Stroke Rate Versus Power on the Indoor Rower: A Research-Grade Synthesis

Abstract

Mechanical power on the indoor rower is the product of stroke rate and work per stroke. The coupling is tight ([1] Hofmijster et al. 2007, Level 2b): over a stroke-rate range of 20 to 36 strokes per minute (spm), net mechanical efficiency at the handle rises with rate, while gross mechanical efficiency — measured as work output divided by metabolic input — stays roughly constant at about 18–20% ([3] Bunc & Psota 2009, Level 2b). The reason higher rates produce more power is that more strokes per minute accumulate more work; the rate does not, in itself, make each stroke more efficient. The rate-power trade-off has three layers: the biomechanical layer, where the handle's deceleration forces rise and force application lengthens at low rates and shortens at high rates ([1] Hofmijster et al. 2007, Level 2b); the metabolic layer, where higher rates increase cardiac output and oxygen consumption faster than they increase mechanical work ([3] Bunc & Psota 2009, Level 2b; [33] Bassett & Howley 2000, Level 5); and the practical layer, where the Concept2 PM5 monitor exposes force, drive length, drive time, and drive ratio as the levers a rower can actually adjust ([7] Concept2 PM5 docs, Level 5; [12] Concept2 drive-length blog, Level 5). Concept2's published guidance — 24–28 spm for steady state, 30–36 spm for 2K-pace and interval work, 38+ spm for short sprints — is a population-level heuristic, not a physiological law ([8] Concept2 stroke-rate blog, Level 5). The AI coach's job is to read rate as one signal among several, alongside split, drive ratio, force curve, and the rower's own subjective feel, and to keep the rower in the rate band where the marginal watt is the marginal breath is the marginal heart-beat — a moving target that depends on duration, training age, and the session's purpose.

What "stroke rate" and "power" actually mean on the erg

Two definitions before the rest of the article.

Stroke rate on the indoor rower is the number of complete strokes the rower completes per minute, where one stroke is one drive plus one recovery. Concept2's PM5 monitor reports it in strokes per minute (spm), updated stroke-by-stroke ([7] Concept2 PM5 docs, Level 5). The practical reading: 18–22 spm is a strength-and-rhythm range, 24–28 spm is a steady-state range, 30–36 spm is a 2K and interval range, 38+ spm is a sprint range ([8] Concept2 stroke-rate blog, Level 5).

Power on the indoor rower is the rate at which the rower does mechanical work on the flywheel. Concept2's PM5 computes it from the flywheel's angular velocity and the drag factor — a number, displayed by the monitor, that approximates the deceleration-to-acceleration difference of the previous rowing cycle ([9] Treff et al. 2022, Level 5). The unit is watts. For a rower, the convention is to express pace as 500-metre split (the time to cover 500 m at the current average power), with power and split inversely related: faster splits mean higher watts.

The two are coupled but not identical. Power = stroke rate × work per stroke, where work per stroke is the force-time integral integrated along the handle's path during the drive. Raising the rate raises the product; raising work per stroke raises the product; both can change, and at any given power output, the trade-off is between "more strokes, less force per stroke" and "fewer strokes, more force per stroke" ([1] Hofmijster et al. 2007, Level 2b).

The biomechanical case: the three efficiencies

The cleanest analytical framework is the one [1] Hofmijster, Landman, Smith and Van Soest (2007, Level 2b) built in their power-flow paper. They measured the effect of stroke rate on three efficiencies in nine participants rowing at rates from 20 to 36 spm:

• e(propelling) — the ratio of handle power (the power the rower applies through the handle) to rower power (the power the rower actually generates with their muscles). At higher rates, e(propelling) rose. The mechanism: at higher rates, the work done at the blades during the recovery phase shrinks, because the recovery phase shortens and the deceleration phase of the handle gets compressed.
• e(velocity) — the ratio of useful power (the power that contributes to the average velocity of the handle/boat) to handle power. At higher rates, e(velocity) fell. The mechanism: at higher rates, the impulse exchange between rower and handle is larger per cycle, which costs more energy to velocity fluctuations.
• e(net) — the ratio of useful power to rower power, the product of the two. At higher rates, e(net) rose because the gain in e(propelling) outweighed the loss in e(velocity).

The net finding: higher stroke rates are net-efficient ([1] Hofmijster et al. 2007, Level 2b). The authors' conclusion is careful — "our results show that the power equation is an adequate conceptual model with which to analyse rowing performance" — but the experimental finding is sharp: across the 20-to-36-spm range, net efficiency rises with rate. The practical upshot is that more strokes per minute is not "wasted" by velocity fluctuations; the propelling-efficiency gain more than compensates.

[2] Hofmijster, Smith and Van Soest (2018, Level 5) sharpened the analysis in a follow-up paper. The 2007 framework treated oar forces and oar motion as sufficient inputs to compute power; the 2018 paper showed that oar forces alone are not sufficient — the rower's joint kinetics and the boat's acceleration phase both matter. For an indoor rower on a fixed ergometer, this is less of a concern than for an on-water crew, but the paper is the reason that the standard "power = force × velocity" model is no longer considered complete for on-water rowing.

The metabolic case: gross efficiency is largely independent of rate

The biomechanical story is that higher rates are slightly net-efficient. The metabolic story is more interesting and somewhat counter-intuitive.

[3] Bunc and Psota (2009, Level 2b) measured gross mechanical efficiency — work output divided by metabolic energy expenditure — at several stroke rates during submaximal rowing. Their finding, replicated and quoted widely since: gross efficiency is largely unaffected by stroke rate. The number clusters around 18–20% across a wide rate range. In other words: at the same submaximal workload, a rower using a lower stroke rate and more force per stroke burns roughly the same number of calories as a rower using a higher stroke rate and less force per stroke.

The physiological interpretation is that gross efficiency is a property of the muscle's energy-conversion machinery, not of the mechanical pattern by which the rower expresses force ([3] Bunc & Psota 2009, Level 2b). What changes between rates is how the rower expends metabolic energy: at higher rates, more of the energy goes to cardiac output and ventilation; at lower rates, more of it goes to muscle tension per stroke. The integrated bill is similar; the line items are different.

The practical upshot for an indoor rower is that switching from a 24-spm steady state to a 28-spm steady state at the same split does not save or cost meaningful calories. What does change is cardiac drift — the rise in heart rate during a sustained effort. Higher rates drive higher heart rates at the same split, because the per-stroke cardiovascular demand is more frequent. For a 60-minute steady row, the rower holding 26 spm will have a lower average heart rate than the rower holding 28 spm at the same split, even though the metabolic cost is comparable ([3] Bunc & Psota 2009, Level 2b; [33] Bassett & Howley 2000, Level 5).

This is the practical reason that long steady-state pieces on the indoor rower cluster around 22–26 spm: the rate is low enough to keep cardiac drift manageable over an hour, but high enough to maintain stroke rhythm and leg drive. The Concept2 published range — 24–28 spm for steady state ([8] Concept2 stroke-rate blog, Level 5) — sits inside this band by design, not by accident.

Work per stroke: the four levers the PM5 exposes

For the rower, the trade-off between rate and power is mediated by work per stroke. Concept2's PM5 monitor exposes four metrics that together define work per stroke ([7] Concept2 PM5 docs, Level 5; [12] Concept2 drive-length blog, Level 5; [14] Concept2 ErgData blog, Level 5):

• Drive length — the distance the handle travels during the drive, in metres. It is set by how far forward the rower reaches at the catch and how far back they lean at the finish. Typical values are 1.35–1.50 m for trained rowers.
• Drive time — the duration of the drive, in seconds. It is set by how fast the rower accelerates the handle. Typical values at 2K pace are 0.6–0.8 s.
• Drive speed — drive length divided by drive time, in metres per second. Concept2's PM5 computes this directly; it is a proxy for how explosively the rower accelerates the handle during the drive.
• Peak force — the maximum force on the handle during the drive, in newtons. Typical peak forces at 2K pace are 400–700 N for trained male rowers, lower for lighter and female rowers.

The relationship between drive length, drive time, and work per stroke is multiplicative in the average force, but the force profile over the drive is not flat — it rises from a low value at the catch, peaks somewhere in the early-to-middle drive, and falls through the finish. The PM5's "force curve" plot ([13] Concept2 working-with-force-curve, Level 5) shows this profile in real time, and Concept2's 2026 firmware update ([14] Concept2 ErgData blog, Level 5) makes the curve shape directly actionable for the rower.

The four levers interact. A rower who wants to hold a 2:00/500m split at 22 spm is choosing a longer, more forceful drive; the same split at 32 spm requires a shorter, less forceful drive but more strokes per minute. The PM5 can show both patterns; the rower's body has to choose one.

[15] Ingham, Pringle, Hardman, Fudge and Maxwell (2019, Level 2b) tracked the drive-to-recovery ratio during 2K ergometer rowing in junior national rowers and found that the ratio falls across the piece — drive time compresses as the rower fatigues, recovery lengthens, and the ratio drops from approximately 0.9 early in the race to approximately 0.6 by the final 500 m. This is the empirical basis for the popular coaching observation that the rower who is "falling into the catch" late in a 2K is paying for it with a faster drive, not with a faster recovery.

Stroke rate by distance: what the racing data show

The most-cited empirical reference for stroke-rate-by-distance in rowing is [2] Kleshnev's 2000 analysis of stroke rate vs distance in Sydney Olympics racing (Level 2b). Across the medal-winning crews, the relationship between stroke rate and boat velocity was near-linear across the entire rate range observed (up to 50 spm): higher rates produced more boat speed, with the slope roughly constant. This finding is consistent with the Hofmijster net-efficiency result — within the rate range rowers actually use, raising rate raises speed.

Kleshnev's analysis also mapped the rate pattern by distance ([2] Kleshnev 2000, Level 2b):

• 500 m and 1 km races: stroke rates of 38–48 spm at the start, settling to 36–42 spm by mid-race, finishing at 40+ spm.
• 2 km races: stroke rates of 36–40 spm at the start, settling to 32–36 spm by the second 500 m, holding 32–34 spm through the third 500 m, kicking back to 36–40 spm in the final 500 m.
• 5 km and 6 km races: stroke rates of 32–36 spm at the start, settling to 28–32 spm by the second kilometre, holding 26–30 spm through the body of the race, with a 32–36 spm finishing sprint.
• Head races and 10 km+: rates clustered in the 24–28 spm range throughout, with brief surges to 30+ spm at landmarks.

The on-water patterns transfer to the ergometer with some adjustment. Concept2's published guidance ([8] Concept2 stroke-rate blog, Level 5) is consistent with the Kleshnev data: 24–28 spm for steady state (the long-race band), 30–36 spm for 2K-pace and interval work (the 2K band), 38+ spm for short sprints (the 500 m–1 km band).

[16] Kleshnev's World Rowing race-analysis compilation (2020, Level 5) confirms the patterns with later data. The 2019 world-championship finals showed the same general shape, with rates in the start of 2K races climbing slightly year-on-year, suggesting that elite crews have continued to find small gains by raising start rates without losing economy through the middle of the race.

Pacing a 2K: what cycling pacing can and cannot transfer

The 2K ergometer race is a 6-to-8-minute all-out effort, and the pacing literature for similar-duration efforts is largely from cycling time trials. The most-cited reference is [5] de Koning, Bobbert and Foster (1999, Level 5), who used an energy-flow model to simulate 1 km and 4 km track cycling time trials. Their conclusions, and the qualifications needed to apply them to rowing:

• 1 km time trial (≈60–70 seconds): the optimal strategy is all-out from the start, because the energy cost of a fast start is small relative to the duration. Power peaks at the start and decays through the race; split times are fastest at the start and slow toward the end.
• 4 km pursuit (≈240–260 seconds): the optimal strategy is all-out for the first 12 seconds, followed by constant anaerobic power output. The result is an "even-paced" race with a fast start.
• Both strategies: even small deviations from the optimal strategy cost meaningful time. A 1% deviation in pacing strategy can cost several seconds over the course of the event.

The transfer to a 2K rowing race is partial. The 2K is closer to the 4 km pursuit in duration than to the 1 km time trial, so the cycling result predicts a fast start followed by constant effort. The on-water 2K racing pattern ([2] Kleshnev 2000, Level 2b) and the ergometer 2K pacing patterns observed in elite competition are consistent with this prediction: 36–40 spm at the start, settling to 32–36 spm by the second 500 m, holding 32–34 spm through the third 500 m, kicking back to 36–40 spm in the final 500 m.

What the cycling model does not capture is the rowing-specific constraint of technique fatigue. A rower who starts a 2K at 40 spm and tries to hold 34 spm through the middle will, in many cases, see drive ratio fall and force curve flatten by the third 500 m ([15] Ingham et al. 2019, Level 2b). The "constant effort" that the cycling model predicts is, in rowing, a "constant technique" — same drive length, same force curve, same per-stroke work — not necessarily the same split. The rower who holds technique will hold the split; the rower who chases the split with shorter strokes will pay for it in the final 500 m.

[34] Tjønndal, Bjørndal and Samuel (2016, Level 5) extended the de Koning variational approach to derive optimal pacing under physiological constraints, including a fatigue cost per unit of power output. The model's prediction for events in the 4–8 minute range — the 2K ergometer race is exactly here — is a fast start (90–95% of peak power for the first 20–30 seconds), a controlled decay through the middle, and a small finishing surge. This is consistent with both the cycling empirical data and the rowing racing data.

Damper setting, drag factor, and the rate question

The Concept2 indoor rower has a damper lever that controls how much air flows into the flywheel. Higher damper = more air = heavier feel; lower damper = less air = lighter feel. The PM5 monitor displays a drag factor — a number, typically 80–220 for an unmodified ergometer, that approximates the deceleration-to-acceleration difference of the previous rowing cycle ([7] Concept2 PM5 docs, Level 5).

The important framing, made explicit by Concept2 in [10] the "What Damper Setting Should I Use?" blog (2019, Level 5): drag factor is what the PM5 uses to compute power; damper setting is the lever that controls it. Two ergometers set to "damper 4" can have different drag factors if the flywheel cages are in different condition. The PM5 self-calibrates to the actual deceleration, so the drag factor on the screen is the more honest number for comparison purposes.

[9] Concept2's "The Damper and Drag of Olympians" blog (2018, Level 5) collected drag factor ranges used by Olympic and national-team rowers:

• Eric Murray (NZ, 2× Olympic champion): drag factor 130.
• Caryn Davies (USA, 2× Olympic champion): drag factor ~125 (damper ~4).
• Gevvie Stone (USA, Olympic silver medallist): drag factor 112 (damper ~3); up to 115 for low-rate pieces.
• Linda Muri (US national team coach): drag factor 115.
• Todd Kennett (Cornell head coach): drag factor 140 for steady state.

The pattern across elite rowers is a cluster in the 112–140 drag-factor range for general training, with higher numbers for low-rate power work and lower numbers for high-rate work. [11] Concept2's "What is the Best Damper Setting for Me?" blog (2025, Level 5) makes the practical case: lower damper (≤3) makes higher stroke rates easier to sustain, while higher damper (≥6) favours low-rate power work because each stroke requires more force.

The rate-damper interaction is real but not large. A rower at 24 spm on damper 4 is producing roughly the same work per stroke as a rower at 24 spm on damper 6, but with different muscle-recruitment patterns; the high-damper rower is producing more peak force and less drive speed, the low-damper rower is producing less peak force and more drive speed. The PM5's force curve makes the difference visible in real time ([13] Concept2 working-with-force-curve, Level 5).

For most trained rowers, the practical advice is to find a drag factor in the 110–130 range that feels like water and to leave it there for most sessions ([10] Concept2 damper blog, Level 5). The exception is interval work at very high rates (above 36 spm) where lower drag factors reduce peak force per stroke and protect the lower back; and low-rate strength work where higher drag factors train peak force ([11] Concept2 damper blog, 2025, Level 5).

What the Concept2 PM5 actually measures: the accuracy question

The PM5 computes power from angular velocity, flywheel mass, and a constant factor ([9] Treff et al. 2022, Level 5). It does not measure the force on the handle directly; it measures the deceleration of the flywheel during the previous stroke cycle and uses the drag factor to translate that into a force estimate. The question of how accurate that estimate is was answered rigorously by [9] Treff, Mentz, Mayer, Winkert, Engleder and Steinacker (2022, Level 5), who tested an unmodified Concept2 Model D against a laboratory reference system (a load cell on the chain) using a motorized test rig.

The findings:

• During steady simulated rowing, the PM5 under-estimates mechanical power output by 2.9–4.3% across the stroke-rate range tested (22, 24, 26, 28 spm). The error is largest at low rates (4.3% at 22 spm) and smallest at high rates (3.6% at 28 spm). The differences are statistically significant (P < 0.001).
• During the first 5 strokes of any piece, the PM5 under-estimates power by 10–70%. The cause is that the flywheel starts from rest, so the deceleration-to-acceleration calculation is unstable until the flywheel has reached a steady-state rotation rate.
• During unsteady simulated rowing (alternating or random variations in stroke force and rate), the random error increases up to 18-fold, even though the mean difference stays in the 2.5–3.9% range.
• After excluding the first 5 strokes, the mean error falls to 0.2–1.9% across all conditions.

The practical implications for a rower:

• Trust the average, not the start. If you are doing 10 strokes at max power, the first 5 strokes will read artificially low. The PM5 is a credible power meter for anything beyond the first 5 strokes; for shorter pieces, the start-stroke error dominates the reading.
• Row evenly for accurate readings. Variations in force and rate across strokes increase the random error substantially. The PM5 is most accurate when the rower is steady.
• The 22-vs-24-spm difference is real. The Treff study explicitly notes that a pace difference of about 2.4 seconds over 6,000 m can arise from identical drive phases at different stroke rates, because the PM5's rate-dependent error means a 22-spm row is read lower than a 24-spm row even when the actual mechanical work is the same.

For an AI coach reading PM5 data, the implication is that the 2–4% systematic error and the start-stroke error should be averaged out across multi-minute sessions. For short sprint pieces, the AI coach should either discard the first 5 strokes or use a different proxy ([9] Treff et al. 2022, Level 5).

What the AI coach does with stroke rate

The MyNextRow AI coach reads stroke rate as one signal among several — alongside split, heart rate, drive ratio, force curve, and the rower's own subjective feel ([26] Foster et al. 2001, Level 5). The coaching logic is roughly:

• For steady-state sessions (UT2, Zone 1–2): the coach prescribes a rate in the 22–26 spm band, with the exact value determined by the rower's training age, recent session history, and target heart-rate zone. The rate is set to keep cardiac drift manageable over the session duration ([3] Bunc & Psota 2009, Level 2b).
• For threshold sessions (UT1, Zone 3): the coach prescribes a rate in the 26–30 spm band, with the rate set by the session's interval length. Shorter intervals (3–5 minutes) get the higher rates; longer intervals (8–12 minutes) get the lower rates. The trade-off is between the metabolic cost of higher rates and the technique cost of lower rates at high splits.
• For 2K-pace sessions: the coach prescribes a rate in the 30–34 spm band, with the rate set to match the rower's race plan. A rower targeting a fast start gets a higher opening rate; a rower targeting an even split gets a more constant rate.
• For sprint sessions (500 m–1 km, 30 s–2 min): the coach prescribes a rate in the 34–40+ spm band, with the rate set by the interval length. The race-pacing literature ([5] de Koning et al. 1999, Level 5) predicts an all-out opening for short sprints; the rowing-specific constraint is that technique must remain intact.

The coach's job is not to push the rate higher; it is to keep the rate in the band where the marginal watt is the marginal breath is the marginal heart-beat — the moving target that depends on the session's purpose and the rower's current state.

The coach also reads the drive ratio ([7] Concept2 PM5 docs, Level 5; [15] Ingham et al. 2019, Level 2b). The drive ratio is recovery time divided by drive time; a typical ratio is 2:1 at steady state (recovery twice as long as drive), and the ratio falls as the rower fatigues. A drive ratio below 1.5:1 in a steady-state piece is a warning sign: the rower is not giving the aerobic system enough time to recover between strokes, and cardiac drift will rise faster than the rate explains. The coach's job is to keep the drive ratio in the band where the rower can sustain the session.

The race plan for a 2K, in detail

Putting the pacing literature and the rate-by-distance patterns together, the practical plan for a 2K ergometer race is roughly:

1. First 200 m: stroke rate 36–40 spm, split 2–3 seconds faster than goal pace. The purpose is to bring the flywheel up to race speed quickly and to bank a small time cushion. The first 5 strokes' power reading is unreliable ([9] Treff et al. 2022, Level 5), so the split is the more honest proxy here.
2. 200 m–400 m: settle to goal split at 32–34 spm. The settle is the transition from anaerobic-burst to race-pace aerobic. The technique must remain intact — drive length, force curve, and posture — even as the rate drops.
3. 400 m–1,200 m: hold goal split at 32–34 spm. This is the steady-state middle of the race. The drive ratio should be approximately 2:1; the force curve should be smooth and repeatable.
4. 1,200 m–1,500 m: build gradually. The split should not change, but the rate can creep up to 33–35 spm. This is the controlled descent into the final quarter.
5. 1,500 m–1,800 m: hold and prepare. The split should be at goal pace or fractionally faster; the rate should be holding at 34–36 spm.
6. Final 200 m: sprint. The rate climbs to 36+ spm and the split opens up. The split is no longer the goal; the goal is the highest sustainable cadence with technique intact.

The empirical data from Kleshnev's on-water 2K analysis ([2] Kleshnev 2000, Level 2b) and Concept2's ergometer 2K race data ([8] Concept2 stroke-rate blog, Level 5) support this shape. The cycling pacing literature ([5] de Koning et al. 1999, Level 5; [34] Tjønndal et al. 2016, Level 5) predicts it from first principles. The empirical observation that the drive ratio falls across the piece ([15] Ingham et al. 2019, Level 2b) is the early warning that the rower is failing the plan; the AI coach should be watching the drive ratio through the third 500 m and flagging any drop below 1.5:1 as a sign to pull the rate back rather than the split.

Limitations and open questions

The rate-power trade-off has several open questions that the current literature does not fully resolve.

• The 20-spm lower bound. The Hofmijster 2007 study tested 20–36 spm; almost no other study has tested rates below 20 spm in trained rowers. The "strength-pulling" range of 16–20 spm is widely used by rowers for low-rate power work ([11] Concept2 damper blog, 2025, Level 5), but the biomechanics and metabolic cost at those rates are less well characterised. [4] Kleshnev (2003, Level 5) modelled ergometer power output mathematically but did not validate the model below 20 spm.
• On-water transfer of ergometer rates. [17] Vinther et al. (2011, Level 5) showed that sliding-based ergometer rowing allows higher rates of force development and changes the drive-length and drive-time profile relative to fixed-ergometer rowing. The implication is that on-water rate patterns may not transfer cleanly from ergometer rate patterns. [19] Legge et al. (2024, Level 5) — the 2024 systematic scoping review — noted that on-water biomechanics research is sparse and that most of the rate-by-distance evidence is from ergometer studies.
• The role of skill. [23] Hofmijster and Van Soest (2009, Level 5) showed that skilled rowers generate more useful power per stroke on a modified ergometer; the skill effect persists even when ergometer resistance is held constant. The implication is that the rate-power trade-off is not purely mechanical — it has a coordination component that the standard force-curve and drive-ratio metrics do not fully capture.
• The damper-rate interaction under load. Most of the rate-by-distance literature ([1] Hofmijster et al. 2007, Level 2b; [2] Kleshnev 2000, Level 2b; [8] Concept2 stroke-rate blog, Level 5) is at standard drag factors. The interaction between high drag factor and high stroke rate — the "damper 10 / 40 spm" combination that some strength-focused rowers use — is not well studied. The Treff 2022 accuracy study ([9] Treff et al. 2022, Level 5) tested drag factor 145 only.
• The 2K race plan in non-elite rowers. The pacing literature ([5] de Koning et al. 1999, Level 5; [34] Tjønndal et al. 2016, Level 5) is for trained endurance athletes. For non-elite and time-constrained rowers, the optimal pacing may be different — a more conservative start, a smaller finishing kick, and a longer steady middle. The empirical data on this population are sparse.

The honest read: the rate-power trade-off is well characterised for trained rowers at standard rates (24–36 spm) on standard ergometers (Concept2 Model C/D/E). The characterisations become less certain at the extremes of rate and at non-standard equipment.

What to do with this article

If you are an indoor rower who wants to apply this:

• For steady-state work, row at 22–26 spm. The rate is low enough to keep cardiac drift manageable over an hour, but high enough to maintain rhythm. Use a drag factor in the 110–130 range; the exact value should feel like water ([10] Concept2 damper blog, Level 5).
• For threshold work, row at 26–30 spm, with the rate set by the interval length. Watch the drive ratio; if it falls below 1.5:1, pull the rate back rather than the split.
• For 2K-pace work, row at 30–34 spm. Use the race plan above as a starting template and adjust for your training age and goal time.
• For sprint work, row at 34–40+ spm, but keep the piece short. The metabolic cost rises steeply at high rates; long sprint pieces are a mistake ([3] Bunc & Psota 2009, Level 2b).
• For technique work, row at 18–22 spm. The rate is low enough that you can think about each phase of the stroke — catch, drive, finish, recovery — without racing the monitor.
• For drag factor, use the PM5's drag factor, not the damper setting, as your reference. The PM5 self-calibrates; the damper does not. Most trained rowers cluster around 110–140; the exact value is a personal choice ([9] Concept2 damper-and-drag blog, Level 5).
• For trust in the monitor, treat the average after 6 strokes as the truth; treat the first 5 strokes as a flywheel-start artefact; treat uneven rowing as the most error-prone condition ([9] Treff et al. 2022, Level 5).

The AI coach reads your stroke rate alongside your split, your heart rate, your drive ratio, and your force curve. The rate is one of four or five signals the coach weighs; the rate by itself does not determine whether the session is good. The session is good if the rate, the split, the drive ratio, and your feel are all consistent with the session's purpose; the session is bad if any of those signals is out of band, regardless of what the others say.

The bottom line: stroke rate is a lever, not a verdict. The biomechanics literature says higher rates are net-efficient ([1] Hofmijster et al. 2007, Level 2b); the metabolic literature says gross efficiency is roughly constant across rates ([3] Bunc & Psota 2009, Level 2b); the racing literature says higher rates produce more boat speed ([2] Kleshnev 2000, Level 2b); the practical literature says the right rate is the highest one the rower can sustain with technique intact ([8] Concept2 stroke-rate blog, Level 5). The MyNextRow AI coach's job is to keep the rate in the band where the marginal watt is the marginal breath is the marginal heart-beat — a moving target that depends on duration, training age, and the session's purpose.