tally (compared to 29,123 across 3,181 entries). Perhaps this bias is an indication that it might be time to pull the plug on collecting data from the Instagram Feed and rely on these newer, more accurate sources?

Social Network Breakdown

We are now collecting votes from a handful of different sources. Here is the break down of how many votes were counted across the different sources:

Source Num Entries Num Votes Votes Per Entry
feed 1888 17383 9.2071
story 490 4330 8.8367
twitter 357 3262 9.1373
reddit_submission_comment 169 1482 8.7692
direct_message 356 3344 9.3933
Sum 3260 29801 9.1414
Sum (after de-duplication) 3181 29123 9.1553

Vote Submissions

It comes as no surprise that the most popular day for people to share their votes was on the day of voting opened with 442 entries collected, followed by the day voting closed with 391 entries collected.

Wrap Up

Whilst we didn’t pick the outright winner, we did successfully predict 25 of the 30 of the songs which landed in the top 30, as well as 79 of the top 100. As mentioned in the 2019 countdown wrap-up, with the shrinking amount of data and general biases across different voting sources, it’s getting harder and harder to pick the outright winner with certainty.

We'd like to thank everyone who shared their votes, visited our site or interacted with us on Instagram. Without our audience, 100 Warm Tunas wouldn't be possible.

We are looking forward to running the prediction again in December for triple j’s Hottest 100 of 2020. Time to start collating that shortlist now!