With the rapid advance of new technology more and more rowers are enabled to train with real-time feedback on mechanical power output during on-water training sessions. The most- known company that provides a system with real-time power output feedback is Nielsen Kellerman (NK; i.e. the EmPower oarlock).

The advantages of training with power output are clear. In contrast to commonly used intensity measures such as stroke rate, boat velocity, and heart rate, mechanical power output is a much more objective measure of training intensity. First, it is strongly related to a rower’s metabolic energy consumption (Hofmijster et al, 2009). Second, it is – in contrast to the commonly used parameters- not influenced by external factors such as wind, temperature and mental state of a rower. Besides, accurate on-water power output measures allow for comparisons of on-water and ergometer training sessions.

When feedback on power output is provided it is highly necessary that the feedback is accurate. Erroneous values of power output may have severe consequences. Rowers who comply with inaccurate power output feedback most certainly train at undesired training intensities and may risk under or overtraining. Unfortunately, most researchers and companies use an algorithm for power output calculations that result in incorrect values of power output.

The most commonly used method to determine power output in rowing is to multiply the moment around the oar with the oar angular velocity. A simple thought experiment (from Hofmijster et al, 2018) shows that this method must be incorrect. Consider a rower in a single scull on the water without oars, with her/his feet attached to the foot stretcher. Indeed, most rowers will fall in the water but imagine that this single scull is very stable and the rower has an incredible sense of balance. Now the rower is asked to move in a periodic motion back and forward relative to the boat. As a consequence, the boat will move periodically back and forward as well. As the boat movements are steady-state (i.e. the velocity of the boat at the end of a cycle will be equal to the velocity of the boat at the beginning of the cycle), no kinetic energy will be dissipated over a whole cycle.  However, during the cycle, the boat movements result in a frictional force of the water on the boat that is opposite to the velocity of the boat. This means that power is required to move the boat. Since the rower is the only ‘motor’ in the boat, the rower must have delivered that power. Though, power output values calculated using the commonly used method will be zero in this example, since the rower does not apply moments around the pin. Using Newton’s laws it can also be mechanically proved that the commonly used method to calculate power output is incorrect (see Hofmijster et al 2018 for a detailed overview).

This thought experiment demonstrates the flaw in the commonly used method: rowers deliver power output via their hands and feet, while the commonly used method only determines the power dissipated via the hands. This flaw has been identified by Kleshnev (2002) but has never been widely recognized by the scientific and practical rowing community. In an experiment, he derived true power output values by measuring forces at the foot stretcher and the oar. However, since these forces are high and opposing, small insurmountable measurement errors may add up and result in bigger miscalculations of power output. Besides, measuring forces on the foot stretcher may not be that trivial in daily practice.

So in order to determine accurate ‘true’ power output values a valid and more practical method is desired. Therefore, we came up with an alternative method to obtain true power output values in which forces at the foot stretcher do not need to be measured directly. This new valid method differs from the commonly used method with an amount that is equal to the product of the mass of the rower, the rower’s acceleration of the center of mass (CoM) and the boat velocity (see the bottom of the blog for a detailed overview of the calculation).

What is a rower’s center of mass?

Every object consists of a mass. This mass is distributed in space. The center of mass is the unique point where the weighted relative position of the distributed mass sums to zero. In other words the distributed mass is balanced around the center of mass. In rowing this point is most probably located somewhere closely in front of the rower’s trunk.

To provide an indication of the miscalculations of the true power output when the commonly used method is used, we quantified the difference between power output values obtained using the commonly used method and the true power output values (see Lintmeijer et al, 2018). Nine rowers with different masses rowed 2 times 250 meter under different rowing conditions (see Table 1). During the trials, the forces at the oarlock, the oar angular velocity, the boat velocity, and the rower’s CoM acceleration were measured. In hindsight, delivered power output values were calculated using the commonly used method and our valid method. Most importantly, we found that true power output values were, on average, underestimated with 12.3% when the commonly used method was used. A small difference between stroke rate conditions was detected: the higher the stroke rate, the more the true power output values were underestimated (see Table 1).

These results emphasize that the commonly used method seriously underestimates a rower’s true power output. Additionally, power output is slightly more underestimated when stroke rate increases. Therefore, we would strongly advise companies such as NK, but also a researcher, to correct their algorithms to determine power output values. The easiest and most practical way is to correct average power output values determined with the commonly used method by the average underestimation (which is equal to 1.14 = 1/1-0,123). When more accurate power output values are desired for example in situations where selection between rowers is based on power output values, we advise determining power output values by measuring the mass of the rower, the acceleration of the CoM and the boat velocity. We would like to recommend coaches and rowers to be critical to the presented power output values. Please be sure, that the correct algorithms have been used.


Hofmijster, M.J., Lintmeijer, L.L., Beek, P.J., van Soest, A.J (2018). Mechanical power output in rowing should not be determined from oar forces and oar motion alone. Journal of Sports Science, 36:18, 2147-2153.

Hofmijster, M.J., van Soest, A.J., de Koning, J.J. (2009). Gross efficiency during rowing is not affected by stroke rate. Medicine and Science in Sports and Exercise, 41:5, 1088-1095.

Kleshnev, V. (2002). This is a chapter. In Y. Hong (Ed.), International research in sports biomechanics: Power in rowing (pp. 224–230). Abingdon: Routledge.

Lintmeijer, L.L., Hofmijster, M.J., Schulte Fischedick, G.A., Zijlstra, P.J., van Soest, A.J. (2018). Improved determination of mechanical power output in in-water rowing: experimental results. Journal of Sports Science, 36:18, 2138-2146.

Table 1: Mean (M) and standard deviations (SD) of the power output values determined using the commonly used method (P_estimated) per rowing condition and the associated differences from the average true power output values (P_true).


    P_estimated in Watt Difference from P_true in Watt Underestimation of P_true
  n M SD M SD
Stroke rate (SR) 18 9 198 48 -23 5 12.0
SR 25 9 280 66 -35 9 12.6
SR 32 8 339 81 -53 9 13.3
Early knee extension (SR 18) 9 178 44 -21 4 11.7
Early trunk extension (SR 18) 9 183 64 -27 8 12.7
Low intensity (SR 18) 9 121 36 -15 5 12.1
High intensity (SR 18) 9 235 58 -28 5 11.9
Overall 62 225  99 28 13 12.3


For those who want to see the maths: