TOPIC:

A long time ago I theorised that the Sand Rover and Holiday buggy would become the "new" SRB in terms of collectability, prices would rise slowly, but in a non-linear fashion, going exponential and peaking at silly values just before the announcement of re-releases. As we now know, there was no spike in prices, and the re-release was barely a re-release at all (and it came very quickly after my hypothesis).

What this did give me though was a set of data relating to prices over time - so I presented that (in another place), complete with graphs with near horizontal lines, and demonstrated that the re-release had no effect on prices, and that the value of both buggies was remarkably flat.

The reason I mention this is this: I know some people read as much as Champion the Wonder Horse, but the post included graphs. Graphs with flat lines. It appears that those who read it was unable to grasp my points as stated in the text - and also unable to interpret a graph with a flat line as meaning _no change_ . I know this because the comments were of two distinct types: firstly "yeah it sucks Tamiya rereleasing everything they shouldn't do it cause it makes everyones cars worth less", and secondly "yeah prices are climbing mental".

To summarise, I'm not overly hopeful about this post being that different to the casual reader, though obvious I'd hope the average Tamiyabase contributor is better informed



First, a few notes about the data:

Wild Willy M38 SWB averages: based on 15 sales July 2012 to December 2014
Wild Willy M38 LWB averages: based on 83 sales September 2011 to December 2014
Willy's Wheeler averages: based on 24 sales October 2011 to December 2014

THe first thing anyone's going to ask about those numbers is why so few? The thing is, for a price to be valid it has to have been publicly visible - anything traded on a forum, over facebook messages and so on can't count, you may know the asking price but there's no way of knowing the final price agreed. Similarly, cars up at "best offer" prices on eBay are no longer auditable, as eBay has been hiding the actual final value for some time. In addition, although most people when they see a car up at a buy it now price of X hundred whatevers assume that when it is no longer visible, it "must" have sold for that - when that simply isn't true, in a lot of cases it will have sold for a lot less outside of eBay, and that price simply isn't auditable.

Consequently, the only prices used are visible eBay final values, and I can't guarantee I captured every single M38 SWB, LWB or Wheeler sale on (English speaking) eBay in the relevant period.
Because the data sets for the SWB and Wheeler are so small, the methodology behind the graphs could be seen as somewhat flawed - the data includes all sales of complete cars (meaning having a body, chassis and at least 3 wheels) were included regardless of condition from junker to NIB (albeit examples at either end, especially the good end were exceptionally rare) as excluding them would have reduced the data set further.

The averages are the arithmetic mean of the price on a sale date, plus all the preceding sales. For example, the average on the 5th sale is the mean of the first 5 sales, the 15th value is the average if the first 15 sales and so on. This has the advantage that any volatility in values is damped out, but can have the disadvantage that as the size of a data set increases over time, trends become less obvious. I don't consider that to be a problem in these small sets.
Note the green vertical lines on the graphs: I'd consider anything to the left of those to be unsafe for the purposes of drawing conclusions, anything to the right as sound. Such measures are necessary given the size of the set and the volatility (variation in prices adjacent in time). For example, sufficient "damping" of the LWB prices has occurred by the beginning of July 2012.



Note that all the values used in working out the averages included local postage as stated by the seller. This seems reasonable to me, as if you're buying an RC car, most of the time you have to pay that cost too, although it should be noted that this will have had a small inflationary effect on prices. Due to the way I've summarised the data it's no longer possible to strip out the postage costs for all the data.

Where no postage cost was stated or the car was pick up only, it was excluded.

This does mean that the value to sellers is not going to be anything like the values stated here - account has to be taken of listing fees, final value fees, Paypal fees, packing costs, subsidising the ever increasing cost of posting to avoid seeming to be guilty of usury or whatever the equivalent sin relating to shipping charges is, ... I think we're looking at least a 15% reduction, and of course the fact sellers don't get to keep the postage cost either.





So what does the data - and the graphs - actually tell us?

Wild Willy M38 LWB: prices have risen from an average of 140 GBP three years ago, to 160 GBP now. This is in line with expectations, although the increase or 15% may be more modest than expected. 5% per annum sounds like quite a good return though, doesn't it?

Wheeler: Prices remarkably stable around the 310 GBP mark. This flies in the face of "common sense", which states that all RC prices are rising, and especially rarer cars like this.

SWB Willy: TBH this result stretched my credulity to breaking point: apparently the average price of a short wheelbase Willy, the rarest of the 3 cars here has fallen in the last two years from 235 GBP to 185 GBP - a premium of only 25GBP over the much more prevalent "Long" wheel base version. I delved deeper, but there are no NIB cars skewing the data, nothing to suggest that all the good ones were sold early and poorer ones make up the sales now ... this apparent 21% fall in two years I can only put down to the small number of figures available - I suspect all that's happening here is there aren't enough figures for the average to have become stable yet.


TBH the only conclusion I'll really stand behind is the modest increase on LWB prices (and to a certain extent, the Wheeler figures): outstripping inflation, but only modestly so. But with our sensible trousers on, didn't we really know that anyway?

Jonny, I absolutely love your in-depth posts!

I have many times thought about keeping statistics on sale prices on ebay, but I have never taken the time to do it.

One comment to your calculations:
Using the mean value as you do, you will see a flattening of the curves when time goes simply because the set grown and single sales will change the average less and less as the set grows.

I would say a "running" average would be better. You could have the average of the last 5 sales or the average for the last x months. That would give a more correct picture of price changes, over time.

-Lars

larbut wrote:

Jonny, I absolutely love your in-depth posts!

I have many times thought about keeping statistics on sale prices on ebay, but I have never taken the time to do it.

One comment to your calculations:
Using the mean value as you do, you will see a flattening of the curves when time goes simply because the set grown and single sales will change the average less and less as the set grows.

I would say a "running" average would be better. You could have the average of the last 5 sales or the average for the last x months. That would give a more correct picture of price changes, over time.

-Lars

I take you point, I even mentioned that damping effect (9th paragraph) ... the thing is, because the sets for the SWB and Wheeler are so small, you got to do something to iron out the volatility to get a meaningful curve, and an arithmetic mean of an increasing set size is the most transparent way of doing that, not doubt there are other ways of doing it but everyone knows what an average is ... and once you've used that, you have to use the same method for the LWB data, other wise you're comparing oranges and bananas.


I didn't just choose the arithmetic mean because it was the first thing I thought of, it was carefully chosen. I've created some graphs with your suggestions for the LWB data set - the first one, the rolling 5 average (apart from the 1st 4 points, obviously) shows a lot of volatility, 3 monthly average does too. In both cases I don' think it's helpful as it appears to shows that the current average buying price for a Wild Willy is over 300 GBP - thanks to one recent NIB sale. If I take that out that figure, then I have to start taking out other anomolous data too - so the results become massaged in a hidden way.

Even a 6 monthly average is flawed - no-one is paying 220 GBP for an "average" LWB Willy.

wow great explanation johny. could you make another statistic for other famous tamiya cars, it would be usefull for vintage owner to see how the price goes from time to time.

bananarc wrote:

wow great explanation johny. could you make another statistic for other famous tamiya cars, it would be usefull for vintage owner to see how the price goes from time to time.

Not so much ... and for several reasons...

Apart from the Willys I'd pretty much stopped collecting data over the last 18 months, some cars I stopped tracking as early as spring 2013, some lasted until Autumn this year, partly out of boredom, but also it being a Sisyphean task - for the reasons I discussed in the first post, even with a lot of work, meaningful conclusions are difficult to draw - the big problem is that RC collecting may be a large and mature market but it's anything but transparent - the only source is a subset of eBay sales.

I feel I've been quite clear on this, but to restate: fully 4 of the 6 graphs so far is this thread are WRONG - one through not enough data, 3 through implying an incorrect conclusion despite a methodology that initially seems reasonable.

I can't help but feel you've somewhat missed the point - I'm saying this is _not working_. Even if you (or one) looks past the flaws with the data collection & trying to present it in a meaningful way, AND the conclusion can be considered reliable, the data shows that prices are stable, or have a modest trend up or down, depending on model - this is not how people perceive how the market is going so see what they want to see in the results.

As a tool for valuing a car it's really kind of flawed, and for the rest of my data even more so as it's purely historical.

That said, I do have a bundle of figures & as this is the only outlet they're likely to have, if any one is really desperate to see graphs based on everything I've already said, these are the cars I have data on, in descending order of reliability:

Meaningful data set:
Biwig
Blackfoot
Cheetah
Clodbuster
Falcon
Fox
Holiday Buggy
Monster Beetle
Ranger
Rough Rider
Sand Scorcher
Sand Rover
Super Champ
XR311
Mardave Cobra
Mardave Meteor


Marginal:
Mondeo
Porsche 959
Mardave Apache
Striker
(+ more)

Very marginal:
Pajero Wheelie
(+ more)

Inadequate:
Blazer
Bruiser
Spital Trial 12
(+ lots more)


If you don't see a car in those lists I may have data on it, I can tell you just how unreliable it will be