I click the button (so I can find it on the graph), then go through the full range of eye movements. I've found that left/right gives a greater difference than up/down. Keeping my eyes closed the whole time, I center my eyes, looking straight ahead for 8 seconds; then I look left for 8 seconds; center for 8 seconds; right for 8 seconds; center for 8 seconds (sometimes I'll repeat the sequence). Why 8 seconds? I think I started using 8 seconds because that's how often the board dumps its memory to flash. Of course, it really doesn't matter (as long as I wait 8 seconds after I'm done before powering down the board).

Spoiler for graph:

My 3/24/14 data does in fact have me running through it both on my back facing up (my normal position for it) as well as laying on both sides. I know I started on my back, then did one side then the other side. I don't specifically remember which order I did it in; but from the data I'd have to say it was center, right, then left (since higher numbers mean closer to the sensor, and laying on a side would push that side's sensor closer to my face). It's interesting, since there seem to be times in which I get the highest reading by centering my eyes and equally low readings looking left or right (such as green in the middle section and blue in the right section). I guess normally the sensors are looking at the corners of my eyes but sometimes end up pointing more at the center.

Spoiler for mask:

The boards themselves don't touch the pillow, even when I smash my face into it. To get the dimensions for the nose board, I laid on my pillow, rolling my head into it as much as possible, then measured the distance from the middle of the bridge of my nose to the point on my eyebrow that touched the pillow, then sized the board to fit into the gap. I did something similar for the head board. I wear them in fabric pouches with elastic bands. Yes, that is rather a lot of elastic; but by having some band that go above my ears and others that go below, it holds it in place much better. It also lets me keep them much less tight than the single band designs I've tried in the past.

As I roll around, I'm sure the pouches pull on the boards some. If you look at some of the overall graphs, you can see lines which are straight but have a gradual slope to them; I suspect that when I roll, the back of my head slips against the elastic making one side slighty tighter and the other slightly looser, and that the mask then slowly slips to the side.