Colors - How They've Changed!

davee123

About a year ago, I was curious to see how the colors in LEGO had changed over the years.  That is, as a kid, I remember grey bricks being pretty rare (I had a bunch of grey plate from space ships, but only a few bricks), until 1984 when the castle sets came out.  And now, grey is everywhere.  Similarly, dark grey and brown were only occasionally used, until they started becoming more common in the late 1990s and early 2000s.

So, I did what I always seem to do-- I wrote a program to do some analysis.  Using BrickLink inventories, and ignoring things like DUPLO, Fabuland, and Dacta/Education, I tried to do a breakdown by year from 1975 - present (then 2014).  Here's what I found:



I haven't tried doing an updated one yet, but I may give it a go.  I know there are a few colors like Dark Tan which are quickly becoming much more mainstream.  And there are enough colors now such that the pie graphs probably aren't the best way to represent these!

DaveE

Roberto

Very interesting, congrats. It seems that only grey, dark grey and black rapresent almost the half of the total production. Good work.

CCC

It would be interesting to see how that data changes based on parts sold. I can image one big SW ship adding a lot to greys, but the number of small creator sets even though sold in large numbers add not very much to the brighter / primary colours.

madforLEGO

Heck just look at the number of shade changes in one color type in a LEGO set these days :-p

Robbzy88

That explains why I've had such difficulty finding white pieces since returning from my dark ages, And not going to the hassle of making PaB orders.

jason1976

Interesting stuff!

One thing to consider though is the amount of pieces and the size of them in sets, for example I have #375 (aka the yellow castle) from 1981, which has a heck of a lot of smaller yellow pieces (1x1, 1x2). Compare this to #6073 (from 1984) which, while mainly grey, contains fewer parts as 2x5x6 panels are used extensively for the castle walls.

Also, I have the first Millenium Falcon #7190, one thing that confused me when I got this out after my dark ages was the variety of colours - there are red and brown bricks which are mostly hidden in the final build. It made me wonder why Lego put those colour bricks in, as in older sets I'm sure these would have been colours more fitting to the main build colours.

Fauch

weirdly I get very little bright blue and bright yellow nowadays, whereas I get a nice amount of bright red. for the rest it seems to kinda match the graphs.

davee123

Robbzy88 said:

That explains why I've had such difficulty finding white pieces since returning from my dark ages, And not going to the hassle of making PaB orders.

Hm, I'm not sure. Even though the percentage of things like white is going down, the overall production of white is probably going up.  The production these days is something like 60 billion pieces per year, while the production back in the 70s and 80s was considerably lower (probably in the 1-5 billion per year range).

By my stats, white was at around 8.7% of production in 2014-- and they supposedly made about 60 billion pieces, meaning there were in the ballpark of 5 billion white pieces made in 2014.  But in 1987, when white was peaking at 22.2%, they were probably only making about 5-10 billion per year, which is in the ballpark of 1-2 billion white pieces.

But ultimately, it depends on your buying habits.  If you don't buy Star Wars or LOTR sets (for instance), you'll probably have a much lower amount of gray and dark gray than the rest of us!

jason1976 said:

One thing to consider though is the amount of pieces and the size of them in sets

I considered adding the size of the pieces based on the weight in BrickLink, but it seemed like there were too many elements that didn't have weights.  I did a similar breakdown by subtheme in LEGOLAND Space sets*, where I DID include the element weights, but that was much easier, since the scope was smaller, and I could put in reasonable guesses for the half-a-dozen or so elements with no weight.

DaveE

* See this thread: http://www.bricksetforum.com/discussion/20734/how-well-do-you-know-legoland-space-colors

jason1976

davee123 said:

* See this thread: http://www.bricksetforum.com/discussion/20734/how-well-do-you-know-legoland-space-colors

Ahh that was you that did that too! I though of it when I saw your pie charts!

Good work on that too!

chuckp

Very nice work. Were tires included in the analysis? Using the completely un-scientific process of looking at my parts collection and assembled sets, I can definitely see the gray dominance, but not the black. Since tires only come in black, I was just curious how they impacted the numbers. 

Fauch

a lot of structural elements are black, they are just not as obvious in the complete models than grey.
also, I think I figured where most of the bright yellow is. probably in the minifigs.

davee123

chuckp said:

Very nice work. Were tires included in the analysis? 

Curse you, I had to go and re-read all my code from a year ago to check! Turns out, I DIDN'T exclude anything except stickers. So all the above includes DUPLO, Scala, Fabuland, Technic, Galidor, Dacta, etc.

So, yes, it includes tires.  But, I think more importantly, it includes things like Technic pins and axles, which are VERY numerous, and might be pushing the numbers out for black and grey.  Almost makes me want to go back and re-calculate using weight to balance it out!  ... Almost.

DaveE

chuckp

^ It's definitely not necessary to re-write any code. I was just trying to get a feel for why the black elements made up such a large portion. I didn't even consider the technic pins and axles, but it makes sense. Regardless, still very interesting. Thanks again for posting the results. 

MrShinyAndNew

This is an interesting analysis. Can you go into a little more detail about what each pie slice represents? Did you essentially group all the sets released in a year into one pie, then count up the individual elements by colour? That would skew the analysis towards small parts, or rare parts or rare sets. It would be really interesting to factor in units sold.

davee123

MrShinyAndNew said:

Did you essentially group all the sets released in a year into one pie, then count up the individual elements by colour? That would skew the analysis towards small parts, or rare parts or rare sets. It would be really interesting to factor in units sold.

Yes and yes.  Oh, and yes!

I basically took every set that had an inventory over at BrickLink (and a release year), counted up each individual element in the inventory, and then made a pie graph of each color's percentage in terms of number of pieces.  Also, there's a limitation of the pie graph software that it can't represent less than 1/360th of the total, so essentially nothing less than 0.2777% is represented in the images.

There's a LOT of room for improvement!

- Some sets in the BrickLink database don't have valid release years, and so couldn't be counted at all.  I might be able to guess a few of them, but I'm not sure.

- Some sets are maintained over several years (like, say, the NXT).  So I counted them for a particular year, but they should have been valid for several years, because they were still in production.  That's pretty difficult to guess for everything up to recent years, when fans started taking notes on when things go out of production.

- Some sets CHANGE inventories drastically, but maintain a set entry, like #9320 (the inventory between 2003 (when I bought my first copy) and... 2006? was so drastic it was a completely different set!)

- Small elements shouldn't be counted the same as large elements (but they were).  I've been tempted to address this using element weights (which BrickLink has data for), but a lot of the data is incomplete.  I actually plan to make a 2nd pass at it this way, but I first have to make an algorithm for estimating unknown element weights (there are too many to do by hand!)

- Some sets are inventoried multiple times, as part of larger "sets" like package deals, or Advent calendars (Each small "window" is a separate "set" within the master set).  The program doesn't deal with this currently, although it probably could if I re-engineer it a bit.

- Production quantity isn't known.  So a set that sold 50,000 copies is counted the same as a set that sold 500,000 copies.  As far as I know, there's no good way of guesstimating that.  It doesn't really skew things in terms of rare elements, but more like rare sets, since a rare element probably doesn't show up in a whole bunch of sets.  There is data in the collector books on general rarity, but it doesn't really give much in the way of "how much" it constitutes, and as far as I know, it's not available to download (you'd have to load it in by hand, oof!)

- It doesn't take sub-elements into account.  So if you have a minifig torso that's blue, with orange arms and yellow hands, it only counts that element as blue.  That's also something I could probably fix, although it would take a lot more doing. Right now, I don't have a mechanism to load "part" inventories-- just inventories of sets, gear, books, and minifigs. :(

Ultimately... I guess I'm not sure how much all those points matter.  I think the biggest thing that needs to be dealt with is the "small vs large" pieces thing.  The other issues... not sure they'd really affect the overall results too much.  The sub-parts might affect things like minifig hands (lots of yellow hands!) and black tires (all the parts that come with attached tires).  Not sure how much of an impact that would make.

Anyway, if I give it another go, I'll be sure to post it!

DaveE

CCC

I guess at least for modern sets you could estimate rough numbers sold by looking at numbers of sets for sale on BL. After all you are only after relative numbers of each. It's bad approximation, but at least it is an approximation.

natro220

Or perhaps you could weight a set based on its type.  Retail, Lego Exclusive, Sdcc set, etc.

porschecm2

Thanks, Dave! This was very informative. I hope you don't mind that I've written up a post over at Brothers-Brick highlighting it. http://www.brothers-brick.com/2015/10/14/the-changing-palette-of-lego-1975-2014/

peterab

I'd be willing to bet that had you chosen to start your analysis in the mid sixties, you’d find white was already in decline by 75 and the eighties only represents a small resurgence. White and red were the dominant colours for many years.

I'm also surprised green doesn't show up in 75, since base plates were pretty much universally green at the time.

Thanks for the interesting and thought provoking post.

CCC

But baseplates are just one part. Even if one per set, the number is minimal compared to say red 2x4 or 2x2 bricks.

1974

What parts came in dark grey before 1977? That's when the mf untensil parts started to show up and not much more until 12V sleepers in 1980 and castle parts in 1984

What dark grey parts was around in 1975?

davee123

peterab said:

I'd be willing to bet that had you chosen to start your analysis in the mid sixties, you’d find white was already in decline by 75 and the eighties only represents a small resurgence. White and red were the dominant colours for many years.

I wanted to do the analysis back through 1949, but the data at BrickLink starts to get pretty spotty when you go back a ways. But yeah, initially I think it was just red & white!

peterab said:

I'm also surprised green doesn't show up in 75, since base plates were pretty much universally green at the time.

Actually, it does! It's just such a small sliver that it's barely noticeable. Only 0.87% in 1975! (Zoom way in and you'll see it)

1974 said:

What parts came in dark grey before 1977? That's when the mf untensil parts started to show up and not much more until 12V sleepers in 1980 and castle parts in 1984

Nothing really that I'm aware of.  I think according to BrickLink data, dark gray started showing up in 1978.  However, if you take the BrickLink data literally, then it showed up before (but not really).

For example, they had the old-school HO-scale cars in the late 50s and early 60s, which BrickLink says were in dark gray (as well as other colors):

http://www.bricklink.com/catalogItemIn.asp?P=664pb01&colorID=10&in=A

And, while technically true that they did have "dark gray" cars, I don't think it was the same color of plastic that was used later in 1978.  They weren't even using ABS at that point, so it definitely wouldn't match!

DaveE

AlphaBravo

Which of these event poster designs looks more "LEGO" to your eye?





I did these for a lecture series in February 2012. I did the top one first, basing the colors for the Hokusai rendering on a 2010 LEGO palette that I found somewhere on the net.

But having entered my dark age circa 1979, something was bugging me, so a couple days later, I did a remake based on the colors of my childhood (but choosing to ignore the translucent colors which had just appeared in the new space sets). The second design is what we ultimately printed.

The dark ages ended ten months after the posters, with a couple Duplo purchases, and I've been crazy about the new colors ever since, particularly the shades of green and all the "bright" (aka girly) colors. Mind, it's going to be a lot harder for my daughter to organize her brick storage by color than it was for me, if she's ever inclined to...

Love the analysis, DaveE!

1974

Qouting on this site sucks, so I'll reply to that 'dark gray' comment like this :

No BL, does not say WHEN a certain part was available in dark grey IF you just search 'colours' .. then you'll just see a part that WILL be released in dark grey much later on

So in 1975, the 3624 hat was NOT released in dark grey, it was made in other colours that year surely, but not dark grey

Even if 3001 was made in 1980 (that particular version) it was only made in old dark grey in 1998 .. and so on

Do your graph take that into consideration?

I see NO dark grey parts at all until '77 and then not common until '84 and proper bricks much later ...

Cheers,

Ole

davee123

1974 said:

No BL, does not say WHEN a certain part was available in dark grey IF you just search 'colours'

It's not really quoted in the interface, no. If you wanted to find it through the user interface, you could figure it out if you have the time. In the catalog, if you click on the color of a particular part (like the 3624 hat in dark gray you're suggesting), you can see a list of the sets and their year of release:

http://www.bricklink.com/catalogItemIn.asp?P=3624&colorID=10&in=A

So, you CAN go through that list, finding the earliest year that a particular part was released in a particular color (at least, as far as BrickLink knows).  But it takes a lot of effort.

1974 said:

Do your graph take that into consideration?

Yep! I didn't use their interface to do this by hand-- I downloaded their inventory data and derived it from that using a computer program. If BrickLink said "here's a set from 1979 that has dark gray in it", then I'd include that in the final tally. It's not derived using the color chart for each part that BrickLink provides on its website. That would be pretty inaccurate!

1974 said:

I see NO dark grey parts at all until '77 and then not common until '84 and proper bricks much later ...

My data actually shows no dark gray until 1978-- which elements do you see in 1977 that are dark gray? (My data is from 2014, so it may be out of date if something's been updated on BrickLink in the last 12-ish months or so) Maybe you're confusing light gray on my graphs for dark gray? Just in case, here's a zoom-in of the colors I found from 1975:



My data shows dark gray as a very small blip starting in 1978-- sometimes not even showing up at all-- and finally starting to explode in 1998. There's a big blip in dark gray in 1980, when dark gray was used in 4.5v and 12v trains as the railroad ties. It jumps from 0.09% to 2.23% between 1979 and 1980! ... But then back down to 0.55%.

DaveE

davee123

1974 said:

What dark grey parts was around in 1975?

Hm, I just noticed-- the dithered trans-clear looks a lot like dark gray if you don't zoom in on the picture in the first post.  Click the picture to zoom in, and you should be able to see dithering, which I used to differentiate between transparent colors and solid colors.  That might be causing some confusion!

DaveE

TigerMoth

davee123 said:

So, you CAN go through that list, finding the earliest year that a particular part was released in a particular color (at least, as far as BrickLink knows).  But it takes a lot of effort.

For old stuff, Peeron might give simpler results:

http://peeron.com/inv/colortable/3624?ip=0

When it was still updated it was reasonably accurate - and when it comes to colours, perhaps more so than Bricklink.

davee123

TigerMoth said:
For old stuff, Peeron might give simpler results:

When it was still updated it was reasonably accurate - and when it comes to colours, perhaps more so than Bricklink.

Yeah, Peeron has pretty good data for the 1980-2005 ballpark.  They also are great (much better than BrickLink) for nuanced data like when a particular part variant started showing up.  BrickLink's historical data in the pre-2005 era is ... sometimes sketchy.  I'll take it with a grain of salt!

DaveE

1974

Yes, I confused trans-clear for dark grey, doh! And yes, it's 1978, not '77. I do seem to remember reading that sets with minifigs where rushed out allready in '77?

OchreJelly

Hey Dave - This data set seems really interesting! But I have to say, a lattice of pie charts is (scientifically) just about the least effective way to visualize it. Would you be prepared to publish the raw (unaggregated) data, so others can analyze and visualize it further? - Iain (TBB).

Fauch

I'm not sure there is a search by color in the database?
like selecting all the sets which contain a specific color and ranking them on the number of pieces of this color they contain.

davee123

OchreJelly said:

Hey Dave - This data set seems really interesting! But I have to say, a lattice of pie charts is (scientifically) just about the least effective way to visualize it. Would you be prepared to publish the raw (unaggregated) data, so others can analyze and visualize it further? - Iain (TBB).

Sure, it's not really my data anyway!  I actually just set it up so that it can run by element weight, and I was going to have a go at pulling the rest of 2014/2015 data as well.

I had it in a pie-chart format merely because that's how I had it set up to do the LEGOLAND Space images done.  Each theme was pulled into a little pie graph, and it looked pretty good as a representation of a theme.  Only took a little tweak to replace each "theme" with a "year", and I could make a swath of pie graphs.  Ideally, if the idea is to see changes over time, I'd probably do each color as a separate line, or do one of those "stacked" charts.

I'll see if I can compile the raw data into something easily digestible this weekend and put it up.

DaveE

OchreJelly

...that'd be sweet! And I don't mind sprawling raw data row-by-row too - I can throw it into an analytic tool and slice-and-dice it to look for trends across factors other than time.

davee123

Alright, here we go:  updated!  Now hopefully a little easier to see trends in colors:


http://www.suave.net/~dave/tmp/color_chart_by_weight.png

So, that's now including all years from 1953 - 2015.  I probably actually shouldn't have included anything before 1970, because the data starts getting sketchy around then.  But what the heck!

It also counts everything by weight now, so you'll notice that green is much more substantial in earlier years, thanks to its appearance in baseplates and the like.

And if you're feeling frisky, here's the raw data:
http://www.suave.net/~dave/tmp/colors_by_weight_raw.txt
http://www.suave.net/~dave/tmp/colors_by_pieces_raw.txt

Oh, and also thanks to Chris & Iain for posting this via TBB!

DaveE

1974

You accidentally art there, mate :)

Venunder

A very psychedelic chart.
You have to wonder whether the Lego designers are taking drugs?  
Or maybe it is just easier to make those new colours now.... :)

Aleydita

That chart is awesome. Nice work. Looks like yellow and green are on the way out. Well, not really, but it's interesting to see how in % terms the increasingly wide pallet makes some colours less needed.

Aanchir

davee123 said:

It also counts everything by weight now, so you'll notice that green is much more substantial in earlier years, thanks to its appearance in baseplates and the like.

Besides the decline in baseplates, another factor in the decline in Dark Green (classic green) is the increasing use of other colors like Bright Green, Bright Yellowish Green, Sand Green, and Earth Green

Venunder said:

A very psychedelic chart.
You have to wonder whether the Lego designers are taking drugs?  
Or maybe it is just easier to make those new colours now.... :)

It's actually a lot harder for designers to introduce new colors now than it was prior to 2003. Since this chart doesn't include themes like Fabuland and Duplo (and since it uses Bricklink data, which conflates some of the more obscure colors with more well-known ones) it's hard to see here, but around 2003–2004, their color palette had really spiralled out of control. However, even if the chart did include those themes, it might not have been wholly apparent, because many of those colors were used for just one or two sets! LEGO had to do a lot of tidying up to get their palette back under control.

While it looks like today there are a lot more colors than there were in the late 90s and early naughts, what you're really seeing is LEGO actually taking greater care to ensure that they actually use all the colors they already have.

Fauch

that's crazy to see that despite the number of different colors skyrocketting, the proportion of black and grey is increasing.

davee123

Aanchir said:

Since this chart doesn't include themes like Fabuland and Duplo (and since it uses Bricklink data, which conflates some of the more obscure colors with more well-known ones)

Actually, it does include Fabuland and Duplo-- I made an error on that first post (corrected a few posts down). Since I made the original program about a year ago, and made it for other purposes, I had some notes saying that I left out strange themes like Scala, Primo, etc. But it turned out that although those notes were correct at some point, by the time I finished (the state of the data at the end), I had included everything (except stickers, which have a BrickLink color ID of "0").

A bit more info on the element weights, too (for the curious): About 5% or so of the BrickLink elements don't have weights. Mostly for strange things like Scala clothing or Toolo girders or something. I ultimately did 2 things:

For some parts, their part ID is made out of a "base element" or "base combination", like "973p83c01", which is a printed version of 973 (a minifig torso) with a pattern number of "83", and a combination ID "01". So if there was a base combo or base element with a known weight, I'd use that to estimate the weight of the part.

For others where that wasn't possible, I did something sketchy, but that seemed to actually give a reasonable result (I was surprised that it did as well as it appeared to). I took all the known weights for elements in the same BrickLink category, and averaged them out, then assigned that weight to the element. It was probably off by a bunch, but in general did pretty well assigning large weights to large parts, and small weights to small parts. So, for the purposes of this exercise, it probably turned out ok.

DaveE

madforLEGO

Fauch said:

that's crazy to see that despite the number of different colors skyrocketting, the proportion of black and grey is increasing.

An increase of the number of Batman related sets likely. :-p

ryan498

Great chart. Interesting to see when the different colours were introduced, looks like the range really started in 2003/2004. I love the tan colours though. 

Fauch

of course, the sand green invasion is because of harry potter, though it looks like it began before 2001? belville sets maybe?

davee123

Fauch said:

of course, the sand green invasion is because of harry potter, though it looks like it began before 2001? belville sets maybe?

Technically, I think it was almost 100% due to the Statue of Liberty (#3450) set in 2000.  It had gobs of pieces, and almost all of them were sand green!

DaveE

MrShinyAndNew

The stacked chart is great. Is there any chance you could make it huger? Huge enough to show colour labels for all the colour on the right? Also dynamic? so you can mouse over it and see the colours, percentage values, etc, at any point? Also build it out of Lego in real life? Also a pony?

But seriously, cool analysis. 

davee123

MrShinyAndNew said:

The stacked chart is great. Is there any chance you could make it huger? Huge enough to show colour labels for all the colour on the right? Also dynamic? so you can mouse over it and see the colours, percentage values, etc, at any point? Also build it out of Lego in real life? Also a pony?

Huger?  I suppose I can do that fairly quickly, although it's not quite as pretty.  Here's a zoom-in from 1990 to 2015 without gridlines, and with automated color name tagging (which overlaps itself sometimes, sorry):

http://www.suave.net/~dave/cgi/color_chart_by_weight_zoom2.png

I could, in theory, go bigger, but I think I'd have to make it around 30,000 pixels tall before it became fully readable, which is Too Huge (tm).  Probably better at that point to annotate something differently.  That's the raw output from my program, so sorry it's a little annoying to look at.

Dynamic?  Well... I'd love to make an interactive mouse-over chart, but I'm not sure if I have the patience to do one.  If you know of a JavaScript/Flash engine that does that sort of thing off-the-shelf or freeware-wise, I'll be happy to plug in the data.  (Or, if you want to do the same with the raw data above).

I am fresh out of ponies :(

DaveE

Sven_F

This is super interesting. 
Does your v. 2 chart include tyres and technic?  Tyres apart from being black  are also significantly heavier.  I think this might skew the black. 

davee123

Sven_F said:

This is super interesting. 
Does your v. 2 chart include tyres and technic?  Tyres apart from being black  are also significantly heavier.  I think this might skew the black. 

Yes, it does include tires and Technic, which I thought might be skewing the data originally, since it was based on piece count rather than weight (and Technic has a lot of black small pins and axles!)

So, out of curiosity, I just tried another experiment-- what happens if we leave out the tires, and just skip 'em?  Turns out, not all that much.  And what happens if we leave out technic sets?  Also, not too much.


So, ok, I know this has been bugging people, so I did another check.  What IS causing all that black?

Top 10 contributers to black (by BrickLink element type):

3760.22 - Plate
2905.12 - Tire & Tread
1491.42 - Plate, Modified
1485.73 - Brick
1480.95 - Technic, Liftarm
1350.79 - Technic, Brick
1311.08 - Hero Factory
772.24 - Tile, Modified
756.88 - Technic, Pin
743.70 - Technic, Connector

So, Tires & Treads (the thing that I removed) is a pretty big factor, but it's not overwhelming by any means.

Ok, what about themes?  Which themes contribute the most to black?

3133.56 - Ninjago
1825.70 - Technic
1317.59 - Star Wars / Ultimate Collector Series / Star Wars Episode 4/5/6
1315.28 - Friends
1148.59 - Ultra Agents
1093.55 - Super Heroes / Avengers Age of Ultron
1070.4 - Super Heroes / Avengers
948.16 - Star Wars / Star Wars Episode 7
913.59 - Speed Champions
874.67 - Technic / Model / Race

Ninjago?  Seriously?  Wow.

DaveE

Fauch

lol and friends contributes more than ultra agent? are there even black pieces in friends?

oh wait I get it, I was thinking about the average amount of black by set. but you actually have to add them all. about 150 sets in friends and ninjago VS about 20 in ultra agent.

Fauch

star wars 7 is actually a very serious contributor, only 9 sets, over 100 per set. in comparison ninjago must be around 20 per set.

Lyichir

Fauch said:

lol and friends contributes more than ultra agent? are there even black pieces in friends?

oh wait I get it, I was thinking about the average amount of black by set. but you actually have to add them all. about 150 sets in friends and ninjago VS about 20 in ultra agent.

There's actually a lot of black in Friends this year, with many of the "pop star" sets practically treating it as a primary color, with the usual pinks and purples and azures only used as accent colors. That said, I still wouldn't have expected it to make THAT big an impact compared to other themes.