> Good engineering is hard to come by. Good test data is even harder.
>

> Test data of composites is worthless: The normal scatter of data
> of identical handlaid panels is 25%. Comparing my panels to yours
> has even more scatter. Adding panels made with the "same fabric"
> made by different manufactures or diffent epoxies adds even scatter.

Just because there is variation in test results in no way implies the tests are bad. It just means there are deviations in the objects under test. This in it self is worthwhile data. However, averages can be determined. Determining averages does require testing multiple copies of the "same" thing.

Just because Gougeon Brothers only tested one example of any given layup does not make the test "bad", it just makes it incomplete. It is a good start, and as empirical data it is better than any mathematical model. The data can now be used to help verify a model. It will still be incomplete, but it is a start.

In the same way George's real-life test is not "bad", it is only one data point in an on-going test. It shows that layup was not strong enough for the conditions George was paddling. Another boat with the "same" layup in the "same" conditions might come through unscathed, and that would say that 50% of the time the layup is not strong enough. Keep doing the same and eventually you will have a significant test.