Get access to everything we publish when you sign up for Outside+.
This safety article came to us via Philippe Batoux, manager of the testing facility at ENSA: the École Nationale de Ski et d’Alpinisme, in Chamonix France, an institute/school for mountain professionals that has existed by this name since 1943. There’s long been a debate, at least in the United States, about how safe it is to use offset knots—knots in which the two tails come out of the same end of the knot and the knot strands lie side-by-side, versus using one rope to follow the other through—to join two ropes for a rappel. Examples include the Offset Overhand (aka the European Death Knot) and the Offset Figure-8, which have been implicated in fatal accidents when the knots either unraveled or rolled
The conclusions reached by Batoux; Alexis Mallon, equipment testing facility; Amélie Gruyelle, equipment testing facility; Marco Perche, SIMOND; Elodie Le Comte – ENSA trainer; Johann Courcelle – ENSA trainer; Michel Fauquet – ENSA trainer; and François Pallandre – ENSA trainer are summarized here, but we recommend you read the full article below to understand the why of their conclusions.
TL; DR: Offset overhand and figure-eight knots are perfectly suitable for joining rappel ropes when using normal rappelling techniques IF:
• All knots are vigorously tightened—each length of rope (the four strands on either side of the knot) should be pulled hard up against the others.
• The free ends are at least 30 centimeters (12 inches) long.
We carried out a series of tests in our climbing equipment testing facility on the most commonly used knots used to join ropes of the same diameter for rappelling. We tested the two most widely used knots, the offset overhand and figure-8.
If these knots are tied correctly, with all four lengths tightened vigorously and 30cm lengths left at the free end of each rope, they are perfectly suitable for joining rappel ropes for a descent under normal conditions. However, if the factors above are not respected (non-tightened knot, short ends), these knots can fail during a rappel descent.
We measured the load exerted on the anchor during extreme rappelling situations (a fall onto the back-up prussik at the start of the descent/a sudden stop during the descent). The maximum loads obtained were equal to approximately 3 times the weight of the person when rappelling on dynamic ropes and 5 times when using static ropes. The knots we tested failed at 3 to 4 times these maximum values.
Climbers use offset knots to join rappel ropes because they sit to one side of the main axis, significantly decreasing the risk of them getting stuck during retrieval. Other advantages of offset knots are the simplicity with which they can be tied and the ease with which they can be untied after loading. For these reasons, offset knots are the most commonly used knots for joining rappel ropes.
Inline knots tend to be stronger, but they sit in the centre of the rope and get stuck more easily when retrieving the rope. Other types of knot may also be stronger (eg. double fisherman’s) but they are more cumbersome to tie.
Offset knots have been implicated in several fatal accidents in North America, where they are known as ‘European Death Knots’ (EDKs) and climbers are strongly discouraged from using them. In Europe, however, these knots are widely employed by both amateur and professional climbers and so we thought it necessary to examine this paradox.
We will start by reviewing three accidents in which rappel knots were deemed the cause. After this we will define the framework of our research project and differentiate between what we have studied and what we have tested, then we will explain how we carried out the tests. This explanation will be split into two parts, one part concentrating on well-tied knots that have been fully tightened and then another on badly tied (non-tightened) knots. In each part we will analyze the data obtained from the tests and draw conclusions in the form of recommendations.
American publications have reported at least five extremely serious rappelling accidents (most involving life-changing injuries or death) in which the knot joining the ropes came undone. We found information about three of these:
13/09/1998, Grand Teton:
– Offset overhand knot
– Team of three climbers
– One new 10mm rope and one previously used 10.5mm rope
– The overhand knot was tightened by the first climber who then descended the ropes. The two other climbers then ‘moved the knot about 12 inches [30cm] forward’. The next person then proceeded to rappel but after several meters her partner ‘watched the knot “unravel,”’ causing the rappeler to fall.
– Offset figure-8 knot
– Team of two climbers
– Two 60m, 10.5mm ropes
The first climber rappelled about 3 meters before falling to his death. His partner didn’t see the knot used but reported that on the previous rappel “the knot was neat, I don’t remember exactly how long the tails were but they didn’t cause me a second glance.”
12/10/1994, Seneca Rocks
– Team of 2 climbers
The first climber descended 4 meters before stopping to untangle the ropes and then fell to his death. The ropes were found untied. There is no information available about the type of knot used or the type of ropes.
Offset knots for joining ropes: An offset knot joining 2 ropes is one where both free ends come out of the same end of the knot, and which has been tied with both ropes positioned side-by-side (rather than one being fed through a knot in the other). We’ll be studying here the offset overhand knot (figure 3) and the offset figure-eight knot (figure 2).
European death knot or EDK: This is the name given to offset overhand or figure-eight knots in English-speaking climbing communities.
Tightened knot: For these tests we considered that a knot has been tightened when each of the four lengths of rope coming out of it have been pulled tight one at a time. An average person can pull a 9mm rope with a force of approximately 20 DaN (1 daN is equivalent to the weight of a mass of 0.981kg).
End of the rope: The end of the rope, or free end, is the excess rope that comes out of the knot and is not loaded during a rappel.
We deliberately limited our study to the type of cases which had been reported in the climbing press and which had been widely discussed in forums.
– rappels made on dynamic or semi-static ropes of the same diameter.
Rappels made on ropes of significantly different diameters (main rope + tagline) do not work in the same way as those in which a knot is used to join double ropes. Only the main rope, blocked by a knot against a maillon, is loaded for the descent, with the tagline simply being used to retrieve the rope. Our study also does not cover rappels on hyperstatic cord.
– knots in which the free ends were of the same length.
However, we varied the length of the free ends from 7 to 50cm.
– offset overhand and figure-eight knots.
– the first 10 meters of descent.
We did 35 tests using 4 different types of rope:
– Two types of dynamic rope: Béal Gully (7.3mm twin + half rope) and Béal Opera (8.5mm triple-rated rope)
– Two types of semi-static rope: Béal ProCanyon 9.8mm and Béal Aquaram 9.6mm
The following variables were studied:
– the type of offset knot: overhand or figure-eight.
– the degree to which the knot was tightened: well-tightened (with all four lengths pulled hard) or non-tightened. For figure-eight knots on Gully ropes we standardized the non-tightened knot, leaving a 3cm gap in both sides (see first image of figure 9). For the other rope types, whether they were tied with an overhand or figure-eight knot, we simply tied the ropes without tightening the knot. The size and shape of the knots depended on the diameter of the ropes.
– the length of the free ends varied from 7 to 50 cm.
– the weight of the person rappelling was 50 or 75 kg.
– the manner in which the rappeler weighted the ropes. This was done either progressively with the rappeler trying to minimize any shockloading, or dynamically with the rappeler letting themselves drop suddenly onto the rope with a fall of between 20 and 40 cm.
Each test was filmed. The force transmitted to the anchor was measured using a SCAIME 2t device set to a frequency of 500Hz. As two ropes were used for the tests, the force on the anchor was twice that on each length of rope.
How strong are offset knots?
We measured the strength of offset overhand and figure-eight knots when pulled apart slowly. Figure 4 presents the static strength of double ropes that have been joined with these knots. As we are studying rappelling, the traction was measured at the anchor and so was exerted on both ropes together. The results show that the maximum strength of an offset knot when pulled slowly and progressively is easily sufficient for use in rappels, even with small-diameter ropes.
The knots behaved differently depending on the type of rope tested. With dynamic ropes, offset overhand knots roll before separating at around 800 daN. Figure-eight knots flip once under this same force and then break at the value indicated in column 3. These same knots tied in Aquaram (semi-static) ropes don’t roll or flip at all and only break when subjected to very high forces.
How easily can these knots be untied after loading?
After pulling the knots with a force of 300 daN, we had no difficulty untying them. This is not the case with all knots. For example, the Romanian knot was very difficult to untie.
Knots pre-tightened by hand
Figure 7 shows the force exerted on rappel ropes joined by a well-tightened offset figure-eight knot over time.
The initial spike corresponds to the phase when the rappeler initially loads the rappel ropes. This was done by a 75 kg person deliberately falling approximately 30 cm onto their back-up prussik. From 1.3 to 1.5 seconds the force measured on he anchor quickly rose to 250 daN.
The knot, which had initially been well tightened, became even tighter. It did not roll and the free ends did not become any shorter. Three further tests were carried out with the length of the free ends always remaining constant. This suggests that a well-tightened, offset figure-eight knot will not slip during a rappel even when shock-loaded. The same result occurred with offset overhand knots. When loaded, pre-tightened knots become even tighter.
Badly tied knots: non-tightened/short free ends
So how would non-tightened offset knots fare? And is a shock-loaded knot more likely to fail than one which is loaded progressively?
The photos and the graph beneath show how an un-tightened, offset figure-eight knot with 10 cm free ends flips and then fails.
The 75 kg rappeler loaded the ropes as progressively as possible. From 3 to 6 seconds the force progressively increased from 15 to 80 daN. The knot then flipped, consuming almost the entire length of the free ends.
After 14 seconds the rappeler descended quickly before suddenly stopping. The force on the anchor reached 190daN then dropped suddenly to 0 as the knot failed.
Without a back-up rope, the tester would have fallen to the ground.
Photo 11 and graph 12 show how a non-tightened offset figure-eight knot with 10cm long free ends flipped and eventually failed.
The rappeler (50 kg) loaded the ropes as progressively as possible and, from 4 to 9.6 seconds, the load increased from 0 to 52 daN. At this point the knot flipped and came undone after the entire length of the free ends was consumed.
The Aquarem is a very stiff rope, and it is difficult to tighten knots in it. These knots have a tendency to become looser over time, especially when the ropes are new and the smooth surface provides little friction inside the knot.
Knots are particularly unreliable in new and/or stiff ropes.
The photo and graph show how a non-tightened figure-eight knot tied in Opera ropes with 10cm free ends flipped and then held.
The rappeler (50kg) loaded the ropes as carefully as possible, and from 2 to 5 seconds the load on the anchor increased progressively from 0 to 70 daN. The knot then flipped, consuming approximately 7cm of rope.
At 22 seconds the rappeler descended rapidly then braked suddenly, causing the load on the anchor to increase to 160daN.
At 36 seconds the rappeler repeated the rapid descent and sudden brake, causing the load measured on the anchor to increase to 190daN.
After the initial flip, the knot tightened and remained stable for the two subsequent shocks during which the free ends were only 2cm long.
Analysis of non-tightened knots
We conducted 20 tests under the following conditions:
– loading of the rappel carried out progressively
– rappel ropes joined by a non-tightened knot
– rappel knot jammed against the anchor ring
– 10 cm free ends before loading.
The results are shown in the table below.
As we have previously demonstrated, non-tightened rappel knots can roll or flip before tightening under load. These actions shorten the length of the free ends.
Which rope characteristics facilitate rolling/flipping of rappel knots?
The amount of rope consumed when a knot flips depends on several factors. It appears that larger diameter-ropes consume more rope, and this is accentuated when the ropes are stiff. A knot that flips in the ProCanyon, a very supple rope, consumes less of the free ends than with the Aquaram, a very rigid rope. The longest length of rope consumed (10cm) was with a rope that was large diameter, stiff and with a slippery sheath. It would be necessary to confirm these trends with a large number of tests but, nevertheless, it should be noted that in two of the three cases of knot failure recorded in the USA, the ropes were new (slippery sheath) and large diameter. These factors accentuate the risk of failure of a badly tightened rappel knot.
How does the way you load the rappel ropes affect the behavior of the knot?
We tested two ways of loading the rappel ropes:
1) dynamic loading, which involved the tester falling 20 to 40cm onto their back-up prussik.
2) progressive loading which involved the tester easing onto their back-up prussik or rappel device as carefully as possible.
None of the dynamic loading tests resulted in knot failure. This was because the knot tightened instantly and didn’t have time to flip. In order for the knot to flip, the ropes must be loaded progressively as seen on diagram 16.
Jerk is the time derivative of acceleration. It equates to the shock felt by the tester. The higher the jerk, the greater the shock. In the diagram we can see two distinct groups with regards to jerk and rope consumed. The highest jerk factors (around 40ms-3) were obtained by the tester falling onto the rappel ropes.
The shock depends upon the length of the fall and the dynamic properties of the ropes. Semi-static ropes will not stretch and absorb as much energy as dynamic ropes and the jerk factor will be greater. The lowest jerk factors (below 10ms-3) were obtained by progressive loading of the rappel ropes.
The horizontal axis denotes the decrease in the length of the free ends expressed as a percentage of their original length. On the diagram you can see:
0% – the black diamond represents a knot in which the length of free ends did not decrease at all. This knot had been pre-tightened.
50% (approx) – knots in which the length of the free ends was halved (from 10 to 5cm).
80% – only 2cm of free ends remained due to the knot flipping and then tightening under theweight of the tester.
100% – the entire length of the free ends was consumed, the ropes separated, and the tester fell (red line).
Obviously the shorter the free ends, the greater the likelihood of an accident if the knot has not been sufficiently pre-tightened, irrespective of the type of knot. It is absolutely essential to always pre-tighten a rope in all circumstances.
When do knots flip?
Our tests show that knots flip:
1) Either when a pre-tightened knot is loaded progressively with a force of more than
2) Or when a non-tightened knot is loaded progressively with a force of less than 120daN
It seems very unlikely that a climber would find themselves in the first situation. Loading the rope with a force in excess of 800kg could only happen during a sudden, catastrophic event such as an avalanche or rockfall and in this case the strength of the knot will probably not be the most critical problem. The second situation, however, is much more likely because climbers often start loading their rappel ropes cautiously, aware of the risks involved. However, non-tightened knots roll when they are progressively loaded, and this can lead to catastrophic failure of the knot.
Analysis of pre-tightened knots
As we saw, a knot pre-tightened by hand will tighten further when loaded, especially during shock-loading. Given that tests demonstrate that progressively loaded knots will hold between 900 and 2400 daN (depending on the type of knot and rope), the margin of safety remains considerable. For classic rappelling situations (one person descending at a time = maximum load less than 300kg), the use of offset knots does not present a risk.
Thirty-two tests were conducted on non-tightened knots linking rappel ropes. Five resulted in the failure of the knot. We managed to ‘simulate’ an accident no matter the tester or the type of rope employed (Gully/Opera/Aquaram).
In order to achieve this, the following conditions had to be met:
– non-tightened rappel knot
– free ends less than 10 cm long
– extremely progressive loading of the ropes.
This condition is very important as it is necessary in order for the rappel knot to flip.
A badly tied knot (loose, or poorly tightened) with short free ends presents a high risk of failure. When tying any knot, it is important to respect the following recommendations:
• All knots should be vigorously tightened; each length of rope should be pulled hard up against the others. The free ends should be at least 30 cm long.
• If tied in this way, offset overhand and figure-eight knots are perfectly suitable for joining rappel ropes when using normal rappelling techniques.
Batoux P. Utilisation de la corde en alpinisme. Publication de l’ENSA. 2021.
Batoux P. Noeuds de jonction pour les cordelettes. Publication de l’ENSA. 2018.
Brass P. Bien choisir son noeud pour rabouter 2 cordes. Montagnes Magazine. Novembre 2017.
Siacci R. Esq., In-defense-of-the-european-death-knot, Climbing Magazine. Octobre 2016
Euro-death-knot flat figure 8 mysteriously fails. Rock and Ice issue 233 (April 2016).