The Wisdom of the Crowd: RESULTS


The test was in fact 4 different tests. Each time a visitor reached the page, a random board was selected from this four:

     

The number of lines/sticks in each test was:

Test1: 348
Test2: 2584
Test3: 7584
Test4: 25978

These images were constructed using Processing, with this script.

#Participants: 705  (test1: 183, test2: 177, test3: 172, test4: 173)

This is the average of the responses, for each test. Is the Crowd so Wise?:

Test1: 441
Test2: 5662
Test3: 395555489
Test4: 53314289374716800

Obviously, it is not.

What about the medians?:  250, 1000, 3137, 9000

Better, yet also quite far the right values.

The first problem is that there are some extreme outliers, joke or fake responses, like: 1, 8, 1234567890, 67890077889 or 9223372036854770000.

After using the Quartile or Fourth-Spread method and removing some outliers, the results are not very satisfying either:

Means:  243, 1073, 3238, 10293
Medians: 216900, 25005000

The crowd loses again. Maybe this was not a very good test to check its wisdom... :/

Now, what about the individuals?

Dichloroethane on Reddit did an impressive estimation for test1 and test4 of 350 and 25737. It's amazing, I recommend you to read the post.

- These are the best entries. I just saved the IP address for each response to prevent multiple answers from the same person, so an approximate location is the only I know about them:

Test 1: 348 (exact!), from Tampa, Florida, US 
Test 2: 2580 (-4), from Norway, Michigan/Iowa, US
Test 3: 7600 (+16), from England (Manchester or Reading)
Test 4: 26745 (+767), from New Zealand (Auckland or Wellington)

Congrats! :) Dichloroethane was more accurate in the test4 (-241) though, but I don't have registered his entry in the database..

Anyway, thanks to all participants, redditors, twitters, friends & relatives who helped with this little experiment!

Thank you, Crowd!

PD: I leave you here all the entries.


Comments
  

 

 

← Return to blog