How accurately should I try to get my Seiko automatic to run?

[Mabouya]

I bet he showers with it too...

[Mark Kelly]

No, but he carried three of them up Mt Rainier in order to demonstrate the General theory of Relativity.

I came across the guy when I was buidling a very high accuracy (< 10^-12) time source for lab use. He makes ordinary nerds look like we aren't even trying.

[Mabouya]

I've made an interesting and potentially useful discovery: I put the Seiko in question on my winder for a few days and it lost very little, if any time. (The winder's a Wolf Module 4.1, which holds the dial in the vertical plane and turns it about the watch's primary axis. I have it set for both CW and CCW turns, since the movement is designed to accept both.)

I want to run the experiment for a week or so on the winder, but assuming this hold true, that suggests to me that either the orientation of the watch or the state of wind of the mainspring is having an effect on accuracy. (I'm assuming that temperature effects can't be that big, although I do keep my bedroom pretty cold.)

Anyone care to guess which of the two effects is more likely to be true? From what I can tell, orientation shouldn't affect things by more than a few seconds a day.