Accuracy and Precision

[Nucejoe]

Picture below shows the difference between  precision and accuracy.

https://circuitglobe.com/wp-content/uploads/2016/09/accuracy-and-precision-compressor.jpg

 How does this relate to watchmaking will be a useful discussion. 

Rgds

[tomh207]

As a former competitive shooter and coach I can relate to the illustrations Joe. In watchmaking, specifically timekeeping would accuracy equate to s/d and precision the positional delta?

 

Tom

[Nucejoe]

Hi Tom .

Exactly as you said.

 Adjustments manufacturers traditionally make to achieve high precision timing including  tools/ approach ( proprietory and general tool)  be useful to discuss,  likewise for regulation and accuracy, furthure more general  tools avilable to repairmans for said tasks.

Rgds

 

 

[RichardHarris123]

If a watch loses 3 minutes a day consistently then its precise but not accurate? 

[Nucejoe]

Hi Richard, 

Interesting point,  when distribution of dots is inaccurate but avg out to accurate timing. the watch shows accurate time but the escapement is inaccurate. 

[tomh207]

56 minutes ago, RichardHarris123 said:

If a watch loses 3 minutes a day consistently then its precise but not accurate? 

I think it really brings in the context of testing in positions. If DU  only 3 minutes could be considered precise as it is consistent. However if it varies to say 1 minute per day DD then the Watch could not be considered precise or accurate. Precision is essentially consistency and accuracy hitting the intended target. Hope this makes sense, great topic for discussion @Nucejoe, I am very interested in this very subject as I am trying to put together a testing/ standard process to look at my service and regulation quality.

 

Tom

[Nucejoe]

 Going back to circles in the image,  the case of accurate  not precise, shows  the dots sorta symetrical, which in some respects can be misleading. A distribution of non-symetrical dots can also avg out as accurate. 

21 minutes ago, tomh207 said:

 Precision is essentially consistency and accuracy hitting the intended target. Hope this makes sense,

Perfect sense tom.

In time keeping , hitting the target can be  avg of the total sum, ie; meantime.

 

[Neverenoughwatches]

1 hour ago, RichardHarris123 said:

If a watch loses 3 minutes a day consistently then its precise but not accurate? 

just the ability to reproduce the same result.  Precisely,  inaccurate by the same quantity . If its accurate then it must also be precise as long as its consistently accurate.

Edited March 8, 2024 by Neverenoughwatches

[Nucejoe]

44 minutes ago, tomh207 said:

 I am very interested in this very subject as I am trying to put together a testing/ standard process to look at my service and regulation quality.

I hope Nickelsilver tells how manufacturers regulate, shows  tools/ equipments etc. 

I guess some manufactures disallow showing their proprietroy tool to the public.

 

[tomh207]

1 minute ago, Neverenoughwatches said:

just the ability to reproduce the same result.  Precisely,  inaccurate by the same quantity . If its accurate then it must also be precise .

Only insofar as we are understanding watches. I think Joe is wanting us to discuss towards defining what these terms mean in watches. It can be “acceptably” accurate but not necessarily precise. The context matters a lot, on the wrist a 1 second a day on your wrist is great, the same watch on my wrist for my daily wear maybe 20 seconds and unacceptable. If however chronometer standard watches you would expect to be very close whether it was you or me wearing it.

@Nucejoe please tell me if I am on the wrong track.

 

Tom

[Nucejoe]

14 minutes ago, Neverenoughwatches said:

 If its accurate then it must also be precise as long as its consistently accurate.

Check accurate but not precise, image. 

It can be consistantly imprecise but accurate.

[Neverenoughwatches]

3 minutes ago, Nucejoe said:

Check accurate but not precise, image. 

It can be consistantly imprecise but accurate.

I see that accurate as an average of being imprecise around a closeness of accurate but not reaching true accurate. 

[Nucejoe]

12 minutes ago, tomh207 said:

Only insofar as we are understanding watches. I think Joe is wanting us to discuss towards defining what these terms mean in watches. It can be “acceptably” accurate but not necessarily precise. The context matters a lot, on the wrist a 1 second a day on your wrist is great, the same watch on my wrist for my daily wear maybe 20 seconds and unacceptable. If however chronometer standard watches you would expect to be very close whether it was you or me wearing it.

@Nucejoe please tell me if I am on the wrong track.

 

On your wrist ,  point might  have certain distribution, and different distribution on my wrist, but both can avg out to accurate.

Precision is limitting distribution of points in a dense group.

In watches, we seek to avg to accurate , whilst torque supply deminishes, so in case of manual winds, better to wind every twelve hours or six for a relatively consistant torque supply. Selfwinders do consistant winding  for us.

[Neverenoughwatches]

15 minutes ago, Nucejoe said:

Check accurate but not precise, image. 

It can be consistantly imprecise but accurate.

Accurate but to a set period of time. Could a watch be specified as accurate for one day due to an average of closeness of accuracy finishing at true accurate at the end of one day. Or a week a month or a year. 

[Nucejoe]

13 minutes ago, Neverenoughwatches said:

I see that accurate as an average of being imprecise around a closeness of accurate but not reaching true accurate. 

Percision is only togetherness of tics in small groups. which can be way off target. its where we seek consistancy of periods of ticking. 

Accuracy is an average of all tics .

Regulation is getting this avg as close to the target as we can.

Rgds

 

[tomh207]

4 minutes ago, Nucejoe said:

Percision is only togetherness of tics in small groups. which can be way off target. its where we seek consistancy of periods of ticking. 

Accuracy is an average of all tics .

Regulation is getting this avg as close to the target as we can.

Rgds

 

So we are getting somewhere, we need to fettle the train and escapement to achieve precision then proceed to regulate to achieve accuracy. Summed up by many folks as “you can’t regulate a watch until it is working perfectly “

 

Tom

[eccentric59]

I think the variable that should be equally considered is frequency and the size of the data set. For example a watch regulated to a 0° BE and loses 30 seconds/day is not accurate but precise. A watch with a positional BE delta of 2° but keeps time +/- 2 seconds a day is accurate but not precise. But this is only at the time of measurement.

Now what if we had a watch that ran 20 seconds fast when fully wound, but 20 seconds slow after 24 hours and the owner wound it once per day? The watch would be accurate if the owner only looked at the time when he wound it.

 

[tomh207]

5 minutes ago, eccentric59 said:

I think the variable that should be equally considered is frequency and the size of the data set. For example a watch regulated to a 0° BE and loses 30 seconds/day is not accurate but precise. A watch with a positional BE delta of 2° but keeps time +/- 2 seconds a day is accurate but not precise. But this is only at the time of measurement.

Now what if we had a watch that ran 20 seconds fast when fully wound, but 20 seconds slow after 24 hours and the owner wound it once per day? The watch would be accurate if the owner only looked at the time when he wound it.

 

That is then bringing isochronism to the table

 

Tom

[Nucejoe]

If it works perfect it can get regulated  accurate and  adjusted precise.

Think high grade watch with split second precision. thats close to perfect.  its both accurate and precise, stable period of oscilation is perfect. 

3 minutes ago, tomh207 said:

That is then bringing isochronism to the table

 

Tom

[Neverenoughwatches]

57 minutes ago, Nucejoe said:

Percision is only togetherness of tics in small groups. which can be way off target. its where we seek consistancy of periods of ticking. 

Accuracy is an average of all tics .

Regulation is getting this avg as close to the target as we can.

Rgds

 

Yep I'm right on the page there with you joe.

51 minutes ago, tomh207 said:

So we are getting somewhere, we need to fettle the train and escapement to achieve precision then proceed to regulate to achieve accuracy. Summed up by many folks as “you can’t regulate a watch until it is working perfectly “

 

Tom

Fettle, that meant something quite different when we were young

[tomh207]

23 minutes ago, Neverenoughwatches said:

Yep I'm right on the page there with you joe.

Fettle, that meant something quite different when we were young

I’m older than you Rich but the meaning is the same, gently manipulating until the desired outcome is achieved

 

Tom

[mikepilk]

I have spent over a week tweaking my Longines 30L, to see how accurately I could get it  running. I had to change the balance staff, so the balance is very slightly out  crown left. I decided it was good enough. I may tweak it the next service.

Fully wound        
        Amp    +/-
DD    296    5
DU    290    3
CD     248    -7
CU    250    15
CL    246    -33

I wind it each morning and I've been wearing it through the days since Monday - when it was set exactly to the correct time GMT.
At 8am this morning (Friday) it was within 1 second.

Throughout the day it loses slightly, As now, at 5pm, it is -4s. But through trial and error, I have found that setting it to +4s/day DU at full wind (the trace is rock steady), sat on the desk overnight, by morning it is back to within a second of the correct time.

I guess this shows that it is accurate, but not precise !  (Though it has never been more than 5s out)

What it has taught me is that the setting on the timegrapher is only the starting point - you need at least a week of wearing and tweaking to get real accuracy, and it will be different for every person depending on their lifestyle and what they do.

Edited March 8, 2024 by mikepilk

[Neverenoughwatches]

1 hour ago, tomh207 said:

I’m older than you Rich but the meaning is the same, gently manipulating until the desired outcome is achieved

 

Tom

Nicely put sir

[RichardHarris123]

1 hour ago, mikepilk said:

Fully wound        
        Amp    +/-
DD    296    5
DU    290    3
CD     248    -7
CU    250    15
CL    246    -33

Definitely not precise but 1s/d is accurate.  The range is high,  50⁰ amplitude and 38s positional variation. Standard deviation on the amplitude is 24.7⁰, on the s/d is 18.3s/d. The relationship between amplitude and rate cannot be linear. 

[mikepilk]

34 minutes ago, RichardHarris123 said:

Definitely not precise but 1s/d is accurate.  The range is high,  50⁰ amplitude and 38s positional variation. Standard deviation on the amplitude is 24.7⁰, on the s/d is 18.3s/d. The relationship between amplitude and rate cannot be linear. 

I would say that the difference between the horizontal and vertical positions is normal. When I see 240-250° vertical, I'm happy. I serviced a friends Rolex and the numbers were almost the same horizontal to vertical.

Do you consider 50° high?  

I fitted a new balance staff (Longines original part) and new balance jewels, chaton+cap.

BTW it's better than 1s/day. It's better than 1s in 4 days 

Edited March 8, 2024 by mikepilk