Watts per kilogram explained: the number and its limits

Watts per kilogram is power divided by body mass. A rider holding 300 watts at 60 kilograms is at 5 W/kg, and a rider holding 390 watts at 78 kilograms is at the same 5 W/kg. On a steep climb those two go up at almost exactly the same speed, despite one producing ninety watts more than the other.

That is the whole appeal of the ratio and also the source of nearly every bad argument made with it. It is the correct comparison for one specific situation, it is close to meaningless in another, and it is completely meaningless without a duration attached. Most of the time somebody quotes a W/kg figure, at least one of those three conditions has been quietly dropped.

Here is what the number actually describes, the physics that makes it the right divisor on a climb, and the five separate ways it gets misused.

What watts per kilogram actually divides

The arithmetic is trivial. Take the power a rider produced, in watts, and divide by their mass in kilograms. The result has units of watts per kilogram, which reads as power available per unit of mass that has to be moved.

The reason that quantity is interesting has nothing to do with the arithmetic and everything to do with what a bicycle is fighting against.

A rider on a bicycle spends power on four things. Lifting mass against gravity. Pushing air out of the way. Deforming tyres and road surface, which is rolling resistance. And overcoming friction in the drivetrain. The relative size of those four changes completely depending on the gradient, and the ratio is useful exactly to the extent that the first one dominates.

On a steep climb at a professional pace, the overwhelming majority of the power is going into gravity. On a flat road at racing speed, the overwhelming majority is going into air. Those are two different problems, and they reward two different kinds of rider, which is why cycling has climbers and rouleurs rather than just fast people.

Why gravity makes the ratio the right number on a climb

The physics of climbing is unusually clean, which is rare in sport and worth appreciating.

The power needed to raise a mass is the mass multiplied by gravitational acceleration multiplied by the rate of vertical ascent. Gravitational acceleration is a constant, close to 9.81 metres per second squared, and it does not care who you are. So the vertical speed a rider can sustain is their available power divided by the total mass being lifted, times a constant.

Rearranged, that sentence is the entire justification for the ratio. Vertical ascent rate is proportional to power divided by mass. Double the power at the same mass and you climb twice as fast. Double the mass at the same power and you climb half as fast. The relationship is linear and there is no fudge in it.

This is why climbing times are the one place in cycling where a single number genuinely predicts performance. Two riders at the same W/kg on the same climb, on the same bikes, in still air, arrive at the top together regardless of whether one of them is producing a hundred watts more than the other.

The qualifications in that sentence are doing work, and the rest of this piece is mostly about them. But the core relationship holds, and it holds more tightly the steeper the road gets, because steeper means slower and slower means less of the power goes to air.

• 390Rider P, threshold power in watts
• 78Rider P, body mass in kilograms
• 300Rider Q, threshold power in watts
• 60Rider Q, body mass in kilograms

Constructed figures, not any real rider's data. Both riders arrive at 5.0 W/kg by completely different routes, and on a steep climb in still air they climb at the same speed.

Why air resistance makes it the wrong number on the flat

Flip the terrain and the arithmetic inverts, for a reason that is equally clean.

Aerodynamic drag depends on air density, on the rider's frontal area combined with how cleanly they shed the air, and on the square of speed. The power needed to overcome it depends on the cube of speed. Mass does not appear in the expression at all.

Mass does appear in rolling resistance, which scales with weight on the tyres. But at racing speed on a flat road, rolling resistance is a minority of the total and aerodynamic drag is the majority, so the mass-dependent term is the smaller one.

Now add the thing that makes this decisive. A bigger rider does have a bigger frontal area, so they are not escaping the drag penalty entirely. But frontal area does not scale with mass. Under rough geometric similarity, area grows with roughly the two-thirds power of mass: make a rider fifty per cent heavier and you make them something like a third larger in frontal area, not fifty per cent larger. Mass goes up faster than the aerodynamic penalty it brings.

The consequence is that absolute watts win on the flat and the heavier rider is advantaged. Take the two invented riders above. Rider P has ninety watts more and perhaps a third more frontal area, which is a trade that favours P substantially. Put them in a flat time trial and P rides away. Put them on a long mountain pass and they finish together. Neither rider got better or worse; the question changed.

The full treatment of what drag actually depends on, and why position matters more than any of it, is in the piece on cycling aerodynamics.

A figure without a duration is not a figure

This is the error that appears most often and is the easiest to spot once you know to look.

Power is not a single property of a rider. It is a curve against time. A rider can produce an enormous number for a few seconds, a large one for a minute, a moderate one for twenty minutes and a smaller one for several hours, and all four are true of the same person on the same afternoon.

Divide each of those by the same body mass and you get four completely different W/kg figures. Here is an invented rider's curve, with round numbers chosen so every division works out cleanly.

Best 5 seconds, 1400 W20W/kg

Best 1 minute, 630 W9W/kg

Best 5 minutes, 420 W6W/kg

Best 20 minutes, 350 W5W/kg

Best 60 minutes, 315 W4.5W/kg

A constructed power curve for a 70kg rider, with values picked so each division is exact. All five figures describe the same rider on the same day.

Show the numbers

Invented worked example: one rider, five durations, five different ratios

Item	Watts per kilogram
Best 5 seconds, 1400 W	20W/kg
Best 1 minute, 630 W	9W/kg
Best 5 minutes, 420 W	6W/kg
Best 20 minutes, 350 W	5W/kg
Best 60 minutes, 315 W	4.5W/kg

Twenty against four and a half. The same rider, the same scales, the same power meter, a factor of more than four between the highest and lowest figure.

So a claim that a rider is "at 6 W/kg" has not said anything until it says for how long. Six for five minutes and six for an hour are different athletes in different sports. And the relationship between a rider's short-duration numbers and their long-duration numbers is itself informative: a rider whose five-second figure is enormous relative to their twenty-minute figure is a sprinter, and one whose curve is flat across the longer durations is a diesel. The shape of the curve is the rider's profile, and the single ratio is one point on it.

The underlying reason the curve exists at all, and the model that describes it properly, is covered in the piece on power meters and threshold power.

The four durations worth knowing, and what each one tells you

Different parts of the curve answer different questions, and knowing which part you are looking at is most of the interpretation.

Five seconds. A neuromuscular measurement. This is the sprint, and it is governed by muscle mass, recruitment and mechanics rather than by anything aerobic. Dividing it by body mass is arguably the wrong operation entirely, because a flat sprint is an aerodynamic problem and a bigger sprinter with more absolute power usually wins. The ratio only becomes relevant for a sprint up a steep ramp.

One minute. Heavily anaerobic, and the duration that decides a short steep climb or a bridging effort. Useful for riders who race on punchy terrain, and a poor guide to stage-race climbing.

Five minutes. Close to the duration at which maximal oxygen uptake is the binding constraint. This is the figure that decides a short mountain-top finish and the one most sensitive to a rider's aerobic ceiling.

Twenty to sixty minutes. The threshold region, and the one that decides long climbs and grand tours. This is the figure people usually mean when they quote a W/kg number without saying, and it is the figure that correlates most strongly with being a good climber.

The practical rule is that the duration of the question should match the duration of the figure. A rider's twenty-minute ratio tells you very little about whether they will win a bunch sprint, and their five-second ratio tells you nothing at all about a thirty-minute climb.

The denominator problem: does the bike count?

By convention, W/kg is quoted against body mass alone. The physics does not care about the convention.

Gravity acts on everything going up the hill: the rider, the bicycle, the bottles, the shoes, the helmet, the food in the pockets. If you want the ratio to predict climbing speed, the honest denominator is the whole system mass.

Including the bike has an interesting effect on comparisons between riders. It adds roughly the same absolute mass to every rider, which means it adds a larger proportional amount to the lighter rider's denominator. That compresses the gaps.

Take the invented riders again and give them both an eight-kilogram bike with bottles and kit. Rider P goes from 390 over 78, which is exactly 5.0, to 390 over 86, which is about 4.53. Rider Q goes from 300 over 60, which is exactly 5.0, to 300 over 68, which is about 4.41.

The two riders who were level on body mass alone are no longer level once the machine is included, and the heavier rider has moved ahead. That is not a trick: it is the correct answer. The bike is a fixed cost, and a fixed cost is proportionally heavier for a smaller rider.

This is also why the minimum bike weight in the technical regulations matters more than it looks. Because professional bikes sit at a regulated floor rather than at whatever a budget can buy, the machine is close to a constant across the peloton, which keeps the comparison between riders meaningful. The reasoning behind that floor is set out in the piece on the UCI technical rules.

The measurement problems, in order of how much they matter

Before any of the interpretation, the number has to be right, and there are four reasons it might not be.

Calibration drift. Strain gauges change behaviour with temperature. A meter zeroed indoors and ridden in the cold reads differently from one zeroed at the roadside. A rider comparing this month's figure against last month's may be comparing two thermal conditions.

Where the meter sits. A hub-based meter reads downstream of the chain and jockey wheels, which consume a small fraction of what the legs produced. A crank or spider meter reads upstream of them. Neither is wrong, and the two will not agree on the same effort. Comparing your figure against somebody else's is therefore comparing two instruments that were never calibrated against each other.

Single-sided estimation. A meter that reads one crank and doubles it assumes symmetry. Most riders are not perfectly symmetric, and the asymmetry can change with fatigue, with position and with injury. The resulting error is systematic rather than random, which makes it worse: it does not average out.

The scales. The denominator gets much less scrutiny than the numerator and is just as capable of being wrong. Body mass varies through a day by a meaningful fraction of a kilogram, and a rider weighing themselves after breakfast and dividing a morning power figure by it has introduced an error larger than most training adaptations.

The combined effect is that a W/kg figure is good to perhaps a few per cent, and differences smaller than that are noise. A rider announcing an improvement from 4.9 to 5.0 has measured something, and it may well not be fitness.

Why race figures quoted on television are models, not measurements

Almost every professional power figure that reaches the public after a mountain stage was calculated rather than read.

The method is straightforward in outline. Take the climb's length and average gradient, take the rider's time, assume a total mass for rider and bike, assume values for air density, rolling resistance and drivetrain efficiency, and solve for the power that would produce that ascent in that time.

The method is also sensitive to every one of those assumptions, and several of them are genuinely unknown from the outside.

Mass is assumed. Rider weights are often approximate, and a few kilograms of error moves the output proportionally.

Wind is usually ignored. A tailwind up a climb reduces the power needed for the same time, and a headwind increases it. Models typically assume still air because nothing better is available, and a stage with a consistent valley wind can shift the estimate substantially.

Drafting is hard to handle. A rider who sat in a group for most of a climb spent less power than one who rode it alone, and the model has no way to see the difference without knowing the group dynamics metre by metre. The economics of that sheltering effect are covered in the piece on drafting in a peloton.

Gradient data is averaged. Real climbs vary, and a model using an average gradient over a segment that includes a flat section and a steep one will misestimate.

None of this makes the estimates worthless. They are useful for comparing like with like, especially the same climb in different years. They are not readings, the margin on them is wide, and a public figure quoted to one decimal place is implying a precision that the method does not support.

The aerodynamic cost of being small

The ratio flatters light riders, and there is a counterweight to it that gets discussed far less than it should.

A smaller rider has less frontal area and therefore less drag in absolute terms, which sounds like an advantage and partly is. But they also have less absolute power, and the balance between those two is what decides flat performance. Because frontal area grows more slowly than mass, the smaller rider's drag saving does not keep up with their power deficit.

There is a second effect that works in the same direction. Drag is not purely about frontal area; it is about the combination of area and how cleanly the shape sheds air. A rider's position on the bike is a large part of that, and position is partly a function of flexibility, bike fit and the dimensional limits in the regulations rather than of body size. Two riders of very different mass can end up with surprisingly similar effective drag if the larger one has the better position.

The net result is that the lightest riders in professional cycling are usually poor time triallists and the best time triallists are usually too heavy to win a mountain-top finish against a specialist. That is not a coincidence or a training failure. It is the direct consequence of the two different resistance regimes, and it is why grand tours, which contain both kinds of terrain, are won by riders who are good at both rather than exceptional at either. The piece on grand tour team selection works through how teams build a roster around that trade-off.

Where the crossover actually happens

The ratio matters on a climb and does not matter on the flat, which raises the obvious question of where the boundary sits. The answer is that there is no boundary, only a continuous shift, and the variable that controls it is speed rather than gradient directly.

The mechanism is worth following. Gravitational power scales with mass and with vertical speed, so it rises in proportion to how fast you are going up. Aerodynamic power scales with the cube of forward speed, so it rises very steeply with how fast you are going along. On a steep road, forward speed is low, the cubed term is small, and gravity takes nearly everything. On a flat road, forward speed is high, the cubed term is enormous, and gravity takes nothing at all because there is no vertical component.

Between those two extremes the two terms trade places gradually. On a shallow drag the split might be something like half and half; on a steep alpine pass the aerodynamic share becomes a minor correction; on a false flat at speed it is the dominant cost again. Nobody rides either pure case for long, which is why real cycling rewards riders who are adequate at both rather than specialists at one.

Two practical consequences fall out of this, and both are counterintuitive until you have seen the mechanism.

A heavier rider loses less on a shallow climb than the ratio implies. Because a meaningful share of the resistance on a shallow gradient is still aerodynamic, the lighter rider's mass advantage is diluted. A climb at a modest gradient at racing speed is a much more even contest than the same vertical gain packed into a steep ramp.

Descending reverses the whole calculation. Going down, gravity is working for the rider rather than against them, and it works harder on a heavier rider. Air resistance is the only thing opposing the descent, and it depends on area rather than mass. The heavier rider therefore descends faster for free, which is why a mountain stage with a summit finish and a mountain stage with a long descent to the line are different races even when the climbing is identical.

That second point is the honest counterweight to every argument about climbing weight. A rider who optimises purely for the uphill ratio has also optimised themselves into losing time on every descent in the race.

What counts as a good figure, and why the answer is a range

The question everyone asks about W/kg is what number is good, and the honest answer is that the question is underspecified in three ways at once.

It needs a duration, for the reasons above. It needs a context, because a figure that would be unremarkable among climbing specialists would be exceptional in a local club race. And it needs a terrain, because a rider whose profile suits flat racing can be genuinely strong while sitting well down any ranking by power to weight.

What can be said with confidence is the structure rather than the values. Figures fall as duration rises, steeply at first and then gradually. Climbing specialists sit at the top of the distribution for the longer durations, because that is the quality their entire physiology and body composition is organised around. Sprinters sit at the top for the shortest durations and well down the list for the longest, because muscle mass that wins a bunch sprint is mass that has to be carried uphill. Riders who win flat time trials and one-day races over rolling terrain often sit in the middle of any power-to-weight table while producing absolute wattage that nobody else in the race can match.

Amateur racing categories are, in effect, a crude sort by this quantity, even where the rules are written in terms of results rather than power. A rider moving up through categories is usually a rider whose threshold ratio is rising, and a rider who plateaus in a category has usually plateaued on that number. The correspondence is loose and the exceptions are the interesting ones: riders who win on tactics, on positioning, or on being the only person in the field who can hold a time trial position properly.

The useful personal benchmark is not a published table. It is your own figure, measured the same way, three months ago. That comparison is free of instrumentation differences, free of assumptions about anyone else's mass, and the only one that tells you whether the training is working.

The ratio in a time trial, where it mostly disappears

A flat time trial is the cleanest demonstration that W/kg is a terrain-specific metric, because it is the event where the number almost stops mattering.

The problem in a time trial is to sustain the highest possible power in the lowest possible drag for a fixed distance. Mass enters only through rolling resistance and through the small amount of climbing the course contains. The binding constraints are absolute watts and effective frontal area, and the second of those is a function of position, equipment and the dimensional limits the technical regulations impose rather than of body size. The mechanics of that are set out in the piece on the time trial.

This produces the most reliable mismatch in the sport. A rider can be among the best climbers in a race and lose several minutes in a flat time trial to a rider they dropped easily on every mountain, and neither result is an upset. They were asked two different questions and gave two correct answers.

It also explains why stage race general classifications are won by a narrower type of rider than either specialism would suggest. A grand tour contains flat time trials and mountain finishes, which means the winner has to be strong in absolute watts and strong in watts per kilogram simultaneously. Those two requirements pull body composition in opposite directions, and the riders who win are the ones who find a workable compromise rather than the ones who maximise either number.

Why losing weight is not a free way to raise the ratio

The ratio has a denominator, and the obvious inference is that reducing it raises the number. This is the single most dangerous idea in endurance sport, and it needs stating carefully.

Arithmetically, the inference is correct. A rider at 300 watts and 65 kilograms is at about 4.62 W/kg, and the same rider at 62 kilograms would be at about 4.84, which is a meaningful improvement for no training at all.

Physiologically, the inference assumes that power stays constant while mass falls, and that assumption breaks at some point for every rider. Muscle contributes to the numerator. Fuelling contributes to the ability to complete the training that builds the numerator. Below a certain point, reducing mass reduces power faster than it reduces mass, and the ratio goes down. Worse, the consequences of getting this wrong are not confined to performance, and cycling has a long and poorly handled history with disordered eating.

The honest version of the advice is that each rider has a mass at which their ratio peaks, that the peak is individual and not knowable from a formula, that approaching it from above produces improvement and overshooting it produces decline, and that the overshoot is much harder to detect than the approach because the ratio keeps looking fine for a while before the numerator collapses.

A rider whose power is falling as their mass falls has passed the point. That is the only reliable signal available from the outside, and it arrives late.

What the ratio cannot see

A power-to-weight figure is a snapshot of a rested rider producing a maximal effort on demand. Races are not that.

Durability. The ability to still produce the number after four hours of racing is a separate quality from the number itself, and it is increasingly treated as an independent parameter rather than a consequence of fitness. Two riders with identical fresh ratios can differ enormously in what remains after a long day, and the one who holds more of it wins climbs the fresh figures say they should lose.

Repeatability. A mountain stage with four climbs asks for four hard efforts with incomplete recovery between them. A single-effort ratio says nothing about the second, third or fourth.

Heat and altitude. Both reduce sustainable power, by amounts that vary between individuals. A rider with a superb ratio measured in cool conditions at sea level may not be the same rider on a thirty-five degree afternoon at two thousand metres.

Everything tactical. Positioning, descending, cornering, reading a race, knowing when to spend. None of it appears in a ratio, and it decides a very large share of results. The piece on team tactics is about the half of the sport that no power figure captures.

Freshness management. A rider who arrives at the decisive climb having spent two hundred watts less than a rival in the preceding four hours has a better effective ratio at the moment it matters, regardless of what either of them could produce fresh.

1. 1Fix the duration firstDecide what question you are answering. A long climb is a twenty to sixty minute question, a short steep ramp is a one to five minute question, and a sprint is a five second question. The duration determines which effort you need, not the other way round.
2. 2Produce a genuinely maximal effortOn a consistent course or indoors, after a proper warm-up, with the meter zeroed in the conditions you are riding in. A submaximal effort produces a figure that understates you and cannot be compared with anything.
3. 3Record the average power for exactly that durationUse the average over the stated window, not normalised power and not a peak. Normalised power is designed for variable rides and will inflate a steady effort's figure if misapplied.
4. 4Weigh yourself consistentlySame scales, same time of day, same conditions. Body mass moves by a meaningful fraction of a kilogram across a day, and the denominator deserves the same care as the numerator.
5. 5Divide, and state both halvesReport the figure with its duration and the mass it was divided by. A number on its own is not a claim anyone can check or reproduce.
6. 6Add the system mass when predicting a climbFor working out how fast you will actually go uphill, include the bike, bottles and kit. This lowers the figure and is the version that corresponds to the physics.
7. 7Compare only against yourselfDifferent meters read at different points in the drivetrain and are not calibrated against each other. Your own figure, measured the same way over time, is a real measurement. The gap between your figure and somebody else's is partly instrumentation.
8. 8Check the curve, not the pointLook at your five second, one minute, five minute and twenty minute figures together. The shape tells you what kind of rider you are and which part is not improving, which a single ratio cannot.

The sequence that produces a figure worth comparing. Every step is a place where a careless answer makes the final number meaningless.

How to read a watts per kilogram claim

Five questions turn a quoted figure into something assessable, and most claims fail at the first or second.

For how long? No duration means no claim. This disposes of a surprising proportion of what gets said about riders.

Measured or modelled? A figure from a rider's own power file is a measurement with instrument error. A figure calculated from a climb time is a model with assumption error, and the assumption error is larger. Race figures in the press are almost always the second kind.

Divided by what? Body mass alone or the full system. The difference is not small, and it moves comparisons between riders of different sizes in a predictable direction.

On what terrain does it matter? A ratio is the right comparison for a climb and close to the wrong one for a flat time trial. A story using W/kg to explain a flat result has used the wrong number.

Fresh or deep into a race? A figure produced in the fourth hour of a mountain stage is a different and more impressive quantity than the same figure produced rested on a Tuesday, and almost nothing published distinguishes the two.

Ask those five and the ratio becomes what it should be: a clean, physically justified measure of one thing, with its boundaries visible. It predicts climbing speed well, flat speed badly and race outcomes only partially, and that is a reasonable amount to ask of a single division. The rest of the cycling explainers here take the same approach, starting with the physics before asking what the number is actually good for.