rossr
WS Benefactor
Posts: 1,838
|
Post by rossr on Feb 26, 2019 22:14:57 GMT -8
This update focuses on estimating the production numbers of 62MAS pieces – from the earliest April 1965 6217-8000 to the last recorded monthly stamping in October 1967. This updates Table 1 and uses additional data I have gathered and received from individual members. A particular thanks to member dasher , who sent through his own data set that he had put together over some time. The other important pieces of information have come from the comments, interpretations and suggestions on this thread, which have forced me to query my own earlier interpretations of the data. Though I have collected additional serial numbers from sales since the previous update, I have not included that information here. An important observation from this work is that it appears that the serial numbers (the last five digits of the stamping) were re-started for each year of production. It may seem obvious and trite, but it is fundamental to the estimation. The broad groups are; April-May ’65 8000 models; the January-October ’66 8001 models; the February-October ’67 8001 models; and the 1967 Horseshoe models. I have not differentiated between the light and heavy stamping of pieces in 1966. I have no evidence to suggest that the Run 1 ’67 Dolphins and the Horseshoes were part of the same serial number batch, so I’ll assume they are separate serial number groups. Another key assumption I make is that the appearance of 62MAS serial numbers (in sales, member posts and so on) across all production runs (1965 8000s, 1966 Dolphins, Run 1 1967 Dolphins, Run 2 1967 Dolphins, Suwa Horseshoes 1967, Daini Horseshoes 1967) has been a random sampling over the entire pool of pieces that were manufactured. I suspect that there may be a bias toward the better Dolphin backs when the pieces are shown off in glamour shots (given the better longevity of the stamping), but I suspect that the frequency of appearance is independent of the production run group. Condition and group should affect the price characteristic, and we are not concerned with that here. Perhaps collectors might hold on to the Daini pieces more tightly than a ’67 Dolphin, but again, everything is for sale at the right price, and I suspect that the infrequent appearance of these is more a function of their low total numbers, rather than their desirability. I tested whether there was a discernible pattern between successive serial numbers for each identified run. I looked at the gap between successive serial numbers to understand whether the variability in the gaps were consistent with a random selection, or whether there was lower variance due to some systematic influence. I couldn’t reject the hypothesis that the gaps are indeed random. A member who knows something about statistical methods could add some valuable suggestions and corrections as to how the data might be better analysed. The serial numbers show that there were no more than 5000-6000 Run1 Dolphin-backs produced, and we have 51 of these in the 62MAS pool. There were likely fewer of these produced than 5000-6000 given that we believe that Seiko stamped other watch models with serial numbers that interlaced those of the Run 1 62MAS pieces, but it is pretty well impossible to even estimate as to what the characteristics were of this practice. Thus, I am anchoring the estimation to the assumption that the number of 62MAS pieces in 1967 Run 1 is 6000. As I mentioned in an earlier post, there only 55 samples drawn from the Run 2 Dolphin group, and though the serial numbers span a larger range, we would expect the size of this group to be not dissimilar to the size of the Run 1 group – of which we know the maximum potential size (6000). Table 2 updates the production estimations that previously appeared in Table 1. I’ve just given a single estimated number rather than a +/- range – though I should probably do that. I have a few more charts to add that I will do in a later post. The Excel sheets I've been using to put this all together are certainly looking a little messy now! So here it is. Comments, observations, suggestions and criticism welcomed. Table 2: Estimation of 62MAS Production NumbersNote that I believe this to be at the higher end of any production estimation. How many of these still exist is anyone’s guess. Regards, Ross
|
|
rossr
WS Benefactor
Posts: 1,838
|
Post by rossr on Mar 7, 2019 0:04:47 GMT -8
Jon passed on another serial number from 1966. I'm still trying to grab them as they come past. I don't have that particular one, though I have a couple close by either side. I imagine that the estimated numbers could adjust as representatives from each pool come in, but it is still grounded in the assumption of there being 6000 from Run 1 in 1967. I did replicate the 1967 charts with data from 1966, though a pattern is not particularly distinct. Chart 18 does show a couple of large gaps which is interesting. Chart 18 And when ordered by month and then serial; Chart 19My spreadsheet is a now a bit of a mess. I need to re-format so I can update the charts more easily when new data comes in. I still haven't updated the charts and estimates with the new data that I have gathered since the updates in October. Regards, Ross
|
|
|
Post by bertnet on May 1, 2020 1:59:51 GMT -8
I have also a serial number from 6217-8000 /1965 N°5506588
|
|
|
Post by bertnet on May 1, 2020 2:03:22 GMT -8
Mine with horseshoe 6217-8001/ 1967 N°7502722
|
|
|
Post by 3Pedals6Speeds on May 2, 2020 5:45:39 GMT -8
If you'd be willing to share the data, I'd be happy to look at helping visualize the data. As the semester just wrapped up and I have turned grades in, I have time on my hands to play around!
|
|
rossr
WS Benefactor
Posts: 1,838
|
Post by rossr on May 2, 2020 17:53:13 GMT -8
If you'd be willing to share the data, I'd be happy to look at helping visualize the data. As the semester just wrapped up and I have turned grades in, I have time on my hands to play around! Yes, not a problem. I’ll tidy it up and send it through. I think there is an opportunity to look at things like the gaps between the serials to consider whether they are part of a pattern, and to consider the confidence of the estimation with the small sample size I have.
|
|
|
Post by yorkiesknob on May 2, 2020 20:37:54 GMT -8
Please add mine. 6217-8001 S/N 7519147 with a dolphin back. Cheers Tony
|
|
|
Post by a1topdog on May 3, 2020 3:31:36 GMT -8
I have an early watch 6217-8000, dolphin back. Due to fading and my poor eyesight it's hard to make out the last 2 digits of the s/n but it looks like 5511894
I would upload a photo, but can see an option to do that
|
|
rossr
WS Benefactor
Posts: 1,838
|
Post by rossr on May 4, 2020 2:08:07 GMT -8
Thanks for the new data points. I'll try to do an update in the next couple of days.
|
|
rossr
WS Benefactor
Posts: 1,838
|
Post by rossr on May 5, 2020 2:31:42 GMT -8
I have an early watch 6217-8000, dolphin back. Due to fading and my poor eyesight it's hard to make out the last 2 digits of the s/n but it looks like 5511894 I would upload a photo, but can see an option to do that Very close to my 6217-8000 @ 5511878.
|
|
|
Post by 3Pedals6Speeds on May 12, 2020 13:51:54 GMT -8
Ross was kind enough to share his excel file, and I've been playing around with the visualization of the data. I have 1966 Dolphin caseback watches 'done' (well, an idea of a draft) and I noticed that the numbers rise within a month, sometimes even continuing the next month (or 2 or 3), and at other times seem to reset each month. So, what I have here is each 'run' of caseback numbers (what is plotted is the last digits of the caseback #, ignoring year and month as all are 1966, and months appear as color), and when they reset, I give the next run a different marker shape. Each different color within a 'run' is a different month's color. Note for instance that Jan, Feb, March, and April appear to all be a single 'run', while May is a complete 'run' of its own, which clearly resets the next month, in June. I'm interested in any feedback as to the representation of the data this way. As a note - more data would obvious increase the knowledge we can take from all this, and Ross does have additional observations you've provided to him that he's coding into the database which I'll then add to this.
|
|
|
Post by 3Pedals6Speeds on May 12, 2020 15:25:00 GMT -8
1967 Daini Horseshoe Caseback watches: and 1967 Suwa Horseshoe Casebacks:
|
|
|
Post by 3Pedals6Speeds on May 12, 2020 15:48:36 GMT -8
1967 Dolphin Casebacks:
|
|
|
Post by 3Pedals6Speeds on May 12, 2020 16:00:44 GMT -8
1967 Dolphin Casebacks without the 2 extreme outliers (huge caseback numbers distorted the graph):
|
|
|
Post by huangcjz on May 13, 2020 0:46:17 GMT -8
rossr , is the factory that the case was made in taken into account? For example, the "X" marked on the inside of the case-back shown by HiBeat here.
|
|
rossr
WS Benefactor
Posts: 1,838
|
Post by rossr on May 13, 2020 16:41:47 GMT -8
rossr , is the factory that the case was made in taken into account? For example, the "X" marked on the inside of the case-back shown by HiBeat here. Unfortunately not.
|
|
rossr
WS Benefactor
Posts: 1,838
|
Post by rossr on Jun 25, 2021 16:22:16 GMT -8
Hello everyone. This post updates previous posts on this topic that were concerned with the serial numbers of the Seiko 6217-8000 and 6217-8001 150m dive watches that were produced between 1965 and 1967. The information and conclusions I reached back then are now superseded by the information in this post. Since my last update on this topic, which included data up to a cut-off date of March 2018, I have collected serial numbers that have appeared in a variety of places – eBay, Yahoo Japan, various forums and Instagram. These add to the serial numbers I had collected in the previous iteration of this project. The samples are those of which I can be reasonably certain as to the specific serial number. There are quite a number of samples I have collected that show unclear or totally/partially obscured serial numbers, and others for which the case back was not shown at all. I have not been able to use these records for this serial number exercise. Thanks to those members who sent me their serial numbers, and again, a mention to member Dasher who had also collected a number of serials. I have split this into several parts. Part 1 deals with the small crown 6217-8000 from 1965 and is probably the simplest series to interpret. Part 1: The Small Crown 6217-8000 The small crown 6217-8000 was produced in April, May and June 1965. These pieces are quite scarce, and collectors seem to recognise this, as good examples of the 6217-8000 tend to fetch a premium in the auction market. I have 55 serial numbers for these pieces. The earliest serial number seen is 5400735, and the latest is 5612440. Note that I have a single observation of a June 1965 62MAS piece. The serial numbers are plotted sequentially in Figure 1. Figure 1 – Serial Number Observations (or samples) Compared to earlier iterations of this chart, a number of the gaps have closed, or have reduced in size. Note that the serials increment smoothly, without unexpected gaps when moving over the April/May and May/June monthly boundaries. The gap appearing prior to the first observed sample (735) might be questioned given that this gap references back to the start of production at 0000 – and who knows what was going on in the early stages of the production of these watches. I did an F-test for population variance between the observed gap distribution and that from a randomly generated dummy set, that suggests that is not unreasonable for gaps of these sizes to exist if 55 samples were randomly drawn from a population of this size. I am just suspicious that large gap presents at the front end of the serials. Is there any evidence to suggest that the serial numbers allocated to the 6217-8000 were not incremented by a single digit? Were all numbers populated between 0 and 12440? It is obvious that Seiko produced watches other than the 62MAS in April and May 1965. To that effect, I have collected images of a small number of King Seikos and (various) Seikomatics that have casebacks with 54 and 55 date prefixes, and serial numbers within the range of that spanned by the 6217-8000. Did these serial numbers displace those of the 6217-8000, or were they produced in a separate line, with a separate set of incrementally advancing serial numbers? I doubt if it can be statistically proven - from the samples I have collected - as to whether the drawn samples are from a population solely of 62MAS pieces 0 to 12440, or whether they are from a population 0 to 12440 that has other models randomly distributed within that serial number range. My strong suspicion is that the 62MAS numbers are fully populated to 12440 (and probably slightly beyond). If this is the case, it would be expected that other Seiko watches (from lines such as the King Seiko) would have full serial numbers identical to those of the 6217-8000, and that these should be observed from time to time. The chance of coming across a serial number identical to the observed 55 samples (out of the population of 12440) is low, but not zero and so should be seen on occasion – if people were looking and comparing. In an earlier post on this topic, I mentioned recalling a post 5-10 years ago (on SCWF) that included a picture of two watches with the same full serial stamping (year, month and serial). Unfortunately, I do not remember the models, and I cannot find this image. The other thing is practicality at the time of manufacture. Would casebacks from different models really be grouped together and directed through the same serial number stamping machine to be incrementally stamped, and then be sorted again into their separate lines for assembly after stamping? This seems unlikely. How many first generation 62MAS (6217-8000) pieces were produced in the April, May and June 1965 period? It may seem simplistic and obvious, but I cannot find any evidence to suggest that there were any less than 12440 pieces produced over this time. The table below shows an estimated split of production over the months, and a total estimated production. The interpolation is done by splitting the gap between the observed monthly serial limits. The earliest serials are extrapolated back to 0, and the latest by replicating the difference between the May to June limits. Table: Estimated Small Crown 6217-8000 Production Note that the ratio of observed samples to assumed production for April and May are similar. The June ratio is based on a single point and an extrapolation into the yonder. The next post (Part 2) - when I can complete it - will deal with the 1966 6217-8001 set. Comments and observations warmly welcomed. Regards, Ross
|
|
|
Post by gerald on Jun 26, 2021 2:42:55 GMT -8
Super interesting - I have a sense as to just how much work is involved in an exercise such as this one, having done the same for the first Grand Seiko.
Just one general observation -
"There are quite a number of samples I have collected that show unclear or totally/partially obscured serial numbers, and others for which the case back was not shown at all. I have not been able to use these records for this serial number exercise."
It would be very useful to know just how many of these there are.
For ones with complete serial numbers, you have observed just 0.4% of the total interpreted production of the -8000. This is a very low "survival" observation percentage - much lower than one might expect.
Clearly the 62MAS was used as intended back then, so perhaps we might not expect to see so many of them still surviving to this day, but by way of comparison, I have personally seen around 5% of the total production of the Seiko Astronomical Observatory Chronometer come to the market in the last 6 years or so.
The AOC is - I think - just about the only Seiko reference from that rough era for which exact production numbers are known, and hence provides a very useful marker against which to judge other populations based on what we see "out there" today. One other example is the first production run of the Astron - believed to be 100 units - from November 1969. In this instance, I know of four examples from this run.
Obviously the AOC and Astron are in a different league price-wise, which might lead one to expect they have been better looked after over the years, despite there being known examples of the references having their cases melted down for the scrap metal value.
However, an order of magnitude difference between what you have observed of interpreted -8000 production and what has been observed of AOC and Astron production feels very high.
Looking forward to the further posts - once again, great work!
Kind regards,
Gerald.
|
|
|
Post by zetamajor on Jun 27, 2021 13:49:15 GMT -8
Hi Ross, I have a 6217-8000 with an early serial number for your records 5400606. Thanks for your work on this topic, really very interesting.
|
|
rossr
WS Benefactor
Posts: 1,838
|
Post by rossr on Jun 27, 2021 17:22:16 GMT -8
Thanks for your comments gerald . As mentioned, there have been a number of 62MAS pieces for which I have not been able to extract a serial. However, most of these have been big crown pieces, and most these have been seen on the auction sites or on Instagram. I haven't made a huge effort to record these pretty anonymous pieces. In terms of 6217-8000 pieces, there are 8 likely unique small crowns with which I had some trouble. There might have been other for which I totally ignored if I could not see a serial number. For three of these, only one or two of the last two digits were missing, and so I set them to 0 or 00, so I could still make use some of the readable digits. There was another, sold out of NZ three years ago that had the last two digits blanked out. As the rest of the digits matched a record I already had, there was risk of duplication, and I didn't use this piece. In terms of the % produced that have come into "view", I'd have to say that this is definitely a function of my ability and motivation (moderate at times) to capture pieces as they come past. There have been periods in which I have not been monitoring the market, and so some have sailed right on by. I've been monitoring (non-continuously) the YJ market for 5 years, and Instagram for 3 years. There are quite a few small crown pieces on Instagram, but the serials are generally not shown, and I have not pursued the owners. As will be seen with the next update, the visibility of the later pieces, starting with the 1966 series, is marginally higher than that of the earlier pieces. Better survivability? Perhaps a greater bias toward showing off a clearer Dolphin caseback? I'm not sure. However, survivability of pieces within the record was not an aim of the exercise. It started as a price trend analysis, which became too awkward, and then morphed into the serial number record. Hi zetamajor , (donning my trainspotters hat here) that is a pretty remarkable piece that you have. It is the earliest serial number that I have seen reported, winds the start back another 129 places, and closes up that problematic gap at the start of the record. If it would be possible to post a picture both of the face and caseback - you should at least be able to post a link to a sharing site - that would be fantastic. BTW it was less than a week ago that I came across the highest serial number from the series - the June 62MAS - in a historically important publication on Seiko Divers by MrSeikosha. Best regards, Ross
|
|