Research Project using a possible new knot

[Dan_Lehman]

Great, thanks!
I'll first just insert/add comments; it will take some time
to put together my list (but I might make some mention here).

But we should be specific & comprehensive, so to enable a correct
and helpful result be obtained.  And I think that it's not merely
a right-or-wrong citing, but maybe just an quite unexpected one,
too (along with right-or-wrong things)!

Ok here are some of the problems that caught my attention:

• It appears that the author tested only one sample of each knot in each type of cordage. ... .
    Here I'll want to depart, for the implied cost of testing
    quickly becomes prohibitive --as there are many factors to cover,
    and the need-so-many-test-cases sends the total skywards!
    .:.  There needs to be some way to make limited testing work.
    (Perhaps by testing multiply knotted specimens, which would in
     less material and one running of device produce "at least this much"
     force measures for the lot, though only one knot might fail (some
     others might show beginnings of failure.  That would be giving a
     "least force" break, which in any case is what users want! )
• The author did not identify the Dyneema and twisted cordage used, ...
    Yes, and from a place that must have just that information.
    Well, one can try to match appearances in the product catalogue ... .
• The author compared measured breaking load to rated breaking strength (average or minimum?  - didn't say). ...
     Was this in fact explicitly stated (vs. to my quick reading, just ignored)?
     I believe that the Cordage Institute, which has some say in how things
     are marketed --but which (also IIRC) differed from some other standards
     bodies' procedures--, had ruled that spliced strength would be quoted line
     strength (i.e., there would be no sophisticated estimation of tensile strength,
     but rather the practical measure (and maybe of 5 samples?).
• The author admits that he was supposed to preload cordage to 50% before testing but did not.
    Again, was the absence of such pre-loading explicitly admitted?
    IMO, I found it surprising.  Granted, usage should do something
    in terms of conditioning lines; but common usage (a) shouldn't
    be reaching 50% (!!) and (b) such usage, well, is just that, and so
    should one be guarding line use (like break-in driving of a vehicle)
    on the risk that as new line it is somehow weaker?  --never heard
    this before (and most testing seems to be done with new rope
    which has been provided by manufacturer or ... for the test.
• The author does not differentiate between knot strength and knot security.  Anybody who has attempted to tie knots in HMPE cordage knows that all the knots tested will slip.  The author, however, reported them as "breaking".
    Well, here we have an issue in the compound structures of
    high-strength (esp. HMPE, it seems) core and protective mantle:
    that loading can be uneven, and esp. that the core does slip within
    the sheath and thus leave the latter to bear force and ... break.
    (Interesting :  I recall some fishline testing in which HMPE line
    was made strong by putting it in some kind of sheath, and even
    then in an overhand knot (!! !?).)
• The author confuses core with cover in Dyneema line.  He repeatedly refers to "Dyneema cover" breaking.
    Eh, I'll see this not a resulting from so much misunderstanding
    --though I do fault the report from not expressly addressing
    the slippage-of-core issue!--
    but just quick shorthand rope ID --"Dyneema" is the name for
    that particular cordage (and, as you note, one that leaves us
    wondering ... , versus a product ID with specifics.
• The author assumes that breaking strength is the primary consideration for selecting a knot.
  ... Even when knotted, they will rip the clew out of the sail before the line breaks.
     Agreed, that seems to be the implication.  (And I recall in some
     boating forum someone remarking that after a hurricane had
     ravaged some port/marina, there were no breaks of bowlines
     in mooring lines (to compare vs. splices) --breaks in any rope
     came from hard edges, if ... .
• In the text, the author notes that it is important to properly taper a splice, but (in the photo provided) it appears that he did not.  Now, I admit that the photo isn't perfectly clear, and I might be mistaken, but if he didn't taper the bury then that would explain why his splice broke where it did, and why it broke at a lower-than-expected load.
    Oh, yes, I was amazed at the splice forces, which as I remark
    above are for some users to be what IS a rope's strength (so,
    by definition, 100%).  And at the end of the stuffing?  --makes
    me wonder what sort of splice, and likely lack of taper, and ... .
    (I don't have much familiarity with splices, but it seems odd,
    and esp. the break forces low.)
• What about other knots?  The author said he tested the Carrick Bend, but did not report the results.  The Angler's / Perfection Loop would have been an excellent candidate to test, especially in stripped cover Dyneema.
    I noted that, too --re "carrick bend" (which, I'll remark, gets
    rather LOW breaks in some old, natural-fibre climbing rope
    testing, I've recently seen!).
    Also, note that it's a double strangle knot used, per image.
    AND that sometimes the pre-tested tied knot is shown
    in a different orientation to the broken specimen!
    (And I'm having trouble e.g. figuring out the broken
     sheet bend, as NEITHER side seems limited enough
     in high-load abrasion to have been the U-part --but certainly
     not the implied side.
     (In some seen samples, it has been the U-PART THAT BROKE,
     where the *loop* part turns around and constricts it.)
• Oh, and Get your knots right!  The author admits that he mis-tied the bowline in the twisted rope, but includes it in his test data anyway.  You can't test knots If you can't tie them correctly!
    Yeahhh, but one can dispose of the myth that the bowline
    was all so "mis-" tied!!  The tail-outside version is the one
    less vulnerable to slippage on ring-loading.  AND one can
    dress the knot such that the tail lies neither so
    in the plane left/right of the returning eye leg,
    but *above/below* it --e.g., to anticipate the SPart's draw
    (so that only upon considerable loading does the tail
     then assume the shown-in-book-images position!).

Given the number and severity of these issues, I think its wrong to draw any conclusions about "best knots" from these tests.
  AH, and we're still awaiting MY citations!

Marlow Ropes provided the test lab and presumably knows how to use it properly.
  Should we really presume this?  --might be more of a leap of faith!

  They should have given the author guidance on devising and performing tests that would yield meaningful results.

Such an obviously flawed article makes me doubt the veracity of everything else published in Yachting Monthy.
  You'd be even more appalled at Practical Sailor's 2001-09? testing!
  Cf. www.practical-sailor.com/issues/27_17/features/4578-1.html
   & i.p. www.practical-sailor.com/newspics/charts/27-17-Bench.pdf
    --look at top line comment re the bowline in Yale Light (rightmost col.) !!!

Regards,
Eric

Thanks.
Others?!

--dl*
====

[NautiKnots]

Well, according to a couple of sample size vs. confidence calculators I just used, your chances of getting the "right" answer with a single sample is about 2%.  That is, any given knot breaking strength test yields the "wrong" value 98% of the time.

I use the terms "right" and "wrong" very loosely here, as population sampling is much more complex than that.  A better way of putting it is that the knot test strengths listed in this "study" have a margin of error of 98%.

Which is to say, that the absolute numbers given, and therefore the percentages calculated are meaningless.  Conclusions drawn from those percentages are more likely to be wrong than right.

Yes, it is prohibitively expensive to test multiple knots in different brands/types/sizes of cordage to 95% confidence within a 3% margin of error, but that doesn't make single-sample testing valid.

An anecdotal test is just that - anecdotal.  One cannot draw meaningful conclusions about a population based on a single sample.

[Dan_Lehman]

I've taken everything onboard and agree it is the tensionless hitch (thanks Tom_Zal)

I still object to this, as that name is rightly indicative
of the criterion for inclusion --i.e., no tension on the
knot of the noose structure--, and in the case at hand,
we see such tension on that knot that it significantly
(?!  well, do we know the % break?) weakens the
>>noose's<< SPart via constriction.
(It would be like calling the tautline h. or midshipman's h.
--those noose structures that are supposed to be effectively
   adjustable (fixed) eye knots--
"non-slipping adjustable eye knots" even though in many
circumstances they unfortunately slip.)

Dan, I will also include your recommended hitch within my testing.

What I aim to do is try to improve on the given knot's
constriction of the noose-SPart by having the >>knot's<<
SPart taken farther around the object (more nearly 360deg)
and connecting to the noose SPart away from where some
contact is initially made, considered from the perspective
of that SPart coming into the knot --at which point there
will be the further wraps of the anchor hitch / Blake's h.
which of course will have less tension than the initial
turn which is close to the object.  And with the latter
knot aforementioned here (B.h.), one additionally *pads*
the noose SPart with the tail of the knot.

So, this should remove rupture --at least until a really high
load-- from the entry of the noose SPart into the knot.
. . . in (my) theory.

(Another such attempted amelioration of this constriction
would be to make the few round turns and then tie off
with a timber h. that has a full/round turn or two
around the noose's SPart, and dressing that wrapping
so that both legs of it come at the near-object end of
the coil (so, sort of *"cascading"* what would be the
away end from the pure helix.
(Friction hitches that have many turns can do this
naturally, the away turn being pulled so that it sort
of *un*wraps at the away end and does some broad
wrap up to there.)


--dl*
====

[Dan_Lehman]

My short answer is that one looks to ranges of values,
not some "right" one --I guess you're thinking of the
mean of many many samples.  One might do some
full testing of a few things so to get an idea of whether
e.g. cordage has much variance,
and then do scattered testing taking the above as an
assumption and not caring for many samples so long
as the few are within an expected range.

There's got to be a way forward.
And for much of a benefit, not so much particular values
but more for seeing how/where/why things are breaking,
so to build theories to explain (and iteratively seek knots
that would via their particular geometry be good tests
of these) knot failure.

(-;

[agent_smith]

Quote from: Tim on June 26, 2018, 12:40:47 PM

    I've taken everything onboard and agree it is the tensionless hitch (thanks Tom_Zal)

I still object to this, as that name is rightly indicative
of the criterion for inclusion --i.e., no tension on the
knot of the noose structure--

Okay...

There is some difficulty in attempting to define what Tim has tied.
It is not new - that is 100% certain.

The knot is actually a 'composite' structure.
It consists of turns around a pipe/rod/post followed by a strangle around its own SPart (standing part).

As a complete entity - it is a noose. It is not a fixed eye knot.

The first segment of the structure is the turns around a pipe/rod/post - and this is making use of the 'capstan effect'. Tim could just have easily tied his knot around a tree or a bollard on a wharf.

I agree that the common name 'tensionless hitch' is a misnomer.
There is tension in a tensionless hitch! Force enters the 'core' via the SPart. That force is quickly dissipated in accordance with the capstan equation. Note that in the capstan equation, friction plays an important role. Obviously, making turns around a tree in comparison to turns around a carabiner has a different coefficient of friction.

Tim could have kept increasing the number of turns around the carabiner.

As with any so-called 'tensionless hitch', the tail must be secured in some way.
Tim chose to secure the tail via a 'strangle' (ie a strangled double overhand knot). There are other ways to secure the tail...using a strangled type method is economical.

Fundamentally, Tim's knot closely matches #409 Poachers noose.
There is no doubt that fundamentally, it functions as a noose.

One can easily modify #409 Poachers noose by adding an extra turn.

So you could argue that Tim has simply tied #409 Poachers noose with an extra turn.
You can just keep adding more turns to modify #409.

I sometimes use the phrase; "Cant see the forest for the trees".

Some people see #409 Poachers noose.
Some people might see a round turn followed by a strangled double overhand knot.

I think ultimately, Tim has tied a 'composite structure' consisting of:
[ ] A round turn or several turns
[ ] A strangled double overhand knot

As a completed structure, it functions as a noose.

The addition of the extra turns makes use of the 'capstan effect'.

A tensionless hitch makes use of the capstan effect.

One could build a strong argument that Tim's creation consists of a tensionless hitch that is secured by a strangled double overhand knot.

Mark G

[NautiKnots]

My short answer is that one looks to ranges of values,
not some "right" one --I guess you're thinking of the
mean of many many samples.

It all comes down to:

• what do you want to know,
• how confident do you want to be in your results, and
• how much risk of error you are willing to take.

If a knot always broke under the exact same load, then you would need only one sample to find out what that load was.  If, on the other hand, different instances of the knot break at different loads, then you need more samples to understand what to expect of that knot in the future.  That is, if our goal is to understand and make predictions about the behavior of the population at large, then our sample set needs to be large enough to have the same distribution of values as the whole population.   As the size of the sample set increases, our confidence in the result increases, and the risk of error decreases.

A sample size of one yields a result (in practical terms) in which we have no confidence, and a high risk of error - regardless of what one is trying to measure (average, minimum, or maximum strength).  All you learn is that the overall population contains that value - not what one can expect from other members in that population.  Tests such as those performed by Yachting Monthly and Practical Sailor, which use a single result to extrapolate the behavior of the general population are worse than useless.  They reveal incompetence that borders on negligence.  The results mislead more than they inform.

One might do some full testing of a few things so to get an idea of whether
e.g. cordage has much variance,...

Depending on what you are trying to determine, and what the distribution of the population looks like, you might be able to get the desired confidence with a smaller sample set, but that size will always be significantly greater than one.  Let's say, for example, you want to know what the probability is of a given knot in a particular rope slipping before it breaks.  So, you start tying samples, and pulling on them until they fail (one way or the other).  If, after reaching 11 samples, you found that 10 slipped and 1 broke, you can conclude that the knot has a 90% chance of slipping, with a 95% confidence in your result, and an 18% margin of error.  If, however, as you test, you find that half of the time the knot slips and half of the time it breaks, you need to take 30 samples to conclude that the chance of slipping is 50% with the same confidence and margin of error.

There's got to be a way forward.

Wishing for something doesn't make it true.  If you want to model the general population, then your sample set needs to be big enough to accurately reflect the distribution of the whole population. 

I hope that helps,
Eric

[Dan_Lehman]

Here are some further comments on the above-cited
Yachting Monthly knots article.

1) "... the strongest knot sailing knot ..."
As noted above already, this "strongest" characteristic
is both not-so-easily-firgured and not really the key quality.

2) To

the RYA recommends a round turn and two half-hitches
for fastening a mooring line to a pontoon cleat because
it can be tied and untied with load in the line. I?ve always
used a pre-tied bowline dropped over the cleat,

I wonder what happened to the cleat hitch, which also
can be un/tied under load, and was of course intended for cleats!?

3a) In the embedded how-to-tie-some-knots video,
the fig.8 eye knot is supposedly dressed to the dubious
rule of putting some end on the outside; the resulting
knot is a poor excuse for something that has been dressed,
whatever the goal!

3b) The fisherman's knot is tied w/discordant (i.e.,
different-handed) overhand compoents.  There is no
discussion about this aspect, and IMO this is likely the
inferior variation.

3c) The grapevine/dbl.fisherman's knot is deemed the strongest
--and it probably was--, but the guy in the video only shows
his ineptness at knot tying by struggling to make the
component strangle knots,
putting them also in a discordant/opp-handed version,
BUT for the final image it seems that someone (else?!)
stepped in and tied the knot concordantly and with
sufficiently long tails for the camera !!  (I wonder if
this mystery tyer can be rented?)

4) Their stated lack of "proof-loading" to 50% of tensile
strength is a surprising remark, IMO.  I've heard of some
cyclic loading done pre-testing, or otherwise no indication
of such conditioning, and of course of testing non-new
cordage, but ... 50% ?!  That seems high.
And

Dyneema and Marlowbraid to see how it affected their known strength,
and found that the break load of both lines was reduced to a startling 35 per cent of their ISO specification.

"Startling" indeed. IIRC, pure Dyneema 12-strand (NERopes)
was said to have broken --don't know about any pre-loading--
at about 33% (for some few knots NER broke for me via
Brion Toss's invitation), but that seems low for conventional
cordage.  (One can wonder what "break" means, here, for
as noted elsewhere the "Dyneema" line often broke in the
sheath only.  With this mid-line stopper knot, though, the
"pulled through" potential didn't exist.)

4) "Then we looked at loop knots" :: yes, how surprising
to read that they included these --RT + 2HHitches, anchor bend?!

5) Although there are two mentions/occurrences of
"carrick bend", there is nothing further --no report of
actually testing it (as stated it would be) or showing
a result for it.  (I guess that along with knotting
editing is a lacked skill at Yachting Monthly!)

 - - - - - T B Continued - - - - - -
(Goodness, there is so MUCH to cite.)

--dl*
====

[Dan_Lehman]

It's interesting to see that for the dbl. sheet bend
the "Dyneema" & Marlowbraid lines broke at opposite
ends.  I figure this :: there was more slippage of the core
in the *hitching* part than the U-part/bight, so the
former taxed its sheath more (in the yellow hitching
end) and broke there; whereas the Marlowbriad's
like greater constriction by the hitching part around
the bight/U-part's ends led to the break coming in
that U-part's SPart.
In other words, greater constriction was the same,
and the difference resulted from core slippage or not.

IN THEORY ...   

--dl*
====

[agent_smith]

Hello Dan and 'NautiKnots'

You guys are drifting way off topic here... but, I definitely like, and am very interested in setting up proper parameters for knot testing and future knot testers in general.

This discussion should be in a new thread.

I find it ironic that the issue of poor knot testing and poor reporting keeps raising its ugly head.
I think the IGKT must accept at least part of the blame/fault that knot testers are continuously getting things wrong.

The IGKT should set up a knot testers page - which could act as a one stop shop for all future knot testers. They can browse to the IGKT forum and review the "Knot testing guidelines'.

Currently, virtually all knot testers act in isolation - with little to no peer reviewing of their work before it is published. They just test and publish = and the same old tired mistakes are repeated endlessly.

I know some members of the IGKT have tried to address this issue in the past but, it has bogged down or run out-of-steam.

Surely we have enough 'experts' in the IGKT to address the issue of knot testing?
Is there any reason in principal why we can't setup a knot testing page to provide a useful set of guidelines and principles for all future knot testers?

I challenge interested members of the IGKT to take this on as an issue to resolve before the end of 2018!

JFK once said... "We choose to go to the moon not because it is easy, but because it is hard"

Surely if we can put boot prints on the moon, we can establish a set of Knot testing guidelines and principles?

Mark Gommers

EDIT / PS

I think a new category needs to be added to this forum - titled:  KNOT TESTING GUIDELINES AND PRINCIPLES

This will make it easier for the general public and interested future knot testers to search and find this site.

[Sweeney]

I think a new category needs to be added to this forum - titled:  KNOT TESTING GUIDELINES AND PRINCIPLES

This will make it easier for the general public and interested future knot testers to search and find this site.

I think it would be more accessible if we put agreed guidelines and principles on the main website, perhaps based on a discussion here about what these guidelines and principles might be. Actual results and ensuing discussions would continue to be posted here perhaps with a link from the main website. In theory this shouldn't take too long or create any great disagreement - that comes from results rather than methodology one would hope.........

Sweeney

[Dan_Lehman]

Further comments on the Yachting Monthly testing.

6) The RT + 2HH will have different behavior on
smaller vs. larger-diameter objects; with relatively large
ones, the clove h. (the *knot* of this *structure*)
grabs onto the (noose) SPart more than it will on a ring.
(I recall test data where this knot rated higher on a ring
but lower on a pile than the anchor bend --which
I presume did so well on the latter by having its
collar yielding enough to leave the SPart more straight
and arriving tangential to the object (my guess)!?)

One can try to ameliorate the noose's effects by settting
it firmly in an iterative manner of pulling hard on the
(noose) SPart and pushing the knot back snug to the
object --of trying to work tight the round turn.   From
such a setting, the draw of the SPart ought be a little
less on the knot and thus the angle of contact better
and ... .   (need that comprehensive testing to see if ...)

7) Marlowbraid tied in a bowline breaking at just 47%
really surprises me.  One can wonder at why --at whether
the assessed tensile strength, e.g., is too high?  (In
some testing of rockclimbing rope the bowline gets
up in the 70%.  Dave Richards's testing of 12.5mm low=
elongation ("static"), 10mm dynamic, & 7mm "accessory"
cordage got (per 5 test specimens each), resp., 63.3-63.1-67.1%.

8.) The fig.8 eyeknots are hard to figure out from the
report's images, as they are shown in lousy dressings
--which apparently was the case with the broken knots.

I find it most peculiar that for the "Dyneema" rope the
break occurs at one of the eye legs --not in the
SPart ?!?!?  Huh?  --and at only "35%" of some supposed
tensile strength, which should mean only 17% in the
broken part (!!) ?!
Were there some slippage of core, that should pull
core from the eye (as seen with the bowline, e.g.)
and LESSEN (remove!) load from the sheath,
not break it!


--dl*
====

[Dan_Lehman]

There's got to be a way forward.

Wishing for something doesn't make it true.  If you want to model the general population, then your sample set needs to be big enough to accurately reflect the distribution of the whole population. 

If you wanted to know the quality/composition
of a gallon of milk, how many samples would
you take for analysis?  I'm thinking that cordage
is for the most part reasonably assumed to be
pretty consistent, and this could be one factor
in knotted-rope strength (to step away from
attributing strength to *knot* en vacuo!) that
can be reduced in number.

And given some experience in test results with
some knots, one might come to believe that
skilled tyers can replicate at least some knots
well enough to not need as big a number as
might be wanted for more complex knots!?

But we surely aren't going to do all the multiples
of sampling per all the multiples of probable
factors that influence strength --a number that
is huge.

Meanwhile, there needs to be a great tightening
of knot-tying skill and reporting detail, irrespective
of test-specimen count.  (An extra dozen or two
tests of fig.8 eye knots (mis)tied as done in Yachting
Monthly would not be a help!)

There is this irony :: Dave Richards remarked in his
report on testing 7mm & 12.5mm low-elongation &
10.5mm dynamic kermantle ropes that he was who
tied all of the knots --the implication being that in
this way he ensured some uniformity to their formation
(it actually doesn't necessarily mean even this, were
his tying skill not great and his discrimination among
possible versions lacking!).
Well, even assuming it was so,
the implication must be that HIS results cannot be
so meaningful/applicable to all the rest of us, who
will be tying our own knots (in different ropes)!


--dl*
====

[NautiKnots]

If you wanted to know the quality/composition of a gallon of milk, how many samples would you take for analysis? 

Well, that depends on what you want to know about milk.  If you only want to know about that particular gallon at that particular time, you might only need one sample.  Don't, however, expect to make useful generalizations about other gallons of milk.  What if the gallon you had tested had spoiled?  Would you conclude that all milk is spoiled?  If you tested 2% milk, would you conclude that whole milk has 2% fat content?

I'm thinking that cordage is for the most part reasonably assumed to be pretty consistent

I think that's a faulty assumption.  Would you assume that knots tied in 12mm double-braid spun polyester would hold the same as those tied in 6mm twisted nylon?  How about 3mm single-braid Dyneema?  In terms of the milk test above, would you assume that all breeds of cattle produce the same quality milk?  If so, you'd be sadly mistaken.

And given some experience in test results with some knots, one might come to believe that skilled tyers can replicate at least some knots well enough to not need as big a number as might be wanted for more complex knots!?

Although unskilled knot tyers might produce knots with greater variability (leading to an increase of samples needed), skilled tyers producing identically dressed knots does not reduce the needed sample size.

But we surely aren't going to do all the multiples of sampling per all the multiples of probable factors that influence strength --a number that is huge.

True.  That's why it is so important to construct your test carefully in order to actually find out what you'd like to know about a particular knot and/or cordage.

Meanwhile, there needs to be a great tightening of knot-tying skill and reporting detail, irrespective of test-specimen count.  (An extra dozen or two tests of fig.8 eye knots (mis)tied as done in Yachting Monthly would not be a help!)

True.  A disconcerting proportion of knot tests are performed by people who are either unskilled at tying knots, or unskilled at testing.  Either one can easily render the test results meaningless.  In the Yachting Monthly test (for example) the test was evidently performed by a reporter who didn't know how to do either.  His only qualification was the ability to write an article.

If you want knot tests you can extrapolate from, they need to be constructed and performed by people who know both how to tie knots, and how to test.

-Eric

[agent_smith]

From NautiKnots:

If you want knot tests you can extrapolate from, they need to be constructed and performed by people who know both how to tie knots, and how to test.

And, what to test.

Its the 'what to test' that is often the key problem.

Correctly tying knots and knowing how to set up your test machinery (including presumably statistically valid sampling) is still nothing compared to knowing 'what' to test.

And this has drifted waaaayyyyyyy off topic from the original poster's points.

[Dan_Lehman]

If you wanted to know the quality/composition of a gallon of milk, how many samples would you take for analysis? 

Well, that depends on what you want to know about milk.  If you only want to know about that particular gallon at that particular time, you might only need one sample.  Don't, however, expect to make useful generalizations about other gallons of milk.  What if the gallon you had tested had spoiled?  Would you conclude that all milk is spoiled?  If you tested 2% milk, would you conclude that whole milk has 2% fat content?

Though, re underscored part, the gallon might itself
be seen as a sample of the greater-batch-of-that milk
I'm pointing to *evenness* of a thing; your supposed
challenges to this miss that point --of course one wouldn't
think this 2% milk gallon implied all milk was so.
And I don't expect this climbing rope to imply things
for THAT one, or yachting ropes, or ... ; and I will
urge the "KNOT strength" is better conceived as
"this-material-so-knotted" strength.

SO, given our gift spool of rope for testing --the milk gallon--,
why test 30 specimens --THAT is my question!

I'm thinking that cordage is for the most part reasonably assumed to be pretty consistent

I think that's a faulty assumption.  Would you assume that knots tied in 12mm double-braid spun polyester would hold the same as those tied in 6mm twisted nylon?  How about 3mm single-braid Dyneema?

Whoa, then I'm poorly stated :: goodness, no,
I meant that a given spool of rope is consistent
with itself,
not that any rope is ... any other.  (See my point
re restating "knot strength" to "this-material-so-knotted strength".)

And given some experience in test results with some knots, one might come to believe that skilled tyers can replicate at least some knots well enough to not need as big a number as might be wanted for more complex knots!?

Although unskilled knot tyers might produce knots with greater variability (leading to an increase of samples needed), skilled tyers producing identically dressed knots does not reduce the needed sample size.

Why not?  Given our milk gallon, what is to gain?
(At least, I think that after some analysis of tests,
one should come to a conclusion that multiple-knots
testing doesn't show much new, unexpected.

(How interesting to see in one of Dave Richards's
tests that it was the grapevine bend that had LEAST
stnd. deviation, even over pure-rope testing!
Similarly, how puzzling to see that both he and also
CMC Rope Rescue thought to test "fig.8 re-woven
AND fig.8 on a bight", as though the knot should
care how it came to be!  NOW, yes, it would be quite
informative to see from field examination if indeed
users came up with different geometries per tying
method --I rather expect that they might,
even to the point of which end is loaded.  BUT,
do that and be explicit about it; just giving the
different names and having same tyers, one would
hope that per "dress & set" action the knots came
out the *same*; in fact, diff.s (5 cases each) were
quite close.)

But we surely aren't going to do all the multiples of sampling per all the multiples of probable factors that influence strength --a number that is huge.

True.  That's why it is so important to construct your test carefully in order to actually find out what you'd like to know about a particular knot and/or cordage.

I'm thinking of moving through the multifactored
*knot* space as an exploratory expedition, and
hoping to travel efficiently so far as test-results
fall in expected ranges,
some of which ranges will become established by
more thorough testing,
but some of that easy movement will come by
doing with less.
.:.  Throw the many-sampled cases out where
a stout pillar looks to be needed.

AND I have argued for --where possible-- the
single *test* of multiple-tokens of a knot,
stringing a line say with 5-10 identical knots
in it, to get a break, which is reasonably argued
to better indicate minimum strength, and then
have all those near-rupture (presumably) knots
to examine --maybe to further load and see ... .

At least :: like eye knots on the ends of single specimen,
and round slings with TWO of end-2-end knots (hoping
to avoid the knot-compression-imbalance that such
a structure with one-knot-only can yield (the knotted
side i.e. lengthening per knot compression and that
not being fully transferred/balanced to unknotted side
by friction at pins --maybe a pulley would do, but ... ).

--dl*
====