Two bikes sit side by side. The riders match cadence, select what appears to be similar resistance and report the same perceived effort. One console shows 180 watts and the other 215. It is tempting to conclude that one rider is stronger or one bike is broken. Neither conclusion follows automatically.
Power is not simply “inside” the bike waiting to be displayed. In indoor cycling Singapore, the number may be measured or estimated at a specific location, processed through a device algorithm, averaged over time and affected by calibration, temperature, component wear and rider technique. Understanding that chain prevents false competition and makes power genuinely useful for training.
Power Is Torque Multiplied by Rotational Speed
In cycling terms, power reflects how much torque is applied and how quickly the crank or flywheel rotates. Cadence provides rotational speed. The system still needs a way to determine torque or resistance.
Some devices use strain gauges that deform minutely under load. Others infer power from flywheel speed, braking force, magnetic position or a manufacturer-specific relationship between resistance and cadence. A direct-force system and an estimation model can both report watts, but they do not necessarily arrive there through the same measurement path.
This is the first source of bike-to-bike disagreement. Equal displayed cadence does not guarantee equal torque, and equal resistance labels do not guarantee the same braking force.
Measurement Location Changes the Answer
Power can be measured at the pedal, crank, spindle, chainring, hub, flywheel or resistance unit. Energy is lost through the drivetrain between some of these points. Bearings, belts, chains and alignment all influence how much power reaches the downstream sensor.
A pedal-based meter may report the work entering the crank system. A flywheel-based estimate may reflect what arrives after transmission losses, or it may use a model that incorporates assumed losses. Two accurate devices can therefore disagree because they measure different points.
This is similar to measuring business revenue at order placement versus cash collection. Both numbers can be valid, but they describe different stages. Riders should ask what the device measures before demanding equivalence.
Calibration and Zero Offset Are Not the Same as Accuracy
Calibration establishes or checks the relationship between sensor output and a known reference. A zero-offset procedure tells the system what “no applied torque” looks like at that moment. Manufacturers sometimes use the word calibration for both, but the processes are conceptually different.
If a sensor’s zero drifts, it can add or subtract apparent torque across the ride. Temperature, installation and material behaviour may influence this baseline. Some devices compensate automatically, while others require a manual procedure or warm-up.
Accuracy is broader. It asks how closely readings reflect a reference under defined conditions. Repeatability asks whether the device produces similar results when the conditions are repeated. A device can be consistently high, which makes it imperfect for absolute comparison but still useful for tracking personal trends on the same bike.
A systematic review of cycling power-meter validity highlights accuracy, sensitivity, repeatability, reproducibility and robustness as distinct properties. It also notes that power range, cadence, torque, rider position, temperature and vibration can affect testing. A single advertised percentage cannot describe performance in every class condition.
Resistance Systems Create Different Curves
Indoor bikes may use magnetic, friction or electronically controlled resistance. The relationship between knob position, flywheel speed and braking force is not necessarily linear.
On one bike, a small turn near the middle of the resistance range may create a large torque change. On another, the same visible turn may have a modest effect. Knobs can also lack absolute indexing, so “two turns” begins from an unknown starting point.
Electronically controlled bikes may target a resistance level or power output through software. Their control algorithms decide how quickly to adjust and how to respond when cadence changes. During a sudden acceleration, one bike may stabilise rapidly while another takes several seconds. The average power for a short interval can then differ even if steady-state readings later converge.
Averaging Makes Identical Efforts Look Different
Displays rarely show every instantaneous fluctuation. They may report one-second, three-second, five-second or longer rolling averages. A longer average looks smoother but reacts slowly. A shorter average responds quickly but appears noisy.
Suppose a rider accelerates from 150 to 300 watts. A one-second display will show the surge almost immediately. A five-second display blends new and old values, so the number rises gradually. If the interval lasts only 15 seconds, the difference in averaging has a meaningful effect on the visible peak and average.
Data recording frequency matters too. One console may store one value per second. Another may use smart recording that saves points when the software detects change. Summary metrics can therefore differ even when the live experience felt similar.
Left-Right Assumptions Can Distort Estimates
Some power systems measure one side and double the result. This assumes that left and right legs contribute equally. Many riders have small asymmetries, and the balance may change with fatigue, cadence, standing position or injury history.
If the measured leg contributes 48 per cent of total work, doubling it underestimates total output. If it contributes 52 per cent, doubling overestimates. A dual-sided or total-system measurement avoids that particular assumption but introduces its own calibration and processing requirements.
The practical lesson is not to chase perfect symmetry. It is to understand whether the device’s watt figure includes an assumption that changes under different riding conditions.
Technique Alters What the Sensor Sees
Two riders can report the same RPE but apply force differently. One produces smooth torque through most of the circle. Another creates sharp peaks and long low-force phases. Depending on sampling and sensor location, the systems may process these patterns differently.
Standing also changes force application and drivetrain loading. Some meters maintain validity across positions better than others. Cadence extremes, very low torque and explosive sprints can challenge measurement systems that perform well during steady submaximal riding.
This is one reason validation should cover the conditions in which the device will actually be used. Accuracy at 200 watts and 90 rpm does not prove identical behaviour at 60 rpm climbing torque or a 900-watt sprint.
Maintenance Creates Fleet Variation
Studio bikes accumulate hours at different rates. Belts stretch, bearings wear, bolts settle, firmware changes and sensors age. Preventive maintenance reduces variation but cannot make every component history identical.
Cleaning also matters. Sweat is corrosive, and repeated exposure can affect hardware if not managed. A bike that feels mechanically rough may change the rider’s technique and perceived effort even if the power sensor itself remains functional.
When a reading looks implausible, check the simplest factors first: bike setup, resistance behaviour, cadence sensor, firmware state where accessible and whether the bike has completed its recommended warm-up or zeroing procedure. Studio users should report consistent anomalies rather than trying to repair equipment themselves.
Compare Within a Measurement Ecosystem
Power becomes actionable when comparisons are controlled. Use the same bike where possible. If that is not possible, use relative zones, RPE and heart-rate response rather than ranking absolute watts across the room.
For personal tracking, record bike identifier, class type, interval duration, cadence and RPE alongside power. If 200 watts on Bike A consistently corresponds to RPE 7, while the same number on Bike B corresponds to RPE 5, treat them as different baselines.
The I.C.G programme described by TFX Singapore translates speed and strength into colour-coded power zones. Zone-based coaching can be useful because it focuses on the rider’s own working range. The interpretation remains strongest when the bike, setup and calibration context are consistent.
Use Power for Decisions, Not Identity
Watts should answer training questions: Is the interval repeatable? Did output fade? Can the rider hold more work at the same RPE? Does heart rate respond differently at a familiar load? It should not become a direct ranking across unknown devices.
If two bikes disagree, first ask whether the difference is stable. A stable offset can often be managed by keeping separate baselines. An erratic reading is more concerning because it weakens repeatability. In both cases, technique and physiological response remain valuable cross-checks.
The most sophisticated rider is not the one who believes every digit. It is the one who understands what the digit measures, how it was produced and which comparisons it can support.
Frequently Asked Questions
Does a higher watt number mean one rider is fitter?
Not necessarily. Body size, bike calibration, sensor type, measurement location and interval execution all affect the number. Fitness comparisons require controlled equipment and appropriately normalised testing.
Should I always book the same studio bike?
Using the same bike can improve trend consistency. If that is impractical, record the bike identifier and use RPE, heart-rate response and relative zones to interpret output.
What is the difference between zeroing and calibration?
Zeroing establishes the sensor’s no-load baseline. Calibration checks or establishes the relationship between sensor output and a known reference. Device instructions may use the terms differently, so follow the manufacturer’s procedure.
Why does power jump when my cadence changes?
Power combines torque and rotational speed, so cadence changes can alter output immediately. The bike’s resistance-control algorithm and display averaging may also make the visible number jump or lag.
