Why don't we create an informal speed rule here and give people a chance to work on outright speed and post results?
Simply, take the average speed over 500 meters from a GPS track log with at least a sampling rate of onces every 10 seconds (or faster). Calculate the average speed by deviding the distance covered by the time taken. DO NOT calculate average speed by summing all the individual speed averages and then deviding it by the number of samples. The latter can be much less accurate.
The track itself and the series of data points must be presented as proof. Also two runs most be made on opposite tack were the slower one of the two is presented. This to take out surface currents.
Everybody with a simple GPS logger can do this. 500 meter will take a shy 50 seconds to cover at 20 knots. This is sufficiently long to take out riding off waves or other effects.
Wouter