2023 Skierg Sprint Analysis and the Importance of Weight

[LeManchot]

The recent Skierg Sprint men's dataset allows for an analysis of the role of weight-normalized power (W/kg) in the competition. I took the available data from the results and calculated W/kg values for those athletes that supplied a weight in their profile. Of the top 100 results, 20 athletes provided their weights. I also did this analysis for the 60-69 age group and the 70-79 age group, as those were ones I had specific interest in. The results are interesting and provide insight as to the accuracy of the Concept 2 in calculation of “power” (and the derived “distance”) on the Skierg as well as the importance of including weight-normalized power (W/kg) in the rankings. As in cycling, performance on a Sklierg (or C2 rowers or C2 bikes) comes down to W/kg. In fact it is arguable that any competition should be time-based with a dependent variable of W/kg as the performance metric, e.g. 3 min TT with the performance metric being average W/kg for the TT. This protocol would be a much more accurate assessment of ability and competitiveness.

Below is the produced dataset from the information provided in the 2023 results. Note: the weights are provided by individual athletes and are not verified. All other data have been verified by Concept 2 and have allowed the W/kg calculations. the forum software does not allow for formatting of the data chart but it is properly organized. The forum software does not allow for imbedded graphics so a link to the graphics is provided.

2023 SkiErg Sprints
athlete age rank time avg power weight W/kg
60-69
Jan Rune Gilde 60 2 03:18.8 356 183 4.29
Gavin Grant 61 4 03:25.8 321 195 3.63
Kev Lovett 62 7 03:33.5 288 212 3.00
Timo Friman 62 10 03:38.6 268 190 3.11
Johan Denekamp 64 11 03:38.7 268 216 2.74
Martin Luirink 64 15 03:41.7 257 182 3.11
Heinz Zippel 64 28 03:54.8 216 211 2.26
Michael Romero 60 29 03:55.2 215 189 2.51
Hannes Botes 64 30 03:55.6 214 212 2.23
Mike Crenshaw 69 42 04:04.8 191 155 2.72
Leo Cunningham 63 58 04:17.6 164 155 2.33

70-79
Phillip de Bruyn 72 1 03:31.0 298 190 3.46
Julian Kennedy. 71 2 03:59.9 203 165 2.71
James Slipetz 70 9 04:10.6 178 175 2.24

overall - top 100
Dan Lake 40 5 02:56.7 508 201 5.57
Joseph SEFFA Silafau 38 13 03:02.6 460 271 3.74
mitchell parkes 34 15 03:03.3 455 243 4.13
Christian Coe 38 16 03:03.5 453 220 4.54
Mark Snape 41 18 03:04.1 449 266 3.72
Freddy Banales 32 27 03:08.5 418 185 4.98
Johan Nilsson 39 30 03:09.0 415 320 2.86
Matt Blake 40 32 03:09.1 414 259 3.52
Brendan Pearson 26 33 03:09.4 412 174 5.22
Jan Egil Andresen 45 34 03:09.8 410 205 4.41
Jacque Peet 29 35 03:10.1 408 174 5.17
Kevin Carter 45 52 03:14.1 383 240 3.52
Denny Locascio 39 57 03:14.9 378 186 4.48
Yevgueniy IZBASH 40 60 03:15.5 375 187 4.42
Robert Cote 57 78 03:17.6 363 195 4.10
Ron Peterson 38 80 03:18.2 360 225 3.53
Jan Rune Gilde 60 86 03:18.8 356 183 4.29
Freid Eggum 49 86 03:18.8 356 214 3.67
Jordy van Langen 28 89 03:19.7 352 176 4.41
Leopold Schwarzer 18 94 03:20.5 347 159 4.81
Henry Strieker 56 95 03:20.6 347 209 3.66


From these data a number of relationships can be developed:

1. The “time for 1000 m vs average power” graph (attachment 1 - https://imgur.com/a/Ro7ijuv) shows a near-perfect logarithmic functionality (fit equation is on graph). This is expected given that power is directly related to time, i.e. Power = Work per unit Time or, alternatively, Power = Force times Velocity. The PM5 uses an angular velocity measurement to compute power and the near-perfect fit of the time data to a logarithmic function with this random selection of individuals indicates that the precision of the C2 Skierg is quite good. The question is how accurate is the Skierg. A recent peer-reviewed study (https://www.frontiersin.org/articles/10 ... 01617/full) of the C2 rower (which uses the same mechanical system as the Skierg) found that the accuracy is 2.9-4.3% with steady application of force and a highly variable result of 10-70% accuracy in the first 5 stokes. Here is what the authors say in their 2022 study:

“According to Van Holst (2014), the actual mechanical power output per rowing cycle, usually the key measure of performance (Soper and Hume, 2004), is calculated by the display-computer of the C2, based on measurements of angular velocity (which is used to calculate acceleration and deceleration of the flywheel), the mass of the flywheel, and a constant factor. This approach differs substantially from the physical definition of mechanical power as work per time. Considering the special calculation and the fact that the C2 is employed worldwide for performance measurements of rowers, it is surprising that there is limited and incomplete information about the quality criteria and particularly validity of the C2.”

Based on the winning time in the 2023 Skierg Sprints, the accuracy of the Skierg (from the study above) would indicate an error of about 7 seconds which makes the top five results from the 2023 Skierg Sprints indistinguishable.

Note: the Skierg is likely to exhibit decreased accuracy due to the significantly less rigid and efficient drivetrain compared to the C2 rower given the degrees of freedom that the hand position can take and the radically different spectrum pull techniques that are used.

2. The “W/kg vs average power for the 1000 m time trail (TT)” (attachment 2 - https://imgur.com/a/Ro7ijuv) is highly variable, particularly at high power values. This hints at increasing inaccuracy at high power output, again questioning the accuracy of the calculated power values. Aside from the deviations at high power, the data do follow the expected increasing trend with average power. A linear fit of the data is shown on the graph although a logarithmic fit has a similar r^2 (about 0.64). There are at least a few sources for the variability:

A. The power calculation derived from angular velocity is not robust across the expected power range: All bicycle power meters have a calibration feature that, ideally, is done prior to any ride. (There are some issues in the industry however - Shimano appear to have significant accuracy problems with their crank-based power meter (https://www.youtube.com/watch?v=R2v7bLiZB90).) Cycling power meters are strain guage-based and make direct measurements of torque. Power in cycling is equal to angular torque times angular velocity (derived from measured cadence). These devices exhibit an accuracy of 1-2% depending manufacturer and include temperature compensation given the sensitivity of the strain gauge measurements to temperature deviations during a ride. I have not seen any statement from Concept 2 about the accuracy of their power numbers, although that may exist.

B. Drivetrain stability: Unlike a bicycle (which has an extensive database for power measurements due to the millions of power meters currently being used on bicycles) that has a rigid and highly efficient drivetrain, the Skierg is considerably less rigid and efficient. Therefore there could be differences from one unit to another. In addition, there is no power calibration on the Skierg to ensure that values being produced are valid. Concept 2 may have a way of internally calibrating that is not visible (or made available) to the user. Although I haven’t done an exhaustive search, I have not seen anyone post about any issues with their power numbers in this forum.

C. Technique/Efficiency: Given the many “techniques” that are recommended by users on the internet and Youtube, it is clear that many of these techniques are not only unfavorable for proper skiing, they are also substantively inefficient from a biomechanics/power delivery perspective. In some cases, a “technique” could lead to higher power numbers but this is produced in an inefficient way and the athlete will prematurely fatigue and not achieve their optimum time for the 1000 m TT.

D. Inaccurate reporting of weight: As in any unvalidated, participant-supplied variable, accuracy can be highly suspect.


3. The “time to 1000 m vs W/kg” (attachment 3 - https://imgur.com/a/Ro7ijuv) shows the expected asymptotic function. Based on these data, the limiting value in time to 1000 m is about 2:40 (min:s) which is consistent with the current accepted Skierg World Record for 1000 m of 2:40.6 (min:s) and with expected human physiology. Since the population here is a random selection and therefore includes a wide range of fitness and experience levels, it is appropriate to look at the top W/kg values as representing a group that are fit and experienced in “Skierging”. Looking at the top five W/kg results from the dataset that represent the 5th (201 lbs), 27th (185 lbs), 33rd (174 lbs), 34th (174 lbs), and 94th (159 lbs) best times, a clear trend of decreasing time to 1000 m with increasing weight is observed. Although much more data is necessary to establish the validity and magnitude of any effect, it would appear that weight is a controlling variable in time to 1000 m. It should be noted that the population represented in the Skierg Sprints is likely skewed to heavier athletes since, apparently, lower weight athletes, by and large, do not participate. The lightest participant in this population is 155 lbs. There may be lighter participants in the population but they may not have reported their weight. As in any endurance sport, lighter athletes tend to produce better results due to a high power to weight ratio, even on flat terrain (as is simulated with the Skierg). The differences become even greater when variable (hilly) terrain is included. Most endurance sports (and particularly in cross country skiing) tend to produce world class athletes with remarkably similar heights and weights across the most competitive population with a few outliers, mostly on the lower weight side. The Skierg Sprints, as currently structured, appear to favor heavier participants.

4. The “time to 1000 m vs weight” (attachment 4 - https://imgur.com/a/Ro7ijuv) confirms the observation in #3. Here there is clear evidence that weight is playing a controlling role and may be exhibiting a power-law functionality (shown in the fit). Again much more data are needed to establish a robust basis for any functionality.

Conclusion

As mentioned above, given the controlling nature of weight in the “time to 1000 m” on the Skierg, an alternative more representative (and more accurate) metric for competitions would be average W/kg over a given time on the Skierg. This is straightforwardly calculated from the average power for the TT and the weight of the competitor. The only issue for implementation would be verification of reported weights.

[LeManchot]

This popped up on YouTube over the weekend (https://www.youtube.com/shorts/1yOx-1o8xFA) showing Johannes Klaebo (arguably the best cross country skier ever (at least in the sprints)) on the Skierg doing 5.62 W/kg for 20 seconds and 7.7 W/kg for 20 seconds on snow. Not sure how they got the watts for the on-snow part, but there are some power measuring poles on the market so that may be a possibility. Note: the 5th place finisher in the 2023 Skierg sprints did 5.71 W/kg for just under 3 minutes, however, at 201 lbs vs Klaebo's reported 165-170 lbs, there would be no contest on snow.

If these numbers are accurate then it shows exactly how much more power one can get with proper poling technique on snow. What Klaebo does in that double pole sprint is not possible on the Skierg as currently designed.

[LeManchot]

Sorry, misread the Skierg numbers as W/kg when they are W... the 562 W for 20 seconds translates to 7.7 W/kg at 160 lbs (72.7 kg)... I wondered about any power measurement on snow but I would expect the value to be significantly higher than on the Skierg given the ability of the body to get into an optimal position for maximal power. That's also my experience - one is much more powerful on snow than on the Skierg.

I wonder what the highest 20 second W/kg value is for the Skierg?

[LeManchot]

I took a closer look at the "time to 1000m" vs avg. power and found a perfect fit (R² = 1) for an exponential function with an exponent of -0.333. The logarithmic fit is close but the exponential function is perfect. Perfect doesn't just happen and is the result of a physical law being followed.

Looking further to the physics of ergonmeters at (http://eodg.atm.ox.ac.uk/user/dudhia/ro ... meter.html), this relationship is expected. The specific relationship for the analyzed SkiErg Sprints data is:

y = (0.0163) (x^-0.333) or

"time to 1000m" = (a constant) ("avg. power"^-0.333)