Look for older reports from rec.climbing. For example: http://canyoncollective.com/threads/some-knot-testing-data.7...

And https://user.xmission.com/~tmoyer/testing/EDK.html

Some of the testing folks at black diamond did a bunch of tests when we were all jabbering about knot safety about 20 years ago. Kinda sad the old posts are getting so hard to find!

There are two basic classes: knots on slippery material like monofilament and knots on ordinary stranded or braided line.

Fishermen have worked out good knots for monofilament, but they jam when used with stranded line.

The strength of a knot or bend in stranded line can be predicted without technical analysis by looking at the curvature of the line as it passes through the knot. The tighter the curve, the lower the breaking strength. The strongest bends, like the Zeppelin, have very soft, symmetrical curves. They are also the easiest to untie.

Yeah, me too, but I have never read of an actual knot failure causing an accident outside of user error (wrong knot, not tied correctly, incorrect application, etc).

edit: in reference to rock climbing accidents...

get into alpining and trad multipitch. Best way to learn a ton of knots, rope systems, and cool tricks!