If you want to do it moving you need 2 sets of 3 bearings as you need to make a significant course and/or speed change in between to alter the relative motion. And both 'target courses' need to be translated to the head of your own speedvectors. Where they cross is the endpoint of the actual target course and speed vector.
Read here:
http://www.filefront.com/13598315/bearingsonly_TMA.pdf
Also, for better accuracy in course, just take longer periods between the bearings. Waiting for 4 degrees bearing change is really too short. I guess you could average it out like that (first three bearings and last three bearings out of a set of 4), but it still has a large margin of error which doesn't go away doing that. You think the bearing is exact when it is reported, but it could be a full degree of. So the difference between 2 bearings has a total margin of error of 2 degrees.But since this method depends on 2 bearing differences the margin of error doubles again. Compare the following bearing sets:
B1=11, B2=15, B3=22, which makes target course 177
(from the original post example)
B1=11.00, B2=15.99, B3=22.00, which makes target course 150
B1=11.99, B2=15.00, B3=22.99, which makes target course 186
186-150=36 degrees margin of error.
I'm not too happy about that!
Yes, that can utlimately happen if you rely on crew reports that are not taken when the bearing crosses the exact degree.
Now lets say you wait longer, like until bearing has moved 7 degrees:
B1=11, B2=18, B3=37, which makes target course 177.1 (our reference)
Then taking worst case margins of error:
B1=11, B2=18.99, B3=37.00, which makes target course 173
B1=11.99, B2=18, B3=37.99, which makes target course 182
182-173=9
Much better, and if you do the math (which I'll spare you) those 7 degrees only took 50% longer, so 22 an a halve minutes per interval.