yosemite finish on a figure 8 knot?

J. Albers wrote:

That was a whole lot of trash talking for what was actually a pretty accurate statement. A sample of one is pretty meaningless precisely because you can't assess how likely the result was in the first place. So your snotty statement about p values is moot because a p value can't even be calculated anyway. Perhaps you should have lectured him about the wonders of bootstrapping or jack knifing...of wait, you can't,  because those are meaningless techniques also with a sample of one (gigantic eye roll). 

We have a sample size of one for each knot test, but we can draw conclusions about how much variance we'd expect to see upon further tests based on other experiments where the same knot was tested multiple times--for example the break tests of Lee's Locked Bowline that rgold linked (I'm grabbing the highest number I saw from the video which was fast, so please excuse misreads--I'm sure Lee has better data):

Trial 1: 2432

Trial 2: 2306

Trial 3: 2367

So with three tests of the same knot, the breaks were all within ~6% of each other. Compare with the tests of the figure 8 follow through with and without the first yosemite follow through:

Without Yosemite: 700+

With Yosemite: 253

So the difference here is >60%.

I'm using rough numbers here because I don't want to overplay my hand: I'm aware that they're different experiments with different methodology. While I could calculate a probability of seeing the difference in the cross-loaded tie-in test given the distribution of the values in the locked bowline test, it would be easy to overstate the significance of that number. All I'll say is that we can't simply dismiss it because of possible variance that would average out in further tests, because the variance between the figure 8 with and without the yosemite follow-through is an order of magnitude larger than is easily explained by sample size.

After watching the first video I would say that it doesn't really matter if it is a sample of 1 or 100 only because it is not only highly unlikely to cross load the knot itself but also because during the (lead) fall the rope takes some of the kinetic energy. That is like saying that seatbelts don't work in a high-speed collisions.

Anyway, my original question was how to tie a yosemite finish on a figure 8 knot?

The failure of the knot by itself, isn't the only evidence.  Ring-loaded figure-8's (and the flat figure 8 EDK) roll because the strands that are pulling in opposite directions stretch and then roll the collar they emerge from over the rest of the knot. This rolling process eats up some of the knot tail and cannot continue indefinitely without knot failure.  The Yosemite finish results in a big collar and a small remaining portion of the knot, and it is significantly easier for that big collar to roll over the the small remaining portion, because it doesn need much rope tension to stretch it enough to fit over. You can easily see this phenomenon at work in the video.  

So what we have here is the outcome predicted by a logical analysis confirmed with, yes, but a single test.  But we already knew the Yosemite finish knot should roll at a lower load than the regular figure-8, the only real question is, how much lower might the failure threshold be?  The test indicates  that failure can happen at a very low value---something like triple body weight.  I think that's cause for serious concern as a worst-case scenario candidate, but without some estimate of how likely that low-threshold failure is, we really don't know how worried we should be.  Maybe we just stumbled on an event of cosmic insignificance, probabilistically speaking. (I doubt this, because it is hard to imagine random variations in the nature of multiple samples tested with the same apparatus that would cause a huge standard deviation in the results, but that's just my intuition speaking.)

To me it doesn't matter much.  I find it clear without testing that the Yosemite finish is going to make the figure-8 more likely to fail under ring-loading.  If the Yosemite finish conferred some wonderful advantage, we might have an interesting dilemma, but that's not the case---it has almost no real use and adds a weakness that might or might not ever become an issue.

Thanks Rgold. I'm going to keep using that knot. So were the 8's tested in the same manner? 

David Kerkeslager wrote:

If you stuff the tail into your hard points and expect it to stay there through a climb, you have either an excessively long tail or unrealistic expectations.

I figure that if the tail is long enough to clip..

Be sensible, have you EVER heard of a rope breaking at a knot in the real world?  They have come untied, ropes have been cut but they don't break. It seems like you are looking for things to worry about.

Forrest Halley wrote: Thanks Rgold. I'm going to keep using that knot. So were the 8's tested in the same manner? 

Obviously, yes.  An eight will roll too when ring-loaded, this is why you shouldn't use a flat 8 as an EDK.  You can see the 8 about to roll in the video.  But add the Yosemite finish and you lower the rolling threshold.

kemple sr. wrote: Be sensible, have you EVER heard of a rope breaking at a knot in the real world?  They have come untied, ropes have been cut but they don't break. It seems like you are looking for things to worry about.

This thread is not about the rope breaking at the knot.

Greg Miller wrote: You have way to much time on your hands

Ankle injury is keeping me off the rocks at present.  Work out during the day and do some posting at night...   

I test things for a living. A single sample means next to nothing. Often we have out lying samples at the high and low end that we disregard. We run a setup sample that gets thrown out, then the 2nd is usually thrown out too. Then the test starts. This is only after several engineers check it out. We dont even talk about tests found on youtube. You have no clue if their equipment is calibrated. The UIAA testing is the way to go anyway. They attach a weight to the piece of gear and drop it. pretty simple. Also, your climbing rope will never see loads like that unless youre pulling your car out of the mud. When you whip on a rope you'll see high loads for a moment but you wont see 700kg for 30 seconds. That test is not representative of any real world scenario. That test is like off roading an airplane and saying it dont hold up

Justin Veenhuis wrote: I test things for a living. A single sample means next to nothing. Often we have out lying samples at the high and low end that we disregard. We run a setup sample that gets thrown out, then the 2nd is usually thrown out too. Then the test starts. This is only after several engineers check it out. 

Doesn't it depend on what you are testing and what the consequences of disregarding an outlier are.? If a rocket blows up on the launching pad, do you view that as a "next to nothing" result?  What if you already have an independent analysis that says the rocket should blow up and then it does.  Is that too a "next to nothing" result?

Also, your climbing rope will never see loads like that unless youre pulling your car out of the mud. When you whip on a rope you'll see high loads for a moment but you wont see 700kg for 30 seconds. That test is not representative of any real world scenario. That test is like off roading an airplane and saying it dont hold up

This seems to be true, but there are cases in which it has been shown that slow-pull tests can be substituted for shock-loading tests, because the results are comparable even though the mechanisms are different.  Whether the knot-rolling is an example I do not know, but it means slow-pulling can't be rejected out of hand.

The sad reality is that there is hardly a statistically valid test in all of climbing, because it is impractical to conduct the requisite number of trials in the right way.  So pretty much all we have as a basis for decisions are results with little statistical validity.  In the face of this uncertainty, we still have to do our best to make decisions that are as sensible as possible.  I don't see throwing up our hands and disregarding all evidence, regardless of how imperfectly obtained, as an optimal way of dealing with the uncertainty.

Justin Veenhuis wrote: I test things for a living. A single sample means next to nothing. Often we have out lying samples at the high and low end that we disregard. We run a setup sample that gets thrown out, then the 2nd is usually thrown out too. Then the test starts. This is only after several engineers check it out. We dont even talk about tests found on youtube. You have no clue if their equipment is calibrated. The UIAA testing is the way to go anyway. They attach a weight to the piece of gear and drop it. pretty simple. Also, your climbing rope will never see loads like that unless youre pulling your car out of the mud. When you whip on a rope you'll see high loads for a moment but you wont see 700kg for 30 seconds. That test is not representative of any real world scenario. That test is like off roading an airplane and saying it dont hold up

As it happens the UIAA/EN certification tests are single item/event tests. You provide the specified number of samples and one is selected at random and tested. in my case (bolts) the sample size is ten and one is tested radially and one axially.

Even as a rappel knot, 700kg is more than 1500lbs. I think its doubtful it would fail in any real climbing scenario. Also a figure 8 follow through with a Yosemite back up is a different knot than a figure 8 bend with a Yosemite backup. I think it would act a lot different. In the real world if you were going to join 2 ropes wouldnt you  back up both sides? What about different size rope? Im sure theres good numbers on all of this somewhere, UIAA website maybe. Or shoot an email to a rope manufacturer. BD seems cool with doing real tests.

rgold wrote: When a test produces a catastrophic failure, that result can hardly be dismissed as meaning "almost nothing."  

Except when we know nothing about the testers, methods and consistency of tying the knots.  When I look at the start of the test on knot 2, it "appears" to be very loosely tied.  So, is that a factor?  Maybe not.  But, why does everyone say when using the EDK that you must dress and tighten every strand for it to be safe?  And, the video that appears right after one of the links you provided asserts how important it is to dress one of the bowlines sequentially.  

Could someone give some example when a tie in knot would ever get ring loaded.  Yes, I know, someone died a few years back while aid climbing and cyclic loading the tie in knot.  But, when would a free climber ring load, even by accident?  Even in the method of belaying off the tie in loop doesn't ring load, as you have asserted.  

I'm not giving and cred to this video, although I would love to see better testing on this very topic and all the variation finishes resembling the YOS finish, which is what the op has asked, not should I start using a bowline.  We know you can go on for 7 pages about how great the bowline is.  But, it would be nice if we could stick to the op's question.

Personally, I didn't worry one iota about ring loading my tie in.  

rgold wrote: Doesn't it depend on what you are testing and what the consequences of disregarding an outlier are.? If a rocket blows up on the launching pad, do you view that as a "next to nothing" result?  What if you already have an independent analysis that says the rocket should blow up and then it does.  Is that too a "next to nothing" result?

This seems to be true, but there are cases in which it has been shown that slow-pull tests can be substituted for shock-loading tests, because the results are comparable even though the mechanisms are different.  Whether the knot-rolling is an example I do not know, but it means slow-pulling can't be rejected out of hand.

The sad reality is that there is hardly a statistically valid test in all of climbing, because it is impractical to conduct the requisite number of trials in the right way.  So pretty much all we have as a basis for decisions are results with little statistical validity.  In the face of this uncertainty, we still have to do our best to make decisions that are as sensible as possible.  I don't see throwing up our hands and disregarding all evidence, regardless of how imperfectly obtained, as an optimal way of dealing with the uncertainty.

In virtually every example of comparative slow-pull to dynamic testing I´ve ever heard of the results are comparable or have the variation we would expect from theory though creep and rolling in ring-loaded knots may be different. Since we know it does occur only the loading is the question, since it is easy to avoid it´s irrelevant to know exactly at what load it occurs. As the Yoemite finish reduces the force required to roll the knot all the more reason to avoid it, particularly as someone such as a guide may become used to using it and then die when rappelling.

David Kerkeslager wrote:

The other reason I hear for using the Yosemite finish on the figure 8 is that the tail can get confused with the belay strand when you're clipping to bolts/gear on lead.

What?  I think your tail would have to be 3 or 4 feet long for this to be even remotely possible.  One only needs a 6 inch tail on a YOS finish to tuck in to tie in points.  Keeps things neat, clean and out of the way.

Greg D wrote:

What?  I think your tail would have to be 3 or 4 feet long for this to be even remotely possible.  One only needs a 6 inch tail on a YOS finish to tuck in to tie in points.  Keeps things neat, clean and out of the way.

Only if you clip above your head. :)

I do agree that if you keep things to about 6 inches you won't have this problem, though.

David Kerkeslager wrote:

Only if you clip above your head. :)

I do agree that if you keep things to about 6 inches you won't have this problem, though.

Even if you clip at your chest, one foot above your tie in, it would be a miracle to clip the wrong strand.  Your tie in would have to be at least two feet to make it even possible and 3 or 4 feet to make sustainable.  You are talking about a non issue, you do realize?   A detraction from the conversation, which you and Eli specialize in.

Greg D wrote:

Even if you clip at your chest, one foot above your tie in, it would be a miracle to clip the wrong strand.  Your tie in would have to be at least two feet to make it even possible and 3 or 4 feet to make sustainable.  You are talking about a non issue, you do realize?

Well, I have seen people clip the tail at the gym. So it isn't a complete non-issue for some people.

Seriously?  How long is their tail?  Sounds like Darwin. 

I’m really confused...how could you “clip” the tail, anyways?  Even an excessively long one would just slide right out and you would immediately know it was wrong.  I like a Yosemite over a barrel knot finish because you can clip lower (extended draws often end up right at the tie-in), but an 8 with no stopper knot is just fine.