My Seamaster AT has gained 3 seconds/day since I purchased it 4 years ago. Every two months I’d put it in hack until it matched the USNO atomic time. Three weeks ago I noticed it was running slow. I’ve checked it several times using the Watch Tracker app and it is now losing 5.1 seconds/day - an 8 second change. Can anyone explain such a large change? Btw, it’s a Master Coaxial spec’d out to withstand >15,000G magnetic fields so magnetism shouldn’t be an issue, right?
I'd be interested to hear what this issue turns out to be. From what I've been reading on here, a magnetized watch would run faster, as it shortens up the spring. Seem something else is happening with your watch.
A lot of things besides mechanical changes, but a change of eight seconds seems a little much for an 8900. Different wear patterns (time on and off wrist), activity levels, even changes in your behavior (digging ditches vs. typing) come to mind. That said, it's a bit much and maybe it's time for a service. As you guessed, magnetism wouldn't be a cause.
Changing how you leave a watch over night can affect the timing. If left upright with crown up it can lose as much as 15 seconds per day, depending on the movement. Flat with dial up it gains a few seconds. other than that the ambient temperature of the room can affect the lubrication, though modern lubes aren't that easily affected by temperature.
There's no straight and fast rule about positional variation and which positions run fast or slow. A watch runs at higher amplitude in dial up or down positions and typically this means slower but not always. I have an SMPO Chrono with a 9300 movement that runs +0.5 s/d with my normal wear pattern, day in and out. This pattern is wearing 12 hour/day and setting crown up for the remainder. When I take it on diving trips it stays on my arm 24/7 and loses about 3-4 s/d. Theoretically, I'm losing 2 s/d on the wrist and gaining back plus a little more while off the wrist. But some of it may be attributed to the fact that during the day I spend hours on the keyboard in the dial up position, and not so much on a dive trip. I doubt your 8 s/d difference is caused by positional variation, not with 8900 movement that's only four years old. If I had it, the first thing I'd do it put it on the timegrapher and see if there are dramatic differences in positional rates. But in your shoes, I'd: 1) make sure it's fully wound before starting (winding by hand). 2) set it against a good standard (radio watch, smart phone, WWV, GPS, not the internet). 3) let it run in a position overnight and check it next day at the same time. 4) record your results. 5) repeat the above (including winding) several times before going to the next position. This will give you a good shot at understanding positional variation on your watch, the old school way. BTW, I've found the clock icon on the home screen of my iPhone is very close to GPS, close enough that I can't tell.