We have implemented two method:
Method 1: Base on the distance of beginning 3min and do calculation of Pace base on 3 sec shift.
Method2: To get pace base on first 1km and then to get next pace by 100m shift. (last 1km/time)
So far, it is not bad! But the two value seems is not equal.
You can´t expect the two values to be equal. That are two different methods.
I think method 2 is more reliable because the relative error is smaller than with method 1.
But why don´t you verify the methods with the real pace?
The way I would do is the following:
Go to a 400 m running-track and mark every 10 m.
Make a beep-pace-file in mp3-format and use a mobilephone or mp3-player for sound-paceing. Every 3,6 sec. one beep is a pace of 6:00 min/km. When you run, you have to be at a mark every time when it beeps. If you are a little to late, you have to run a little faster, ifyou are behind the mark run a little slower. You can control your speed very comfortable this way. Than have a look at the watch if the pace there is 6:00 min/km or how much it is different or varies. This is the first step to test if your method 1 is more reliable or method 2. You can do more tests with other speeds. One beep every 3,3 sec. is a pace of 5:30 min/km for example. As told, this could be the first step. The next is to change the speed after a while in order to watch the reaction of the watch.
If I should help you making the mp3-pace-files let me know.