WatchUSeek Watch Forums banner

Accuracy Tracking

50K views 259 replies 31 participants last post by  Trias 
#1 · (Edited)
As a little exercise I have been tracking the accuracy of several of my watches, 4 TC, a precisionist and a standard quartz that looked to be keeping fairly good time. I have not been concerned about temperature as a variable and realize that this could impact my results significantly. What I was more looking for was to see how well the various watches were keeping accuracy over time, as they are part of a larger collection that gets worn on the basis of a watch change every 1 - 3 days depending on mood and needs. As I also have several AT watches that are part of the wrist rotation it means that any given watch could conceivably not see wrist time for a couple of weeks (or longer for lesser favourites).

The watches in the test were:
- Breitling Airwolf Raven
- Bulova Precisionist Claremont
- Maurice Lacroix Miros Diver
- Seiko Worldtime Flightmaster SNJ017
- Sinn UX

I also am testing the Christopher Ward C70 Brooklands COSC, but currently have too few data points to use for the chart.

My methodology was to check the watches by the stopwatch method using my Citizen Skyhawk AT (after making sure that it had made a successful sync within the last day). At the start of the test I set all the watches and then took an initial reading to account for the small setting inaccuracy (the worst was off by ⅓ of a second based on my pathetic reaction times). At each test point I took 5 readings (sometimes more if one reading was significantly different from the others, replacing that reading). The readings were based on a 15 second comparison period, starting the reading as the second hand passed a certain marker on the "checked" watch, and taking the stopwatch reading on the AT watch as it's second hand hit the same marker plus 15 seconds. I took the average and then computed an expected SPY value based on the following formula:

SPY (est) = (Reading - 15 - original offset ) * 365 / (test date - start date)

I tabulated the values for each of the 5 watches, and while seeing relatively as expected performance from most of the watches I was noticing that for some there seemed to be an increase in the projected SPY as time progressed. It was at this point that I decided to create a plot of the SPY versus days (test date minus start date), and then use Excel's trend line tool to project out to 365 days. This captures an "acceleration" effect. The chart below shows the results:

Text Line Plot Parallel Slope


The Breitling and Sinn show relatively flat curves (lines) projecting to -2.76 spy (versus an average estimate right now of -7 spy) for the Breitling and 16.76 spy (versus an average estimate right now of 12.5 spy) for the Sinn. Both fall within COSC (at 25.5 spy), and the current average values are quite acceptable to me. The Sinn will get corrected at the fall and spring time changes, but the Airwolf will go the full 365 days on the test, as time change can be achieved without touching the minutes/seconds.

The Seiko is not providing any real surprise, and even projected out to 365 days is showing about 82 spy, which is better than the monthly range of +/- 15 spm would project to at an annual rate (180 spy).

The Precisionist is not looking anywhere near as good as the advertised claims, but then too is getting very little wrist time. I am sure that its numbers would be better if worn daily, but based on comments I have heard from a number of sources the 10 spy number is suspect.

The real shocker is the Maurice Lacroix Miros Diver (ETA 251.232). While early numbers were projecting a quite low spy, the last three checks were giving 13, 15 and 18 spy. This projects out to a very un COSC, un TC, 64 spy at 365 days! I will continue to monitor this situation, and will contact ML.

I am 121 days into my test and will continue monitoring until 365 days have passed (the Sinn and Precisionist will end the test at the fall time change as their seconds/minutes hands are affected when changing the time). The other 3 can adjust the hour hand independently.
 

Attachments

See less See more
2
#2 ·
Great work!
Thanks for doing the donkey work. I hope you can finish an entire year.

I was at first puzzled by what you call the "acceleration" effect. The rate increasing over time.
Even with small periods to measure over and extrapolating to a year, the rate should remain constant?

My guess it is that summer is having an effect here.
Your measurements will tell.

If you finish a year we can nicely see how each of the watches are affected by the temperature changes over the year.
 
#3 ·
I was at first puzzled by what you call the "acceleration" effect. The rate increasing over time.

Even with small periods to measure over and extrapolating to a year, the rate should remain constant?
I would have thought so too. The early data (under 35 days) was a bit erratic, but then a small error in reading say 0.1 sec after say the first 3 days would translate to 10 spy. The same error at 121 days is more like 0.3 spy. The chart and trendline exclude the early data.

There seems to be a definite growth trend in some of the watch data (the ML for example).
 
#5 · (Edited)
If I'm understanding this correctly, you have run the test for 121 days, but have drawn graphs out to 365 days. If we look between 0 and 121 on the graph, we don't see much "acceleration" except for the Miros and Seiko. The Breitling, Sinn, and Precisionist are fairly flat over the time period that has actually taken place. There is absolutely no reason to expect a rising curve when the data points are spy values. Since the watches are in storage for most of the time, unless you are experiencing substantial inside temperature fluctuations (which seems unlikely), I can see no reason whatsoever for the curves to be anything but flat. You are, after all, measuring the same phenomenon simply at different times. This is quite different from measuring rate over time (without the adjustment to spy), and it is quite different from graphing spy with a changing temperature abscissa, running, from, for example, 50 F to 90F. This all leads me to think that the construction of a trend line beyond the current state of timing is portraying something that will not actually occur. By far the most revealing part of your graph is at the point 121 on the abscissa. These estimates are quite solid being based on about 4 months of elapsed time.

One or two suggestions for the continuation of this project: (a) use 10 timings (at least) for your average offsets and (b) use a more precise clock as your reference--the NIST clock being one possibility. I've noticed that the signal lag varies with my RC watches.

By the way, none of the above is meant as criticism. Your results are very interesting and valuable and will be even more so after a year's timing.
 
#6 ·
I realize that there is significant danger in extrapolating future data based on current data, especially when the projection period is longer than the historical data period. At the same time while the Breitling, Sinn and Precisionist lines are flat, the other two do show a pronounced growth over time. The last three readings for the ML were at an annual rate of 13, 15 and 18 spy respectively.

I will continue to monitor these watches, and as time goes on my measured data will increase and my trend line will be shorter, and based on a larger set of data.

As to the accuracy of the Skyhawk AT, I check it against a NIST source on a few occasions and usually find that it is within 0.25 sec, so I doubt that I am picking up much error in the more recent readings. At 121 days the effect at 365 days would be no more than 0.75 spy. Also if the Skyhawk was causing so much distortion to the results, the Sinn, Breitling and Precisionist data would also be showing disturbing trends.

I'm not drawing any conclusions yet, but I do find the rapid rise in the ML readings a little disturbing. I doubt that reading accuracy is an issue because I did two separate sets of readings the day that I got the 15 spy value and basically got the same result.
 
#8 ·
OK, I decided to take another look at my data. Instead of plotting the SPY estimates for each reading I instead plotted the actual seconds off atomic time at each period, and then fitted a trendline out to 365 days.

Line Text Parallel Plot Slope


The trendline was computed from the last 4 readings. As can be seen, the ML Miros line definitely takes a bend upwards about 4 readings back. The dashed line marked "ML" is where I would have expected the Miros' trendline to actually go.

The table below shows the projected SPY values based on my original method (upper line) and my revised method. The Seiko is looking better at under a minute a year, the Breitling and Sinn are showing 365 day values that are consistent with my average weekly reading based estimates of SPY (7 and 12.5 respectively). The Precisionist is still showing worse than spec at about 29 SPY. And the ML Miros is definitely showing off spec for a TC caliber at around 35 SPY, whereas had it maintained its original trend it would come in at about 11.2 SPY.

SeikoBreitlingSinnMirosPrecisionist
81.79-2.7616.7663.9338.38
47.77-6.4913.6935.1729.16

I'll follow both methods and see where it takes me. I'd expect both sets of curves to merge (or come close) at the end of the test.
 
#9 ·
OK, I decided to take another look at my data. Instead of plotting the SPY estimates for each reading I instead plotted the actual seconds off atomic time at each period, and then fitted a trendline out to 365 days.

...

The trendline was computed from the last 4 readings. As can be seen, the ML Miros line definitely takes a bend upwards about 4 readings back. The dashed line marked "ML" is where I would have expected the Miros' trendline to actually go.

The table below shows the projected SPY values based on my original method (upper line) and my revised method. The Seiko is looking better at under a minute a year, the Breitling and Sinn are showing 365 day values that are consistent with my average weekly reading based estimates of SPY (7 and 12.5 respectively). The Precisionist is still showing worse than spec at about 29 SPY. And the ML Miros is definitely showing off spec for a TC caliber at around 35 SPY, whereas had it maintained its original trend it would come in at about 11.2 SPY.

Seiko
BreitlingSinnMirosPrecisionist
81.79-2.7616.7663.9338.38
47.77-6.4913.6935.1729.16

I'll follow both methods and see where it takes me. I'd expect both sets of curves to merge (or come close) at the end of the test.
Very good work - some of the 'strangeness' in the result might come from not having a very clear measurement on how big the thermal effects are - but in the long-term the average should still be very interesting!
 
#10 ·
Thanks for doing all the work. After tracking my SBCM023, I started on all of my current quartz watches, but gave up as too much work.

I also used my Skyhawk AT as the time standard and the offset (average -.15) was constant with NIST. I found it was easier and probably more accurate when comparing indexing second hands
side by side while using the stop watch method.

Walter
 
#11 ·
Increasing deviation over time from zero offset may be a sign of factors that we haven't yet fully understood. With my two tweaked 9F commemorative GS quartz models, I've seen a little of this and have been unable to explain it. The deviations have grown about 2 spy over period of 7-8 months (e.g., 2 spy in the first 3 months of testing, as opposed to 4 spy in the last 3 months of testing which have begun about 6 months after the end of the first set of readings). I've written it off to random error at this point. Crystal aging seems like another reasonable explanation, and Catalin can comment further on this. If our theories about the latter are correct, we might see a plateauing after a number of months in the graph of spy × elapsed time.
 
#12 ·
Doesn't a jump from 2 spy to 4 spy in only a few months seem awfully fast for crystal aging? I thought Seiko and Citizen both used only pre-aged crystals in their high end movements?
Anyway, I hope your two commemorative 9Fs learn to behave and level off in their behavior. Neither of my GS are tweaked (one is a 9F and the other 8J) and I haven't had either for long enough to know whether it is accelerating...... time will show.
 
#14 ·
Ten days later:

Line Text Parallel Colorfulness Slope


I have added the latest readings (today) so we are now at 131 days into the test. I have also added the COSC +/- 25.55 SPY boundaries (based on 0.07 seconds per day).

The Sinn, Breitling, and newly added CW C70 (only 40 days of data) are projecting well within COSC. The Precisionist is slightly outside COSC, not bad, but definitely not the +/- 10 SPY claimed.

The standard quartz Seiko isn't doing too poorly (well within the +/- 120 to 180 SPY of standard quartz). Although I haven't been accurately monitoring my older (non-AT) Citizen Skyhawk, I'd estimate that it is projecting to approximately 23.62 SPY (an earlier test had it at around 33 SPY, neither result too shabby for a non-TC movement).

The Maurice Lacroix Miros Diver is certainly not behaving like a COSC chronometer. I have contacted ML and will have to send the watch back to Switzerland. I'll be curious to see if the CW develops a similar behaviour as it uses the same ETA caliber.

The table below shows the SPY projections for the last two sets of readings:

Text Font


Maybe when the fall time change comes which will take the Sinn and Precisionist out of the current test I'll start a new test with the Sinn, Precisionist and non-AT Skyhawk.

Time will tell................
 
#15 ·
You could adjust the Miros yourself. It does have that feature.
 
#18 ·
Very interesting. Good work!

But I guess there's a factor overlooked. The voltage level. Most clocks and watches speeds up with decreasing voltage level unless they contain compensation for that too.
While I do not have logged it seems quite consistent with my normal clocks and watches I set regularly. Whether it's the dollar store clock in my bathroom (don't want to ruin an expensive one here) that is normally slow but in the few last months of a battery it is eerily dead on. Or my Seiko bedside alarm clock with is normally a fast but the last few months it is very fast.

Although silver cells have a flatter discharge curve than alkaline they're still not as water level flat as the old mercury cells. So to tell anything about aging you either need a constant voltage source or compare several battery discharge cycles with each other. And then you even can't be sure all the batteries are behaving exactly the same - achieving same voltage level at same discharge state etc. even though they may be same designation from same brand - a lot of factors have influence of a battery's performance.
 
#20 ·
But I guess there's a factor overlooked. The voltage level.
This is not a factor that, in my recollection, has been discussed much in connection with TC watch accuracy. My only recollections about it have had to do with whether or not light-replenished rechargeable batteries might produce less-than-optimal performance in TC modules (as in the "eco-drive" TC Citizen movements). In such cases, so went the theory, fluctuating levels of battery charge might cause decrements in performance. The consensus was, again if I'm remembering correctly, that this shouldn't be a factor affecting accuracy, and I don't believe any empirical evidence was presented suggesting such degrading of accuracy.

I'd really like to see some well-developed theory or, far better, solid, replicated empirical evidence of varying-voltage effects before considering this factor worthy of consideration.
 
#19 ·
Battery discharge level could be a factor, but all the watches are relatively new, although how long the watch sat "in-stock" before I bought it is anybody's guess. As it is adjustable per an earlier response I am going to continue to monitor it for a while longer, but that low battery theory may just be the answer, and if not, then a little adjustment may just be in order here.
 
#28 ·
Todays updated graph and numbers (we're now at 139 days, 48 for the C70):

Line Text Parallel Slope Colorfulness


Text Font Line


The drop in projected SPY may be influenced by the generally cooler temperatures that we are experiencing of late. It's getting cooler outside and we don't have our furnace turned on right now, so over night the bedroom does get fairly cool, and the daytime highs are lower (except next week when we are supposed to get some summer like daytime temperatures).

Assuming that I adjust the watches after 1 year (Seiko, Breitling, Miros & C70) or at DST/non-DST time change (Precisionist & Sinn) the worst that each watch would likely be out is:

White Black Line


That means that along with my radio controlled watches I currently have at least 6 watches that will be within +/- 10 seconds accuracy at any given point in time. And one more, the precisionist that will be within +/- 15 seconds.
 
#29 ·
Nice and interesting work Sabresoft! I follow your thread with great pleasure while I'm in a similar procedure myself. I've had 6-7 watches on the test-bench for just over 6 months now, three of them identical to yours:
Sinn UX
Breitling Airwolf
Chris Ward C70
In due time results will be presented, and I will try to adopt your methodology. For now just a comment on this statement:

I am 121 days into my test and will continue monitoring until 365 days have passed (the Sinn and Precisionist will end the test at the fall time change as their seconds/minutes hands are affected when changing the time). The other 3 can adjust the hour hand independently.
I ran in to the same problem last spring on DST-day, and decided to use this work-around:
Immediately before setting for DST I did an accuracy-test and then another just after the watch was set for DST.
I then correct all test-results after the DST-setting for the difference between the two readings. In this way I get continuous results also for watches without an independent hour-hand.
 
#30 ·
Day 146 (55 for the CW C70). The chart doesn't look too different from last time, so I'll just post the projected SPY values:

View attachment 511629

As most claimed SPY performances from manufacturers require a certain amount of wrist time I'm sure that the watches would perform better if worn daily. I'd try wearing only one watch for an extended period, but just can't bring myself to ignore all the other watches for so long. And no I won't wear two watches just for the sake of "science".

That said this test was intended as a "practical" test of a collection of watches to see which will keep the best time while not in use, so that when I wear them they are reasonably accurate.

At this point I have six watches that I can wear on a moment's notice and be confident that they are within +/- 10 seconds or better at any given point in time (assuming a once-a-year reset at minimum):

- Citizen Skyhawk AT (radio controlled)
- Citizen Attesa ATD53-3081 (radio controlled)
- Casio G-Shock GW3500B-1A (radio controlled)
- Breitling Airwolf Raven
- Sinn UX (necessary resets at DST changes will keep it under 10 seconds)
- Christopher Ward C70 Brooklands

The Precisionist is close (as it would have to be reset at DST changes).

I'll need to look into getting the Miros adjusted.

I am looking to eliminate some watches from my current total of 15. The Seiko and Precisionist keep coming under consideration, but I just keep saying no to letting them go. Although the Seiko is non-HEQ it keeps a place in my heart because it has a perpetual calendar and is ana-digi, my favourite format (I have 5 ana-digi watches). The Precisionist might go if Bulova comes out with a model with more complications or even just a perpetual calendar and adjustable hour hand.

While I am not a fan of pure digital watches, the Seiko Brightz SDGA003 does have some attraction for me (solar, RC and EPD display) picture courtesy of Seiko, and would add that "fourth" RC watch to my collection.

View attachment 511674

Then I'd have 4 TC, 4 RC, 1 Precisionist (the accuracy brigade) and 4 standard quartz and 3 mechanical. Of the standard quartz, one, my Independent is up for sale, the Seiko is on the fence right now and the Campanola and non-AT Citizen Skyhawk are keepers. The non-AT Skyhawk was my only watch for 10 years and it was when it went in for maintenance (the pushers needed de gunking) that this horology virus hit me. Of the mechanical watches one, a du Bois, was a gift from my grandfather and stays in the collection, another is a cheap Chinese no name which I kind of like despite the lousy accuracy, and the last is a Rotary with annual date (it doesn't understand leap years), but has too small a dial so it too will be up for sale soon. I'd like to add a Breitling Navitimer or Sinn 903 (similar look to the Navitimer at ½ the price), and am always on the lookout for TC watches.
 

Attachments

#34 ·
Data update from this weekend (day 153).

I have changed the way that the trendline is calculated. Previously I used the last 8 or so readings. I have now decided to use all but the first few readings, but including zero and this has changed the lines slightly. I decided to eliminate the first three or so readings because small errors on early readings were having a magnified distorting effect on the trend.

The effect of this change was to bring the trendlines more in alignment with the projected SPY values computed by taking a reading and projecting the SPY based on the 365/days@reading ratio. The upshot of this is that the Precisionist and Miros are now projecting close and within COSC limits respectively.

Line Text Parallel Colorfulness Plot


While most of the trendlines are straight, the Miros line has a bend downwards and then goes straight (which is the projected portion) so I am certain that there is something strange going on with that watch. But I am less inclined to believe that it is due to any fundamental problem with the watch and am tending toward the failing battery theory. While I acquired the watch in April of this year, the COSC certificate was dated July 24, 2008. Although I believe that only the mechanism is tested, I imagine that assembly and mating with the current battery would have happened some time in 2008, which puts the battery at around 3 years old (probably close to it's limit). The ML manual is notably unhelpful in this area, but the CW 70 which also uses the ETA 251.232 caliber has an expected battery life of 2 to 3 years, and the Breitling Avenger Seawolf Chrono which I believe is also based on the same 251.232 caliber also indicates a 2-3 year battery life. I will let this play out, but I wouldn't be surprised to see the battery die shortly. The certificate showed 0.03 SPD at 23C, which would support a 10-11 SPY value, and the Miros was projecting less than that in the first 106 days. It was after that this acceleration effect showed up.

The table below shows the projected SPY from the trendlines, including the old and new methods for this period.

Text Font Line Number


A little further reading and I am now totally confused. The ML uses not the 251.232, but a 251.262, which the ETA website lists as a Flatline not a Thermoline. Yet I have a COSC certificate and the watch has COSC on the dial.
 
#35 ·
This thread has evolved to be exiting! b-)

Please lets us know what happens to the accelerating Miros! It wouldn't surprise me a bit if the battery is almost done and it jumps back to normal with a new one.

Or you could of course also just have it changed now and see what happens.
If it doesn't get better instantly you can always send it back for repairing afterwards. ;-)
 
#36 · (Edited)
Day 167

Latest Chart

Line Text Parallel Plot Slope


Latest SPY Summary

Text Font Line Number Design


Since I suspect that the Miros misbehaviour may be due to near end-of-life battery behaviour I have been running the chronograph continuously 24-7 to expedite final battery failure. We'll see what happens.

As my chart/table doesn't fully identify the models, here is a refresh on the models under test:

- Seiko Worldtime Flightmaster SNJ017 (Standard Quartz)
- Breitling Airwolf Raven (TC)
- Sinn UX (TC)
- Maurice Lacroix Miros Diver Chronograph (TC)
- Bulova Precisionist Claremont (Precisionist High Frequency Standard Quartz)
- Christopher Ward C70 Brooklands (TC)
 
#37 ·
Day 174, getting pretty close to ½ year.

Just the summary table today.

Text Font Line Design Number


As mentioned earlier the old method involved using only the later data, the new method included zero and used all the data except for the first few readings (too erratic due to the high error potential in the small readings).

Clearly the Breitling Airwolf Raven and Christopher Ward C70 are showing exceptional performance, the Sinn UX a little less so, but still quite acceptable (Sinn makes no other claims other than COSC Chronometer which this unit easily meets).

The Bulova Precisionist is at around 26 second SPY, just outside COSC, higher than their claimed 10 SPY, but I believe that that requires wearing daily, and actually this watch gets limited wrist time, so the performance is probably acceptable.

The Seiko Flightmaster is a non-TC, non-HEQ watch, and so the 40 SPY performance is quite exceptional considering that standard quartz can be up to +/- 180 SPY.

The Maurice Lacroix Miros continues to "accelerate". I have been running the chronograph 24-7 lately in the hopes of running the battery down quicker, as it is very likely 3 + years old which should be its limit.

Just out of curiosity I looked at the COSC certificates for the TC watches:

The Breitling certificate has no data.

The CW certificate (2010-07-12) has 0.01 spd @ 23°C, 0.05 spd @ 8°C and 0.06 spd @ 38°C - My current measurements indicate a range of 0.0095 to 0.0170 spd (in a temperature range of probably 16°C - 28°C).

The Sinn certificate (2008-02-20) has 0.01 spd @ 23°C, 0.01 spd @ 8°C and 0.02 spd @ 38°C - My current measurements indicate a range of 0.0360 to 0.0365 spd (in a temperature range of probably 16°C - 28°C).

The ML certificate (2008-07-24) has 0.03 spd @ 23°C, 0.02 spd @ 8°C and 0.07 spd @ 38°C - My current measurements indicate a range of 0.0669 to 0.0772 spd (in a temperature range of probably 16°C - 28°C).

Curious that the Sinn and ML are showing current performance at 2.4 and 3.6 times the original tested values @ 23°C. Could this be due to crystal aging?
 
#38 ·
Nice data, Sabresoft. I don't know whether you have the C70 and Precisionist data in your table because, on my screen, it seems to be truncated after the Miros results. Just to refresh my memory, let me ask you this: In your latest spy estimates have you simply taken the actual drift over the 174 days and multiplied this by 365/174 to get the spy values? To me, those would be the most accurate spy estimates with your data.

Your question about the possibility of crystal aging is an interesting one. I have noticed the same phenomenon with my three Grand Seiko 9F models, with later timings indicating larger estimated spy values than earlier timings did (and, I believe, independently of weather-caused temperature changes), although still within specs. It would be good if we could do some accurate modeling of crystal aging. With two of my GS 9Fs, Seiko offer to re-tune them in the third year of use. This might be tacit acknowledgement of crystal aging. I guess our hope would be that if it has to occur, its effects are fairly rapid with a steadily decelerating function, so that by Year 2 or 3, it is in full plateau mode. This all might suggest that we not worry too much at first about accuracy, getting serious by, say, Year 2 and making sure that any re-calibration takes place no sooner than that.

Just out of curiosity, how many trial replications are you averaging to get your drift values?
 
#40 ·
I guess our hope would be that if it has to occur, its effects are fairly rapid with a steadily decelerating function, so that by Year 2 or 3, it is in full plateau mode. This all might suggest that we not worry too much at first about accuracy, getting serious by, say, Year 2 and making sure that any re-calibration takes place no sooner than that.
Well the ML & Sinn are 3 years old based on the COSC certificates. I can live with the Sinn (I'd prefer under 10 SPY, but then I have to reset it twice a year for DST/non-DST due to there not being a quick hour feature, so the worst it will be out is about 8.5 seconds).

I am not sure that battery is the problem on the ML, but it seems to be a feasible explanation considering its probable age (I doubt that the vendor would have replaced it, so I am assuming that it dates from when the unit was first assembled). If so then the true normal multiple of original drift may be much lower. We'll see.

Just out of curiosity, how many trial replications are you averaging to get your drift values?
I take six readings, and then if one or two are extreme (say one high, one low) I'll throw those out and take two more readings.

For example the last readings for the Breitling were:

1.53
1.51
1.51
1.5
1.54
1.5

for an average of 1.515 (I used 1.52 in my SPY calc). These values were the difference between the time that I started timing on the tested watch and the point where the reference source passes 5 seconds later).

I then compute the actual difference as:

measured average reading minus 5 seconds minus initial offset at time of starting the test

so in the case of the Breitling the actual difference is: 1.52 - 5 - (-0.35) = -3.13 seconds

and in the case of the ML it is: 17.44 - 5 - (-0.55) = 12.99 seconds

Even if I used 1.5 or 1.54 this time for the Breitling the projected SPY would fall in a range of 6.39 to 6.40 per the trendline.

What I did not do was take multiple readings when capturing the initial offset. The next time I launch a test I will take more care in establishing the initial offset (which ranged between -0.55 and +0.39 seconds for the various watches in this test).

The general fluctuations in my readings (including the initial offsets) is a testament to my poor hand-eye coordination, and my slow reaction time. Still for the most part there is a consistency in the periodic readings that shows an SPY projection that is fairly consistent from week-to-week. With the exception of the ML, which shows a continual growth.

I don't think that procedural errors in my method could explain the ML situation, because if my procedure was flawed other watches would also be showing strange behaviour, and they aren't. Doing the readings for the ML is more challenging than for most of the others because of the short, but fat seconds hand (the same caliber CW has a much thinner seconds hand and a tick for every second, not 5 second increments, so it is much easier to do the readings). Also the alignment of the ML seconds hand is poorer, and so when I do readings I tend to get ⅓ to ½ with a full second difference from the others. As a result I may do 10 or more to get a consistent sample of 6 actual readings that I use in my average.
 
#42 ·
Well we're almost at 6 months (181 days, 90 for the C70):

Text Font Line Design Number


The Sinn, Airwolf and Precisionist continue with very similar projected SPY values. The Seiko seems to be improving a little. I wonder if this is due to the generally cooler temperatures of late.

The Miros continues to show an increased SPY projection. Still waiting for the battery to die. The C70 is also showing a slight increase in SPY projection from week-to-week, but the line still looks to be relatively straight unlike the Miros which is showing a decided upwards curvature.

Line Text Parallel Plot Slope


The rather strange shape of the Seiko and to some extent the Miros curves is partially due to limitations of the Excel trendline function. I am not too worried because as we get closer to 365 days these lines will make more sense.

Considering that I do not wear the Precisionist very much I think that the 26 SPY is acceptable. I'm sure that if I could wear it daily I'd see closer to the claimed 10 SPY, although I'm not sure it would be 10.
 
#43 ·
It will be interesting to see whether there is any change to your trend lines as the weather gets distinctly colder. This will only have an effect, of course, on those you wear outdoors. All things considered (cost, etc.), that seems like pretty decent performance for the Precisionist.
 
#44 ·
I don't anticipate too much difference from the temperature outdoors because in the winter the watch is inside my winter jacket sleeve and with the possible exception of a long session shoveling snow I wouldn't expect my arm to get too cool. Most of the time it will be short durations outside.

The bigger impact though may come from inside temperatures because we allow the bedroom to cool significantly during the overnight and daytime hours, only running heat when we are getting up and getting ready for bed. It is not unusual to have the temp at 18C and during the really cold weather it can even drop to 16C.

The sudden change in the Seiko lately may be due to the already cooler daytime and overnight temperatures.
 
#45 ·
Day 188

The latest chart:

Line Text Parallel Slope Font


Noticeable is the truncated line for the Seiko. There is a reason for this. As of Yesterday I no longer own that watch. I had been on the fence about selling the watch over several months now.

Reasons to sell:
- Not HAQ
- No backlighting of the digital display
- Didn't get much wrist time because in many ways was too similar to the Breitling and Skyhawk in style

Reasons not to sell:
- Ana-digi style (which I like)
- "pilot" style watch with slide rule bezel (another of my favourites)
- Perpetual calendar
- Multiple complications (chronograph, world time, alarm)
- Pretty good accuracy for a non-HEQ

Reason it sold: was showing a picture of my collection to a co-worker on Wednesday and he pointed to it and said he quite liked it. Impulsively I asked if he wanted to buy it and he said yes. In some ways I miss it already, and I was curious to see where the curve was heading (I almost wanted to say wait till Monday so that I could do one more reading at least), but as pointed out in the reasons to sell above, it just wasn't getting much wrist time, and as far as I am concerned my collection has to be a "wearing" collection, not just something that gathers dust in a drawer.

The SPY data:

Text Font Line Number Design


The Breitling, Sinn, Precisionist and CW C70 are showing fairly consistent SPY projections, (the C70 with a little growth but seems to be stabilizing). The Miros continues its upwards projection well outside COSC.

The plan forward is to continue the test till next April for the Breitling, CW and Miros (assuming no battery failure), and to start a new test after the no-DST time change for the Sinn, Precisionist and add in my non-AT Skyhawk. Probably do an update on each test on alternating weeks.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top