I analyzed 67 years of LEGO data

Comments

• bobabricks Member Posts: 1,842

  July 2016

  Very interesting. Looks like the first set to break 1,000 pieces was a teacher resource set, ha ha.

  0
• ricecake Member Posts: 878

  July 2016

  Very cool!
  

  0
• chuckp Member Posts: 684

  July 2016

  Nice work! The color palette by theme was very interesting. Gray is really taking over, even the Friends line wasn't spared. 

  0
• LusiferSam Member Posts: 574

  July 2016

  bobabricks said:

  Very interesting. Looks like the first set to break 1,000 pieces was a teacher resource set, ha ha.

  I believe it is correct that first to break 1,000 pieces was an educational set, but the not the one shown here.  7100 is more twice as large as 1034 and predates it by 20+ years.
  
  Very interesting overall.  The graying trend is interesting.  The minor color increase is pretty self evident, but is fun to have a visual of.  I like the color palette by theme.  Maybe the top five colors should be shown as three seems like it might be too few.  The node network is the most interesting to me.  There must be something about the methodology I don't understand because the results look slightly odd.
  
  I'm going to be critical of the use of Rebrickable as the set database.  They are one the worst databases when it comes to pre-1972 sets.  They basically have no Samsonite sets, which would have a major impact on what you see in the 60's.  Set inventories from this period are notoriously incorrect or incomplete, as such I'm not sure they really be relied on.  I'm not even going to get into regional and year differences of the 50's sets.  And one finally nitpicking thought, 2016-1950=66, not 67.  Where are the 1949 sets?
  

  0
• ricecake Member Posts: 878

  July 2016

  LusiferSam said:

  And one finally nitpicking thought, 2016-1950=66, not 67.  Where are the 1949 sets?
  

  2016 - 2015 = 1, but includes 2 years' worth of data (2015, 2016), then extrapolate.
  
  If you are including the endpoints, then the number of items in a range is (end - start + 1).
  

  

  1
• LusiferSam Member Posts: 574

  July 2016

  ^
  That's what I get for writing too quick, very red faced.  Yes, you're quite correct.  But the point still reminds:  Where are the 1949 sets?
  

  0
• KingAlanI Member Posts: 2,355

  July 2016

  ricecake said:

  LusiferSam said:

  And one finally nitpicking thought, 2016-1950=66, not 67.  Where are the 1949 sets?
  

  2016 - 2015 = 1, but includes 2 years' worth of data (2015, 2016), then extrapolate.
  
  If you are including the endpoints, then the number of items in a range is (end - start + 1).
  

  It's called a fencepost error, because a fence with X sections needs X+1 fenceposts.
  
  the article is both pretty and insightful.
  
  

  

  1
• KingAlanI Member Posts: 2,355

  July 2016

  ricecake said:

  LusiferSam said:

  And one finally nitpicking thought, 2016-1950=66, not 67.  Where are the 1949 sets?
  

  2016 - 2015 = 1, but includes 2 years' worth of data (2015, 2016), then extrapolate.
  
  If you are including the endpoints, then the number of items in a range is (end - start + 1).
  

  It's called a fencepost error, because a fence with X sections needs X+1 fenceposts.
  
  the article is both pretty and insightful.
  
  

  0
• Istokg Member Posts: 2,366

  July 2016

  I have to second what @LusiferSam said.... the input data to this very interesting research for sets prior to 1972 is incomplete.  From 1954-72 the majority of sets had parts in white and red... pretty equally divided.  No idea why there was a spike for 1960... maybe because only a few sets were introduced that year.  1964 never had blue as the predominant color... that was still mainly a red/white year.  Also black was not introduced as a set color until 1961 (only available as very rare PAB colors from 1950-56).   Ditto for Gray.... which was introduced as plates in 1962 and bricks in 1974 (only 1 set... 1650).
  

  0
• LusiferSam Member Posts: 574

  July 2016

  @Istokg I think the 1960 spike is pretty easy to explain.  It has to do with both the small number of sets released and the fact that there is a single very large set in the data (700K). 
  

  0