Price per Piece - An outdated idea?

[nerdsforprez]

A caveat prior to reading this:

The below is not an argument for value of a set.  That is highly subjective – I get that.  So, please no rants on the subjectivity or personal nature of set’s value.  If you are willing to buy a helicopter set, at any cost, because you love helicopters - that is terrific, but doesn’t have anything to do with this post.  If you are willing to buy a set because it has 6 Claas tires and you value them above all else, regardless of the economic value – great, terrific, but again, it doesn’t have anything to do with this post. 

I preface the below because I don’t want this post to morph into some debate on who values what. We have plenty of those discussions.  So, please actually read the below.  If there is a post about the subjective elements of value, I will kindly refer you to read this portion of the thread.

Okay – with that out of the way, here we go!

I have alluded to this in the past, and I did some real simple analyses on this stuff years ago, but I did not feel like digging through my hard drive to find the files and I know the work I have done in the past did not cover anything past 2014.  I felt like once again examining the issue of Price-Per-Piece (PPP) as it relates to a measure of monetary value for Lego Technic Sets, but with all new data beginning with sets after 2014.   

I have not added other Lego genres in the data set, so I do not know if the following also holds for other genres. 

Issue-at-hand is PPP.  It is tempting to use this as a metric for the value of a set.  Easy-peasy right?  Simply divide the cost of a set by how many pieces it has and you get a price-per-piece value.  The cheaper every Lego element is, the better the value of the set, at least from a strictly economic standpoint – Right?  

This comes with obvious flaws.  The most obvious, is that the economic value in a set is not dictated by pieces, but actual LEGO.  It is not difficult to understand that a set may have few pieces, but if those pieces are large and cost TLG a lot of ABS or difficulty in production/assembly, then they will be expensive.  A set with tons of panels will be relatively expensive because panels, typically, are heavy (relatively speaking) because they contain a lot of product (ABS). 

With this understanding I propose that PPP is an outdated metric and I would encourage discontinuing using it as an idea of the value of a set from a strictly economic standpoint (NOT counting subjective factors).  I don’t make this recommendation because it is meaningless.  On the contrary.  In the information I share, the amount of pieces in a set and its price are HEAVILY correlated.  So, obviously it contains some value.  I only discourage its use because there is an equally simple, yet fairly more effective way of measuring the economic value of a set.  And that is measuring the Price-Per-Gram (PPG).  Take the cost of a set and divide it by its weight (I obtain my values from Bricklink but there are other sources).  In this case I think grams is the easiest way to get this, giving us the PPG measure. 

Let’s look at why I make this proposal.  In a data set starting with 2014, I randomly selected two sets from each year group from the following size categorizations:

Large sets (over 2,000 pieces)

Medium sets (1,000-2,000 pieces) and

Small to Medium sets (under 1,000 pieces)

And simply looked at their correlations between price and piece count versus weight.  If I do that we get a correlation of 96% between set cost and number of pieces and a correlation of 99% between set cost and the weight of a set.  Without getting too technical (there are some young readers and I want them to grasp this) that means that number of pieces roughly accounts for 92% of the cost of a set and the weight accounts for roughly 96% of the cost.  That is AMAZING.  Despite our complaints that means that Lego sets (at least Technic sets) are reliably priced (on average) and things like license from like Porsche or Volvo means very little when pricing sets.  This should be great news for any Technic fan.  It also means that using weight or number of pieces are nearly identical ways of measuring the economic value of a set (even though using weight still beats piece count.  It explains 4% more of price than does piece count). 

However, there is one caveat.  I left some information out. 

I excluded all sets with an abnormally high PPP.  Anything over 13 cents a piece (examples are 42030, 42070, etc).  Believe it or not, according to the range of sets I sampled described above, there were relatively few.  But I left this information out for a simple reason and that is sets with an abnormally high PPP usually have large or heavy pieces in them.  Obviously, if set has a great PPP, but contains like 50% pins (small, light, relatively little ABS) then this is important to know when comparing to a set with like panels, 15L liftarms, motors, battery boxes, etc.   

When adding sets with an over 13 cents a piece value, the numbers change from the above.   For piece count, the number of explained price drops to 86%.  If we use weight, we still explain 94% of a set’s price.  The difference here changes from 4% difference (above, with NO abnormal PPP sets) to a difference of 8% when abnormal PPP are included. 

To further drive this point across, if we reach back to 2010 (admittedly, I don’t like to do this because going back that far inflation begins to eat at our results) we find the famous but economically infamous 8043.  Tons of motors, tons of actuators, etc. … tons of non-traditional Technic elements. Still LEGO just not in its traditional form.  Adding just this one set further separates the delta value of explained variance of piece count versus weight in relation to Lego costs by another three percent (11% difference).  This sounds like a very small number but when it can add several percentage points to a data set of nearly 50 samples, well it really drives the point home. 

I write all this at a time when 42099 and 42100 are coming out.   42099 numbers are already released. At only 958 pieces but 250 USD, that is like an atrocious 26 cents per piece value.  However, the central point of this post is that if that is the metric one is using to decide or emotionally respond to 42099 then they are cheating themselves.  PPG is not perfect, and when adding abnormal sets (as defined by many non-traditional Technic elements, i.e. motors, LAs, etc.) it is equally important to recognize that its correlation to price also goes down.   JUST NOT TO THE EXTENT THAT THE RELATIONSHIP BETWEEN PIECE COUNT AND PRICE DOES WHEN ADDING THESE ABNORMAL SETS.  So by no means is it perfect but PPG easily beats PPP when discussing their relationships with set cost. 

I have 42099 values for cost and piece count.  If I add it to the data set our correlation drops drastically.  9% to be exact.  From 86% to just 77%.  Weight for the set is not out yet (or at least I have not found it) but I suspect the relationship between cost and weight will drop as well, just not to the extent of cost and piece count because motors, large, heavy elements, etc. will be taken into consideration.  I suspect 42100 will further this theme. 

Fun thing to examine.  Eager to hear other’s opinions but please try and stay on topic. 

The above is certainly not a perfect formula but it is better that what people use right now.  And yes, I get that using a set’s weight contains non-ABS parts.  Manual, packaging, etc.  I would argue it is all LEGO, I mean TLG produces it right?  But, I don’t want to get into that because it does not matter.  The relationship is improved by using weight versus piece count – you really can’t argue with that.    

Edited August 10, 2019 by nerdsforprez

[Mr Hobbles]

So, firstly, yes, PPP is silly, for the reasons you mention. Size of parts, size of packaging, shipment costs, molds, etc.

But, you mention 42099 and 42100.

42099 and 42100 are interesting cases, and probably only comparable to sets that came out in 2009-2011 during the introduction of Power Functions.

42099 and 42100 are flagship models designed to showcase the power of Powered UP in technic models through Control+. As a result, at this early stage, the cost of the set has to include:

1. The cost of the ABS plastic that went into them (weight/quantity).

2. The cost of new molds for those parts.

3. Additional new materials for the new components - PCB's, chipsets, metal, wires, etc.

4. The new supply chain requirements for those parts.

5. The cost of developing the new apps, and firmware for the electronics components.

6. Recouping the costs for 2-4 in a reasonable timeframe.

The sixth one is important. A business functions by return-on-investment, and that time period is not indefinite. They can't say "we can spend this much up front because we think we'll get it back over time" like Power Functions, because at this point they have no measure of whether Powered UP will be a success yet. They may be forced to replace it in a couple of years time due to people not liking it. They'll have had to make the same decisions at the start of Power Functions too.

We know Lego typically works in 3 year blocks from previous experience through themes, products, etc. If it hasn't turned out to be succesful by whatever internal measures they have in place by year 2, they've typically had no qualms in calling it quits by the end of year 3. This means that they'll have likely worked out amortizing the cost over 3 years.

No doubt, as Powered UP sticks around for longer the cost of parts will come down as they did for Power Functions, and future sets 3+ years out will be perceived as better value for money - the costs of hardware, firmware and app development will come down as most of the work will have been done, and they'll be able to focus on charging for material and production.

[weavil]

When looking at sets for me or my daughter, I always look at these points when deciding a sets value (except for basic type sets):

• Licensed: Is the set a licensed product? If so is it a big name like Harry Potter/Marvel? or not as big a name like Jurassic World?
• Complexity of Parts: Does the set have large parts like panels/boats/supports/animals? If so what size are those parts? How many of the large parts? Is it one T-Rex? Twenty Technic panels? One small paddle boat? Are the large parts easy to make (like a 2 or 3 piece mold)? Or are they difficult to make (like a 5 or 6 piece mold)?
• Printed: Do any of the parts have printing on them (other then regular figs)? How many of the parts are printed on? Is it a little bit like a 1x2 tile or is it a lot like a big animal/creature/figure?
• Pre-Assembled: Does it have any pre-assembled parts? Are they simple or complex assemblies? Are they small or big? How many does it have?
• Design: Did the set take a few months to design? Did it take 10+ months to design?

There will always be a debate about the value of a set. A collector will not worry too much over price. A builder will look at what parts they get for their money. A person who plays with the set will look at how many functions/entertainment they will get for their money. Right now we can use Bricklink to assess the the value of set, sans new parts for that set/year.

[tafkatb]

I'm not sure you can ever view this from a "purely economic" angle, because exactly what it is you're buying may not be as clear-cut as it seems. (And economics is a much softer science than most are willing to admit anyway.) What I'm getting at is: a LEGO set isn't just A Thing, it is A Thing That You Build, which means that the building process is an inextricable component of the product. And with that taken into account, I'd say PPP is still a valid metric, because piece count is pretty well correlated with how long a set will take to build. All else being equal, a 3,000-piece set is still a better value than a 2,000-piece set of the same weight and price simply because it will take roughly 50% longer to build, and that building experience is part of what you're buying. Now, that may be veering into more subjective territory than you're looking for here since it will vary depending on the speed/skill of the builder and thus can't be definitively quantified, but the build time deserves to be considered alongside the build size. And since piece count is likely the best at-a-glance indicator of build time, it can't be completely discounted.

To put it another way, the comment that "it is all LEGO," even the manual and packaging, can be more holistically expanded to encompass the manual, the packaging, and the time spent putting it together. I don't know about everyone else, but that's what I'm paying for.

[deraven]

So, to your point about metrics for objective value of a set, I'd say I'm 95% in agreement.  I would prefer to see your same set of data looking only at the weight of the Lego elements and not total packaged weight... but that brings me to the only reason I don't think PPP is "outdated" vs a more accurate metric like PPG:  Information on the weight of the set is not easily and readily available for quick comparison purposes.  Since the piece count is right on the box along with the price being obvious, the quick math is the easy way to go.

Now, I would say if large numbers of people were basing their purchasing decisions solely on the PPP, it might be more important to push for something like PPG... but I don't think anyone looks at sets in a vacuum that way and can pretty quickly identify the other selling points of a set if they see that the PPP is higher than average.

I don't think PPP was ever meant to be more than a quick-and-dirty comparison method, and as that I think it still serves its purpose and it's outdated.  However, if more services like Brickset started automatically calculating the PPG on sets the same way they do the PPP, I think that would go a long way toward making it a better tool for comparing the value of sets against each other in a more accurate way.

8 minutes ago, tafkatb said:

To put it another way, the comment that "it is all LEGO," even the manual and packaging, can be more holistically expanded to encompass the manual, the packaging, and the time spent putting it together. I don't know about everyone else, but that's what I'm paying for.

I very much agree with this as well, and shows how even mostly objective measures keep spilling over into subjectivity when it comes to Lego!  

[JonathanM]

Is there a dataset available with weight, price, number of pieces, year, licensed, electronics etc? I guess with brickset/rebrickable pulling down the data would be reasonably easy? Once done, model price with everything available and then take a look at outliers to refine things.

[icm]

I recently looked at this from the perspective of City sets when trying to decide what price I was willing to pay for the Space line.  I didn't go into great detail or conduct a statistical analysis, but I did look at the Bricklink weight and Brickset original RRP of most of the sets in my collection, which spans forty years and contains more sets than I care to admit.  Using inflation numbers from the US Consumer Price Index, I found that the typical (not average, since I didn't study the statistics) unadjusted price per part has slowly drifted up over the past forty years, but the adjusted price per gram has stayed remarkably constant.  In 2019 dollars, a fantastic deal like the Saturn V has a PPG of about 4.7c, most sets have PPG between 5c and 6c, and the really overpriced kits have a PPG of about 6.7c.  Those numbers may be wrong, since I didn't write anything down and I'm repeating them from memory, but they're what I generally rely on now when trying to assess what a "good price" for a kit might be.  However, I was surprised to find that some kits that I'd previously considered a great value are relatively overpriced by PPG, like the 2012 X-wing and the 2017 Y-wing.  So there's obviously a lot more than weight in Lego pricing, but I agree that as a quick and easy measure of value PPG is more reliable than PPP.  I contacted Huw at Brickset to suggest that PPG be displayed alongside PPP, so he's aware that there's some interest in it.

Edited August 10, 2019 by icm

[howitzer]

Did you consider weight of the set as it sits in a shelf of store, or did you only take into consideration weight of the completed model? In other words, how do the packaging, manual, etc. affect the numbers? If you separate the weight of the model itself and the weight of everything else, is there a proportional difference between different sized sets?

[syclone]

I personally prefer PPP as I don't care how big or heavy parts are - the more parts you get, the better (except black 2L pins ). It is supposed that a person knows that there are special elements that will drive the price up - electronics/wheels/panels/flexaxles/licenses - so that is taken care of mentally. Let's say set #1 has heavy remote control parts&motors, and is built around those, but set #2 doesn't have electrics and instead has frames and other functions, increasing it's part count significantly. Both are the same price and approximate weight, but PPP for set#2 will be smaller than PPP for set #1, yet PPG will be simliar-ish. So a buyer that needs parts and doesn't care much about electronics will see that he will get a better deal off of set#2 because it's PPP is smaller. And the buyer that doesn't care about specific parts will see that one set has a better offer for quantity of parts but other set has specific elements that drive that price up.

Conclusion for me: I personally need a value that counts every element in an equivalent manner, without changing it up because of the nature of certain elements. Lego could release a superheavy part that will drive up the PPG unrealistically, yet PPP would count it as a part. Maybe I'm wrong on the mathematical&economical level (not my strong points) but on a personal level PPP is my preference.

Edited August 10, 2019 by syclone

[nerdsforprez]

12 hours ago, Mr Hobbles said:

So, firstly, yes, PPP is silly, for the reasons you mention. Size of parts, size of packaging, shipment costs, molds, etc.

But, you mention 42099 and 42100.

42099 and 42100 are interesting cases, and probably only comparable to sets that came out in 2009-2011 during the introduction of Power Functions.

42099 and 42100 are flagship models designed to showcase the power of Powered UP in technic models through Control+. As a result, at this early stage, the cost of the set has to include:

1. The cost of the ABS plastic that went into them (weight/quantity).

2. The cost of new molds for those parts.

3. Additional new materials for the new components - PCB's, chipsets, metal, wires, etc.

4. The new supply chain requirements for those parts.

5. The cost of developing the new apps, and firmware for the electronics components.

6. Recouping the costs for 2-4 in a reasonable timeframe.

The sixth one is important. A business functions by return-on-investment, and that time period is not indefinite. They can't say "we can spend this much up front because we think we'll get it back over time" like Power Functions, because at this point they have no measure of whether Powered UP will be a success yet. They may be forced to replace it in a couple of years time due to people not liking it. They'll have had to make the same decisions at the start of Power Functions too.

We know Lego typically works in 3 year blocks from previous experience through themes, products, etc. If it hasn't turned out to be succesful by whatever internal measures they have in place by year 2, they've typically had no qualms in calling it quits by the end of year 3. This means that they'll have likely worked out amortizing the cost over 3 years.

No doubt, as Powered UP sticks around for longer the cost of parts will come down as they did for Power Functions, and future sets 3+ years out will be perceived as better value for money - the costs of hardware, firmware and app development will come down as most of the work will have been done, and they'll be able to focus on charging for material and production.

They are interesting cases.  Time will tell, but I agree with all you mentioned.  Folks need to account for all these things when considering the price.  Interesting note, despite all the complaints about the Bugatti, it was reasonably priced (as defined by BOTH PPP and PPG).  However, the Porsche was a little extreme on both PPP and PPG.  Both had quite a few unique pieces, had to deal with licensing, etc.  However, the Porsche was a foray into the new, whereas the Bugatti was only following tracks.  I wonder if the extra two cents (ish) per piece AND gram (weight) were due to the point number five you make.  Development.  Though the Bugatti also had development of course, perhaps the Porsche had more, because it was starting a whole new line (UCS Technic)?  All this is just speculation of course, but interesting in terms of your points.

11 hours ago, tafkatb said:

I'm not sure you can ever view this from a "purely economic" angle, because exactly what it is you're buying may not be as clear-cut as it seems. (And economics is a much softer science than most are willing to admit anyway.) What I'm getting at is: a LEGO set isn't just A Thing, it is A Thing That You Build, which means that the building process is an inextricable component of the product. And with that taken into account, I'd say PPP is still a valid metric, because piece count is pretty well correlated with how long a set will take to build. All else being equal, a 3,000-piece set is still a better value than a 2,000-piece set of the same weight and price simply because it will take roughly 50% longer to build, and that building experience is part of what you're buying. Now, that may be veering into more subjective territory than you're looking for here since it will vary depending on the speed/skill of the builder and thus can't be definitively quantified, but the build time deserves to be considered alongside the build size. And since piece count is likely the best at-a-glance indicator of build time, it can't be completely discounted.

To put it another way, the comment that "it is all LEGO," even the manual and packaging, can be more holistically expanded to encompass the manual, the packaging, and the time spent putting it together. I don't know about everyone else, but that's what I'm paying for.

No one is arguing for a "purely economic" anything.  Your words, not mine.  In fact, I like how other's word the measurement (below). It really is just a "quick-and-dirty" way of looking at things - but even though it is quick and dirty I really need to emphasize something.  One does not need to be statistically sophisticated to understand this point.  And that is, even in the messy data (with like 8043, etc.) weight accounted for 94% of the variability of a sets cost.  That is AMAZING.  It really is.  So even though there are issues like licensing, etc. only 6% of a sets price is left explained by factors other than weight.  Perhaps not "pure" or "perfect" - but pretty darn close.  A whole 12% is left unexplained (which is still really small) when we use piece count.    

However, I do really like your second point. I hadn't thought of that.  Admittedly, that is somewhat a subjective interpretation of value that I had hoped to avoid, but it does provide a great explanation of why PPP continues to be useful.  Something PPG does not capture.  Thank you for sharing. 

11 hours ago, deraven said:

So, to your point about metrics for objective value of a set, I'd say I'm 95% in agreement.  I would prefer to see your same set of data looking only at the weight of the Lego elements and not total packaged weight... but that brings me to the only reason I don't think PPP is "outdated" vs a more accurate metric like PPG:  Information on the weight of the set is not easily and readily available for quick comparison purposes.  Since the piece count is right on the box along with the price being obvious, the quick math is the easy way to go.

Now, I would say if large numbers of people were basing their purchasing decisions solely on the PPP, it might be more important to push for something like PPG... but I don't think anyone looks at sets in a vacuum that way and can pretty quickly identify the other selling points of a set if they see that the PPP is higher than average.

I don't think PPP was ever meant to be more than a quick-and-dirty comparison method, and as that I think it still serves its purpose and it's outdated.  However, if more services like Brickset started automatically calculating the PPG on sets the same way they do the PPP, I think that would go a long way toward making it a better tool for comparing the value of sets against each other in a more accurate way.

I very much agree with this as well, and shows how even mostly objective measures keep spilling over into subjectivity when it comes to Lego!  

From the gut-reactions at least on this side of the forum (I don't visit other places very often) I actually do think that folks base opinions of sets off of PPP.  Perhaps it is some of our younger members, but this type of stuff comes up all the time.  In fact, and this is only pulling from memory, admittedly I did not go back to check all this, but every year the Technic flagship models come out and the costs keep hiking on up I see PPP all over the place.  Sure, it is easier to obtain than PPG, but not by much.  Quick search on BL yields a sets weight.  

But I do agree with your other point.  The whole PPG would be better if it only included ABS (and things like motors, etc.) but not manual, etc.  However, I really don't think it matters much.  Again, anything explaining 94% of ANYTHING is quite remarkable.  Like it or not, both PPP and PPG show excellent relationships with a set's price.  It is just that PPG is better, and seems to take into account at least some degree size or weight of the elements, not just individual elements themselves, and I don't think it is that hard to get.  Not as easy as PPP.  But both take at least a little work because there are very few than can do that form of mental arithmetic in their heads. 

8 hours ago, icm said:

I recently looked at this from the perspective of City sets when trying to decide what price I was willing to pay for the Space line.  I didn't go into great detail or conduct a statistical analysis, but I did look at the Bricklink weight and Brickset original RRP of most of the sets in my collection, which spans forty years and contains more sets than I care to admit.  Using inflation numbers from the US Consumer Price Index, I found that the typical (not average, since I didn't study the statistics) unadjusted price per part has slowly drifted up over the past forty years, but the adjusted price per gram has stayed remarkably constant.  In 2019 dollars, a fantastic deal like the Saturn V has a PPG of about 4.7c, most sets have PPG between 5c and 6c, and the really overpriced kits have a PPG of about 6.7c.  Those numbers may be wrong, since I didn't write anything down and I'm repeating them from memory, but they're what I generally rely on now when trying to assess what a "good price" for a kit might be.  However, I was surprised to find that some kits that I'd previously considered a great value are relatively overpriced by PPG, like the 2012 X-wing and the 2017 Y-wing.  So there's obviously a lot more than weight in Lego pricing, but I agree that as a quick and easy measure of value PPG is more reliable than PPP.  I contacted Huw at Brickset to suggest that PPG be displayed alongside PPP, so he's aware that there's some interest in it.

Interesting.  Yes, I would like to know how this plays out in other genres.  I do remember in the old data set I looked at I did look at Star Wars models.  Lots of ABS in the large UCS SW ships - I remember like the Super Star Destroyer and Imperial Star Destroyer were relatively heavy because of all the large plates (i.e. heavy - lots of ABS) and when considering PPG over PPP it really improved the model. 

 

Edited August 10, 2019 by nerdsforprez

[tafkatb]

1 hour ago, nerdsforprez said:

No one is arguing for a "purely economic" anything.  Your words, not mine.

Apologies for misquoting you; I thought you had said "purely" but then looked back at your post and saw the phrase was "strictly economic," which does have a bit different connotation.

1 hour ago, nerdsforprez said:

[...] weight accounted for 94% of the variability of a sets cost.  That is AMAZING.  It really is.  So even though there are issues like licensing, etc. only 6% of a sets price is left explained by factors other than weight. 

It really is! I love those moments in statistical analysis where you start off with scattered data points and just by controlling for a certain factor suddenly come up with a nice clean line. It really speaks to what an efficient production process LEGO must have if all those other factors like unique molds, recolors, deco, etc. ultimately amount to such a tiny variation in pricing.

[kodlovag]

OK, I made these graphs to check the PPP vs. PPG, because I like to visualize things, I underastand things better in this way. I used the same sources as @nerdsforprez, USD RRP prices are from Brickset, weights are from Bricklink. Weights are containing box and instructions too, but it would be too much work to calculate the net weight. On the other hand, all sets have box and instructions, and bigger sets have bigger box and lomger instructions. So I think weights are still proportional.

First the price vs piece:

I checked several special elements, like the long linear actuators, the pullback motors, the pneumatics and the Power Functions in the sets. I found, that only PF counts significantly in the price. The PPP scatters a lot, but only the PF sets pops out.

Then, the price vs. weight:

The graph is much more clear. What we can see? (42099 is not included in the graph, because there is no weight data available)

• Inflation: 25-30% price increase is visible over the 15 years.
• Big sets are cheaper.
• Only several overpriced sets. All heavily motorized, or licensed. I think Porsche is also overpriced compared to big sets.

Edited August 11, 2019 by kodlovag
42099

[nerdsforprez]

Great Job!  This is where it gets fun.  Have you tried to do a trend analysis comparing a linear vs. logarithmic trend for the second graph?  Your trend seems to perhaps better fit a logarithmic trend versus linear. Especially given that one yellow guy, between 3000-4000 grams, at about 125$ dollars USD.   What set is that?  Extremely heavy but cheap.  Looks to be in the 2005-2007 time frame.  Also, if your conclusion that larger sets get relatively cheaper, then yes, you are looking at a log. trend versus linear. 

If you do the same graph, but set a trend analysis you can both compare whether a regular linear versus log. would best fit the data, but you will also have a set, average line through the data points and you can then visually see (or even calculate) how much each data point (set) deviated from the central line. 

I agree it is fun to look at data this way.  Couple of things.  In the dorky trends I have examined in the past, your interpretation of sets being cheaper as they grow larger is correct.  Check out this post I made years ago:

I believe this is a function of what my piece versus lot count post above is all about.  Larger sets, on average, have a larger ratio of overall parts to new parts than small sets. Think if it this way.  In a set of, say, only 200 pieces most of them will be unique.  Other than pins, there will not be many redundant parts.  However, in really large Technic sets, the exact opposite occurs.  There is tons of redundancy.  Large Technic sets usually have hundreds of pins (which require little to no ABS - they are so light and small), and even other elements come in groups of 10's, 20's, etc.  In fact, in larger sets, it is actually rare to find an element that is not duplicated at least once.

So, TLG needs less molds, relatively speaking, for large sets versus small sets.  Summarized, there are a lot less unique parts relative to overall piece count in large sets than small sets and that makes them cheaper to produce. 

I think it is really cool to have someone look at similar data, in a similar way you did years ago, and verify similar results. 

One last lesson, and I fully expect to get some blow-back from this, is that I know Technic fans gripe about the high prices in the last years of some of our large sets.  Sets above 2000 pieces is now not all that big news whereas prior to 2012 it had never occurred.  And folks are complaining about the prices.  Well, I hate to break it to folks, but looks like TLG is pricing Technic sets on a logarithmic curve, meaning that the larger the set, the cheaper it is (relative to smaller sets).  There is a hidden cost effectiveness in producing large sets, and they are passing the savings to us.  Look at @kodlovag's last graph. If you draw a straight line through the red marks, as you approach 4 and 5,000 gram sets only the Porsche and Bugatti really fall in line. Many of the other sets fall below the line meaning they could actually be more expensive (course there are many that are more expensive as well.). 

Oh and I did find one error - though I didn't look for others.  the BWE looks like an expensive set based on weight but somewhat a rock star at one 7 cents per piece for PPP. I think you have the weight wrong.  It looks to be nearly 6000 grams, not 4,000.     

Crap.... I just realized that BL does have manual weights.  So... theoretically it would be possible to factor out manual weight.  I think they also have box weights for many sets.  Crap... I can see myself going down a fox hole here....

We'll see.  I have on a trip right now, but will be returning home soon.  Not sure how much time I will have to look at this.....

[Bartybum]

There's a whole bunch of interesting plots you could obtain from historical data. I'd be really interested to see two 3D plots of sets since the release of the 8275 PF bulldozer:

- PPG (adjusted for inflation) vs model weight vs release date
- PPP (adjusted for inflation) vs piece count vs release date

These two plots alone would allow a comparison of six important distributions/metrics when viewing orthogonal planes:

- PPG vs model weight (how the amount of resources affects price)
- PPG vs release date (evolution of PPG)
- model weight vs release date (evolution of set weight)
- PPP vs piece count (how the piece count affects price - effects of mass production)
- PPP vs release date (evolution of PPP)
- piece count vs release date (evolution of piece count)

It'd also be nice to see both plots superimposed on one another to view how PPP and PPG compare.

Colouring data points by set size (piece count), PF and pneumatics could also show how pricing is affected by each.

Edited August 11, 2019 by Bartybum

[kodlovag]

6 hours ago, nerdsforprez said:

Especially given that one yellow guy, between 3000-4000 grams, at about 125$ dollars USD.   What set is that?  Extremely heavy but cheap.  Looks to be in the 2005-2007 time frame.

It is the flagship set 8285 Tow Truck from 2006. It has big wheels, metalized parts and pneumatics, and these do not reflected at all in the price. Only PF increases the price, and only when there is more than a battery box and an M motor is present in the set.

6 hours ago, nerdsforprez said:

compare whether a regular linear versus log. would best fit the data

I think there is no physical or economical reason, why a log would fit better. A logarithmic curve bends down much faster than our data points. Anyway I tried, and it fit worse. A log vs. log fit tells nothing new, just compressing the upper segments, and it is harder to see, that big sets are cheaper. Any slowly bending curve could fit better, but there are too few data points in the upper segment. I would simply propose to use two linear fits, with a break point around 2000-2300g. It will work fine.

What we actually see here (I think) is the "volume" pricing. The price of a set is composed from volume related (like weight of plastic, longer instruction, etc.) and fixed costs (like box design, standard process of accepting a set design(kid proof?, using bricks currently in production? etc.). And some others in between, like the number of lots described by @nerdsforprez... The fixed costs increasing the price of a small set faster, than a big set. On the other hand small sets are selling faster, than big sets, so the fixed costs are distributed better. What I want to say, there is no way to tell the exact shape of the curve. Maybe TLG could, but we have no chance. Therefor a simple broken line is good enough.

6 hours ago, nerdsforprez said:

Oh and I did find one error - though I didn't look for others.  the BWE looks like an expensive set based on weight but somewhat a rock star at one 7 cents per piece for PPP. I think you have the weight wrong.  It looks to be nearly 6000 grams, not 4,000.

True, I'm correcting it. It is 42070, I confused weight with piece count when manually marked the point.

Edited August 11, 2019 by kodlovag

[kodlovag]

5 hours ago, Bartybum said:

There's a whole bunch of interesting plots you could obtain from historical data. I'd be really interested to see two 3D plots of sets since the release of the 8275 PF bulldozer:

- PPG (adjusted for inflation) vs model weight vs release date
- PPP (adjusted for inflation) vs piece count vs release date

These two plots alone would allow a comparison of six important distributions/metrics when viewing orthogonal planes:

- PPG vs model weight (how the amount of resources affects price)
- PPG vs release date (evolution of PPG)
- model weight vs release date (evolution of set weight)
- PPP vs piece count (how the piece count affects price - effects of mass production)
- PPP vs release date (evolution of PPP)
- piece count vs release date (evolution of piece count)

It'd also be nice to see both plots superimposed on one another to view how PPP and PPG compare.

Colouring data points by set size (piece count), PF and pneumatics could also show how pricing is affected by each.

I have all the data required for these graphs (except inflation correction). But it took me about 6 hours to collect the data, draw the graphs and analyze them. So it's quite time consuming, but I will try to make them later.

8275 released in 2007, so it is included in the data. I went back to 2005, because that was the time when TLG migrated to the new beam/liftarm design, and the darkest year with only 5 sets released. Definitely many things changed at that time.

I already have the colored data for PF, pneumatics, long linear actuators and pullback motors. Every pullback sets are 20 USD RRP, no exception. If a set contains pneumatics, then both PPG and PPP are in the trend line, only exception is the 42080 with the new pneumatics +PF. Only heavily motorized and licensed sets can jump out, and thats why I only posted the colored PF chart. And marked the Porsche and the Bugatti. I think the licese fee is about 15% of these sets.

[kodlovag]

8 hours ago, Bartybum said:

There's a whole bunch of interesting plots you could obtain from historical data. I'd be really interested to see two 3D plots of sets since the release of the 8275 PF bulldozer:

- PPG (adjusted for inflation) vs model weight vs release date
- PPP (adjusted for inflation) vs piece count vs release date

These two plots alone would allow a comparison of six important distributions/metrics when viewing orthogonal planes:

- PPG vs model weight (how the amount of resources affects price)
- PPG vs release date (evolution of PPG)
- model weight vs release date (evolution of set weight)
- PPP vs piece count (how the piece count affects price - effects of mass production)
- PPP vs release date (evolution of PPP)
- piece count vs release date (evolution of piece count)

It'd also be nice to see both plots superimposed on one another to view how PPP and PPG compare.

Colouring data points by set size (piece count), PF and pneumatics could also show how pricing is affected by each.

I cannot do a 3D plot, however I think my original colored charts were a kind of 3D plots. But I made these six comparisions, but I have no idea what we can learn from them. Prices are adjusted with US inflation data, and 2019 equivalent prices are displayed. I'm not really sure about the method to use US inflation, because the manufacturing is in the EU, so it would probably be better to use some kind of EU inflation (?) and EUR/USD conversion rates. But it should not be too different.

The orange spots are the sets with Power Functions.

My original charts told me more, I cannot see the added value from these ones. I think I will not create more charts. Let's discuss these ones.

[Bartybum]

9 hours ago, kodlovag said:

My original charts told me more, I cannot see the added value from these ones. I think I will not create more charts. Let's discuss these ones.

Nice work, could you also highlight the pneumatics sets in green (if both pneumatics and PF, do a green dot with a red outline), stick a number next to each PF set that describes how many PF components (and pneumatic cylinders/pumps) it includes, and label the outliers with their respective set numbers? I think you’ve also forgotten to highlight 42082 as PF.

Reason I ask is that I’d like to see how the inclusion of PF and/or pneumatics throw the usefulness of PPP and PPG for a loop, and how that scales depending on how many PF/pneumatics components are added.

The first two graphs indicate to me that both PPP and PPG seem to be equally appropriate since they exhibit really similar behaviour. Both remain constant enough, but really begin to break down once you either a) include pneumatics and/or PF, or b) go below 1k pieces.

The time history graphs should indicate how constant or varied the usefulness has been over time, and whether Lego has been upping the PPP or PPG or not.

The comparison of PPP and PPG shows me that both are pretty much equally appropriate to use to determine the worth of a set.

Edited August 11, 2019 by Bartybum

[nerdsforprez]

On 8/11/2019 at 5:15 PM, Bartybum said:

Nice work, could you also highlight the pneumatics sets in green (if both pneumatics and PF, do a green dot with a red outline), stick a number next to each PF set that describes how many PF components (and pneumatic cylinders/pumps) it includes, and label the outliers with their respective set numbers? I think you’ve also forgotten to highlight 42082 as PF.

Reason I ask is that I’d like to see how the inclusion of PF and/or pneumatics throw the usefulness of PPP and PPG for a loop, and how that scales depending on how many PF/pneumatics components are added.

The first two graphs indicate to me that both PPP and PPG seem to be equally appropriate since they exhibit really similar behaviour. Both remain constant enough, but really begin to break down once you either a) include pneumatics and/or PF, or b) go below 1k pieces.

The time history graphs should indicate how constant or varied the usefulness has been over time, and whether Lego has been upping the PPP or PPG or not.

The comparison of PPP and PPG shows me that both are pretty much equally appropriate to use to determine the worth of a set.

Both are appropriate, but PPG still outperforms in its relationship to price.  Not prediction, but relationship. Graphs are fun and visually help with literacy, but overall numbers do not lie.  Weight, compared to piece count, on average, better accounts for set price.  In a sample of 50 sets over the last 6 years, it accounted for for over 11% more of the variance of cost than did piece count.  Also, by adding more sets with a heavy load on PF elements, this discrepancy widens.  As @kodlovag pointed out, there is also a restricted number of sets with high PF elements and piece count.  What this means is when any new sets are added that fit these criteria, such as 42099 and 42100, this discrepancy will widen even further.   When sets without PF elements are added, it narrows. 

Very fun stuff.  Special thanks to @kodlovag for all his work. Regrading the log vs. linear interpretation, in laymans terms what this means is that PF elements (in amounts described by him) and other sets really are different animals altogether.  Perhaps not appropriate to lump them together in a PPP camp when considering sets. 

One can really get in the weeds with all this.  It is fun and academic, but really I wanted this post to be practical and I think we have some practical solutions or recommendations.  We still get a large number of "what set should I buy" posts here typically from new users.  While this is fine I guess I would provide some simple, but more objective guidelines. 

In summary:

1) Take care of your subjective needs and wants first.  No one can dictate those for you.  Desires/wants of helicopter versus supercar, truck and trailer versus farming equipment, etc. - no one is going to be able to take care of that for you. 

2) get opinions from others in terms of building experience, etc. 

3) take a good look at parts list, - obviously. 

4) - when, and only when, you have really boiled sets to a few options, and you just want to make an economic decision (other factors are taken care of) - know that weight of a set more accurately accounts for price than does piece count.  This is especially true for large, over 2,000 piece sets with many PF elements, but obviously true for ANY set with tons of PF stuff.  8043, 42099, 42030.   Also understand because sets are including much more than standard ABS (wires, metal, etc) price prediction will be more variable and unpredictable. 

This also goes for snap-judgments on sets based on only PPP.  Not only would I not recommend it, but especially on sets with high PF (or other systems, say Control+) elements AND high piece count.  PPP is simply not going to be an accurate measure of anything valuable.  COmparing a PPP value to say regular primarily ABS sets is like comparing apples and oranges. 

Lastly, I am very clear not to say this is a way to "determine the worth of a set" - you will get all kinds of subjective comments.  Better put, I would say this whole post, of PPP versus PPG are only measures that account for the variance of a sets cost.  Whether or not that constitutes as "value" or "worth" will be up to each and every reader.  @kodlovag listed a very good one at the first of this thread, there are also many others. 

[Bartybum]

1 hour ago, nerdsforprez said:

know that weight of a set more accurately accounts for price than does piece count

I'd really amend that word to consistently, it seems more technically correct

[nerdsforprez]

Not sure why.  Consistency relates to reliability and that is not the discussion here.  I think (hopefully) that this has all been about is validity, not reliability. 

Reliability has to do with the temporal stability of a measure, validity its diagnostic accuracy. 

[Bartybum]

I mean, the outcome of this entire study has to be based on statistics. Your question poses two linear problems:

"How much does every kilogram cost?"

"How much does every piece cost?"

Both of these require that we model our data using a single linear regression for each, then see which line is better represented by its respective data, i.e. which line has data that lies closer to it, i.e. which line has the lowest R^2 value.

Here, accuracy is directly a result of the internal consistency of each data set to its respective best fit.

Obviously, modelling this as a single linear curve is a bad idea, because as we've directly seen with the data, the inclusion of PF and pneumatics throws everything for a loop.

In this essence, I've gotta agree with Kodlovag that my curves aren't that valuable compared to his.

Edited August 13, 2019 by Bartybum

[nerdsforprez]

Ummm, I think we are getting our wires crossed a bit.  Its not necessary to get too far in the woods with statistics here, but based on some of the things you mentioned, I can see there may be some misinterpretations. 

I think you are misinterpreting internal consistency.  Data sets of an outcome variable don't have internal consistency.   Internal consistency has to do with test or measurement construction, not outcome variables as nature provides.  Low R-squared values are not desired outcomes. High are, but not low. 

Anyways, validity encompasses reliability.  You cannot have a valid measure of something without it being reliable.  However, the converse is not true.  You can have a reliable measurement without it being valid. 

Ex: scale for weight cannot be a valid measurement of one’s weight without it being temporally stable (reliable). However, a scale can be very temporally stable (again, reliable) but off by ten or so pounds and therefore invalid. 

So, making a specification of temporal stability is redundant IMO.  Stating accuracy of weight in accounting for cost implies temporal stability. That is the reason for my word usage.  But either way, if it makes better sense to you to use something that indicates reliability over validity.  Great.  Whatever works 

 

Also, as far as the questions you pose in quotations - those are not my questions at all.  They may be your interpretations.  Fair enough -   but they were not the questions I posted.  Succinctly put; my question was:

We have the cost of a set.  Many folks think that PPP most accounts for this.  That is not correct.  PPG accounts for cost better.  So, is it time to abandon the PPP?  Based on the input from others.  I don't think it is.  Not because PPG doesn't outperform PPP is accounting for cost, because it does, just in that there are other reasons to use PPP for. 

If others want to branch out and examine the set of all possible variables that may account for set price - fair enough.  That is great.  But that is not my endeavor nor the reason for the post.  As put out by another contributor to the thread, I think the idea works best by some "quick and dirty" measure for accounting for set cost.  Nothing elaborate.  I am a utilitarian at heart, therefore I focus on that which may actually be put to use, that which may be functional.  No one is really going to use an elaborate formula to account for set cost.  Something that accounts for 86% of set cost variance, especially that which can be done in just a few seconds, is pretty darn good. Something that accounts for like 94%, which can be done almost as quickly, IMO, is even better. 

 

 

 

 

 

[Bartybum]

18 minutes ago, nerdsforprez said:

Low R-squared values are not desired outcomes. High are, but not low

Sorry sorry, I meant R^2 closest to 1. Entirely my bad, I got R^2 inversely confused with relative error or whatever the stats term is.

Edited August 13, 2019 by Bartybum

[Bartybum]

Okay, I think I understand. Since the data varies wildly, none of it is really consistent.

Edited August 13, 2019 by Bartybum