Dynamic Recovery: A Rower's Secret Weapon

Motivation

The goal of rowing is to achieve maximum efficiency of your strength. This is done by rigging or adjusting oars and riggers to your specifications and by having perfect technique. Some coaches prefer having slower hands away from the body over faster hands after the pull or drive of the oar through the water and on the way up to lower the oars into the water at the “catch”. Other coaches prefer the opposite. I want to know what will give me maximum efficiency. Is there truly a difference between dynamic and traditional recovery and is there a better way to measure it?

This project uses a measurement of the motion of the seat relative to the boat combined with the available data of the (all-important!) motion of the boat relative to the water to compare the two recovery techniques and to establish a reliable way to measure the efficiency of the recovery part of the stroke cycle.

The results are spectacular: dynamic recovery with its active use of the muscles not normally engaged during traditional recovery can provide a brief 7% increase in boat speed at each stroke. Over a full race, this provides a huge competitive advantage.

Elements of a rowing stroke

Getting a boat to move as fast as possible involves a combination of strength training and of the rowing technique development.  Every little thing counts, and even a very small increase in efficiency multiplied by a large number of strokes the rowers complete during a race produces a significant improvement in performance. Much of training goes into improving the drive part of the stroke cycle, when the extending legs move the athlete’s body forward (in the boat travel direction) and the oars are in contact with the water, pushing the water backward and the boat forward.

The recovery part of the stroke is when the athlete’s body moves back (opposite to the direction of the boat travel) and the oars are out of the water. This motion first accelerates the boat (the body accelerates toward the stern and the boat receives a reciprocal pull forward), and then decelerates the boat. The traditional recovery advocates for a relatively slow and even motion, letting the boat to “run under the rower” and avoiding “slamming” into the footholds at the end of recovery just before the catch. The dynamic recovery advocates for an active initiation of recovery using the quad muscles for a rapid flexion of the hips and knees of the rower. This gives the boat a stronger forward pull at the beginning of recovery, but is it cancelled out by the rapid backward push at the end of recovery as conservation of momentum requires?

The seat motion measurement

The seat movement gives precise information about the position of the rower’s body as a function of time, but to my knowledge has never been measured directly. I designed an optical detector of the seat movement: two infrared reflection detectors a fixed distance apart attached to the seat are able to detect the timing of transitions between reflective and non-reflective pattern of strips attached to the floor of the boat. The sensor package uses an ESP32 microcontroller to measure the timing of the transitions and wirelessly send the data to a RowPal device, a novel portable rowing computer (RowingPrecision.com) that simultaneously records this data as well as the high-resolution GPS coordinates of the boat and its speed, measured every 100ms. The device worked extremely well on the test bench but water splashing in the boat rendered the infrared measurements unreliable. Fortunately, every run was recorded on video from the coach boat running parallel with the rower at matching speeds. This was intended as a backup measurement but became the primary source of the usable data shown here.

The lesson: always have a Plan B!

Data analysis

Frame-by-frame analysis of the video; rescale the values from pixels to meters using a known bow-to-RowPal fixed distance;

Download and parse .json files of the data from RowPal (GPS-measured speed of the boat);

Convert both data sets to a common absolute time base, to match the data stroke-by-stroke;

Chop the data set into individual strokes, keeping only those strokes for which both data sets are available;

Allow for a single time shift of the video time base (the start time of the video is only known to 1s precision) to align the data based on the characteristic features of the stroke curves;

Plot all stroke curves superimposed, slight variations between strokes results in a band of curves instead of a single one.

To understand the resulting plots, consider an example of a comparison between traditional (in red) and dynamic (in blue) speed curves of Karolien Florijn recorded during her 2024 Olympic final, as reported by Mike Purcer (reproduced from purceverence.ca). Traditional recovery resulted in the time of 1:55.15 at 33.0 spm; the dynamic recovery improved it to 1:53.64 at 36.4 spm.

The seat sensor data obtained in this project allows us to understand how such an improvement is achieved, aiding athletes and coaches alike.

Results: dynamic vs. traditional recovery

The bands of curves represent superposition of data from multiple strokes. The bands are narrow, indicating consistent performance by the athlete. The downward slope of the blue curves represents the speed with which the seat movement during recovery is initiated. This slope is steeper, and the duration is shorter, for dynamic recovery (graphs 2 and 4) vs. traditional recovery (graphs 1 and 3), both with and against the current. Although the effect is subtle, a significant boat speed boost (the green so-called "speed curves") is seen during recovery. This boost is much more pronounced during dynamic recovery without a decrease in speed at the end of recovery. Conservation of momentum is not violated as an early catch breaks the symmetry of the initial and final part of the recovery.

Conclusion: The results demonstrate a spectacular improvement in the contribution of recovery to the boat speed, from 0.25/4.25 m/s (6%) for traditional to 0.5/3.75 m/s (13%) for dynamic recovery! Dynamic recovery with its active use of the muscles not engaged during traditional recovery can provide a brief 7% increase in boat speed at each stroke.

This result could only be achieved by a rower who was specifically trained for dynamic recovery: a similar measurement for an athlete attempting dynamic recovery for the first time yielded very similar speed curves for both dynamic and traditional recovery (not shown).

Conclusion: Using dynamic recovery requires a change in the training regimen.

Sensor redesign and future work

The first-generation sensor was redesigned to use magnetic detectors and a sealed magnetic track which keeps itself in place and allows easy repositioning in a wet environment. Incommensurate sensor placement provides improved velocity and direction detection.

Late-season testing was limited. Additional athletes of various levels of rowing experience need to be trained and tested for dynamic recovery.

A bigger picture: muscle oxygenation sensors for simultaneous measure of energy use; oar sensors for separating centre-of-mass and upper-body movement; comprehensive training guidelines; live monitoring.

A few % gain on every stroke has the potential to yield elite-level results.

Acknowledgments

A big thank you to: Rowing Precision (rowingprecision.com) for hosting my Co-Op term and for the use of their novel RPX-270 (but I call it “RowPal” because it is!) device; to Mike Purcer for generously sharing his extensive knowledge of rowing; to the Physics Department, Brock University, for technical assistance; and to my father, Dr. Edward Sternin, who helped with data analysis and piloted the safety boat during the test runs.

References

Tracker. opensourcephysics.github.io/tracker-website/. Accessed November, 2025.

Wilmington Youth Rowing Association. https://www.wyra.org/player-resources/diagrams-of-the-sport-of-rowing/22013. Accessed March, 2026.

Espressif. Esp32 series datasheet. Technical report, espressif.com, 2025.

Guide to using Tracker. www.physics.brocku.ca/Courses/1P91/lab-manual/Resources/Collisions_Tracker/collisions.php. Accessed November, 2025.

The Brock News. New tech startup drives forward with VPMI support. brocku.ca/brock-news/2025/07/new-tech-startup-drives-forward-with-vpmi-support/. Accessed September, 2025.

Michael Purcer. Traditional vs. dynamic recovery. purcerverance.ca/my-blog/f/traditional-v-dynamic-recovery. Accessed March, 2026