I was bored a couple of years ago and worked out a reliable method of navigating by the sun. It works on the principal that you can equate the elevation of the sun from horizon, to its horizontal distance from either east or west where the sun nominally rises. But of course this is subject to two errors:
1. The sun does not rise in the east or set in the west other than at the equinoxes
2. The elevation of the sun does not exactly equate to the horizontal distance from where it rose or set. This error varies with time of day and time of year.
I wanted my method to account for these errors in a usable, practical way.
To do this I got from the internet the azimuth at sunrise and sunset and sun elevation throughout the day, on various days throughout the year, and put this into a spreadsheet. I could then work out the difference between elevation and horizontal distance traveled from sunrise (or yet to travel for the afternoon) throughout each day, and how this varies. E.g. a difference 2 means the sun has travelled horizontally twice as far as it has elevated.
I also entered into the spreadsheet the difference (in degrees) between sunrise/set azimuth, and actual east or west.
This involved a lot of numbers, so I then condensed the numbers into something usable and memorable and in which any remaining error is acceptable to me.
My process in the field (in the morning):
1. I measure the sun's elavation using my hand, and apply that same distance horizontally back towards east.
2. I then apply the first correction (in this example it would be 1.5 fold correction in morning between March and November i.e half the distance again correction in the morning is 2 fold in the winter months). Then I apply a further correction using my feet as a protractor to account for the difference between sunrise position and actual east (difference is 0 at equinox 21 march, 20 degrees at 21 Apr, 35 on 21 may, 40 degrees at summer solstice, then it reverses as the year continues.
With a bit of practice I can figure out east and west in under a minute using this method. Much quicker and more accurate than stick shadows and analogue watch faces. I'd be happy to type out my set of condensed numbers if anyone is interested.