Why paddling needs its own numbers.

A power meter measures the force the rider puts through the pedals and how fast the crank turns. Force times crank length is torque, and torque times cadence is power. That number is the athlete's output, so a hard effort reads hard into any wind, and cyclists train by it.

Most paddlers only have speed, and speed is shared with the conditions. In the Locus Fit archive, a 10 km/h wind along the course made the downwind leg 5% faster than the upwind one at the same stroke rate (1,105 sessions). In the median session the two directions differed by 3.1%: about five seconds over 500 m at training pace.

Current is harder, because GPS can't see it. On a flowing river, a kayak's power tracks its speed through the water closely and its GPS speed much less well (Hogan et al. 2022).

Today, every session on Locus Fit is read with the weather for its time and place, and pieces paddled into the wind are set against pieces paddled away from it. Locus One is designed to add a wind sensor and speed through the water, so the conditions can be taken out rather than only noted.

Heart rate takes a minute or more to catch up with an effort. On long steady work that doesn't matter. On the short intervals sprint paddling is built on, it reads the rest better than the work, and it misclassifies the time athletes spend in each zone (Hogan et al. 2021).

The same training block can look very different depending on what measures it. Over eight weeks, one group of German kayakers spent 91% of their time easy by heart rate, but only 68% to 73% by speed (Matzka et al. 2022).

The cheapest measure that handles intervals well is the oldest: minutes multiplied by the athlete's own 1-to-10 rating of the session (Foster et al. 2001). The Locus Fit watch app can record that rating when a session ends, and the read uses it.

We read every paddling session on the platform: 4,471 of them, from 11 athletes, recorded on Garmin and Polar watches between 2016 and 2026. A channel being in the file is not the same as the data being usable, so each one was checked.

GPS every second

1,327 of 4,471 sessions. More than half (2,447) logged a point every 4 seconds or slower, Garmin's default smart recording, where a 20-second sprint is five points or fewer.

A heart rate from a real heartbeat

Of the 2,022 solo paddles with a heart-rate channel, 1,165 behaved like one: it followed boat speed with a lag and reached a working rate. 585 didn't: it never reached a working heart rate at working speed, or never followed the effort. A watch on the boat still records a number.

A usable stroke rate

In kayak, the watch's own profile counted a left-and-right cycle as one stroke in 209 of 232 sessions: half the real rate. In canoe, none of the off-the-shelf apps we tried counted strokes reliably; only the Locus Fit app's count is used.

All three

272 of 4,471 sessions had one-second GPS, a usable stroke rate and a real heart rate. In canoe, 7 of 1,656.

Some of the gap is settings, and the free Locus Fit Garmin app closes that part by counting strokes on the watch athletes already own. The rest is where the watch sits, which is why Locus One puts the sensor on the paddle.

Elite endurance athletes do about 80% of their sessions easy and most of the rest hard, with little in between (Seiler 2010). In a randomised trial of runners, cyclists, triathletes and skiers, that polarized pattern produced the largest gain in aerobic capacity (Stöggl and Sperlich 2014). Many elite programs are pyramidal instead: mostly easy, some threshold, a little very hard.

In a season-long study of German junior sprint paddlers, they trained 80% to 91% easy by heart rate. More time easy went with more improvement over 1,500 m, and more time in the middle zone with less (Matzka et al. 2021).

The middle is also where group training can settle without anyone choosing it. When a squad paddles hard intervals together, the pieces tend to converge on a shared pace, and easy days creep up to meet them. That is hard to see from one boat and plain in the data, which is why Locus Fit reads what was paddled against what was planned, zone by zone.

A 200 m race is about a third aerobic; a 1000 m race is three-quarters or more (Byrnes and Kearney 1997). The athlete built to win one is often not built for the other, and the same program does not suit both. An athlete's build has three parts: the engine, the frame and the range of motion.

The engine. Muscle fibre type is part of it. After three 30-second all-out sprints, athletes with a fast-twitch profile had still not recovered five hours later; slow-twitch athletes had recovered in twenty minutes (Lievens et al. 2020). In a three-week overload block, the runners who overreached were the ones with more fast-twitch fibre (Bellinger et al. 2020). Fibre type itself takes a muscle biopsy or a specialised MRI scan, and no club does either. It shows in data a squad already records: how far an athlete's top speed sits above their aerobic speed, the speed reserve used to profile 800 m runners (Sandford et al. 2019); how fast their pace fades across a set; and how long they take to be ready for the next hard day.

The frame. Stature and proportions set the levers. Olympic sprint paddlers are strikingly alike: lean, with proportionally large arm and chest girths, and narrow hips in the men (Ackland et al. 2003). Body size predicted where Olympic kayakers set their foot bar and how wide they held the paddle, and moving three national-level kayakers off their own set-up slowed the boat (Ong et al. 2006). In teenagers the frame is still growing: the fastest under-14 kayakers sat taller, had paddled for more years and were further through puberty than the rest (Fernandes et al. 2021), so part of a junior result is timing.

The range. Flexibility decides the posture an athlete can hold in the boat. Young kayakers with tighter hamstrings sat in the kayak with more spinal flexion and a pelvis tilted further back, though they stood the same as everyone else (López-Miñarro et al. 2012).

That profile is what we are building next: the engine read from training data, with the frame and range from a tape measure and a few simple tests. It will estimate a type; it will never claim to identify muscle fibres.