I posted a note recently (http://publiclab.org/notes/Ecta64/05-16-2017/video-tracking-software-for-monitoring-kites) regarding video tracking of kites for various applications. During the experiment I flew a TALA anemometer kite (http://publiclab.org/wiki/kite-anemometers) with a approximately 15 foot long tail that terminated in a 2.83 gram weight. After getting the video that formed the basis of the screen capture below, I reeled in the kite to find that the weight was missing. The video track rendered in Kinovea was grouped tightly at first then moved all over the sky rapidly. I suspect now that the weight fell off around 3-5 seconds into the track and the kite was being stabilized by drag alone versus the "gravity-sensing" provided by the TALA weighted tail system (for more see this note: http://publiclab.org/notes/Ecta64/07-05-2016/weighted-tail-for-high-wind-range-kite-stability and here for some interesting anecdotes with other kite designs: http://www.arch.ced.berkeley.edu/kap/discuss/index.php?p=/discussion/4187/kite-tail-stability-enhancement).
Flights tracked via Kinovea (such as in the main note picture) in which the weight stayed on show a much more tightly grouped track (even when altitude differences are compensated). While a controlled study is needed, the kite tail of the TALA system (whether 36 feet of 15 feet) performs remarkably better with a sub-3 gram weight on it.
The main thing the weighted tail seems to do is correct a kite from attitude changes that because of the kite's AOA cause the kite to begin moving (or looping and diving into the ground) in a new direction. The weight "senses" the ground and pulls the kite towards earth keeping it in the proper attitude into the wind.
Now losing my weight in flight does bring up safety concerns for those using a weighted tail. I was flying in a cleared remote area so there was little chance of injury but as with camera rigs, do make sure if you use a weighted tail that it is secure and don't fly over others or other's property.
Note: All video tracks were gathered at the same focal length, only altitude and time varied.