Cycling Speed Is Mostly a Fight With Air

Three resistances, and only one of them runs away with you

The whole model fits on one line. The power arriving at the road is P = Crr·m·g·v + m·g·sin(θ)·v + ½·ρ·CdA·v_air²·v, where v is your speed over the ground, v_air is your speed through the air, θ is the gradient, ρ is air density and CdA is your drag area — the frontal area you present multiplied by how badly shaped you are. Divide the whole thing by about 0.97 to get the power your legs must produce, because a chain and derailleur lose roughly 3%.

The first term is the tyres deforming against the road and the road deforming back. As a force it barely changes with speed, so its power is proportional to v. The second term is gravity, and on flat ground it is exactly zero — which is why flat riding is such a clean experiment. The third is the one that matters. Drag force goes with the square of air speed, and power is force multiplied by ground speed, so on a windless day the aero power term goes with v cubed. Linear against cubic is not a close race for long.

Where air overtakes the road, and it is slower than you think

Set the two terms equal and solve. Crr·m·g·v = ½·ρ·CdA·v³ cancels one v from each side and leaves v = √(Crr·m·g ÷ (½·ρ·CdA)). Feed in a rider and bike of 176 lb, Crr 0.005, ρ 1.225 kg/m³ and CdA 0.32 m², and the answer is 4.47 metres per second — 10.0 miles per hour. Below that the tyres are your main enemy; above it, the air is. Ten miles an hour is a gentle cruise on a towpath. Almost every mile you have ever ridden with intent has been spent mainly on air.

Read the crossover formula again and notice what it does not contain: it does not contain your fitness, and the speed itself has cancelled out of the ratio in a very specific way. What it does contain is mass and CdA, both of which you can change. Heavier rider, higher crossover; more aerodynamic position, lower crossover. Better tyres move it down too, and this is the practical reason a set of supple tyres feels like free speed while a lighter frame does not: the tyre term is the one you are actually paying at low speed, and the aero term is the one you are paying at every speed that matters.

Squares, cubes, and why doubling your speed is not four times the work

The cliché says eight times the power for twice the speed. The cliché is exactly right about the air term and wrong about your legs. Going from 10 mph to 20 mph takes the air term from 17.5 W to 140.1 W, which is a factor of 8.000 to three decimal places — the cube of two, precisely as the formula demands. But rolling resistance only goes from 17.5 W to 35.0 W, a factor of two, and the two piles add. Total power at the pedals goes from 36.1 W to 180.5 W: five times, not eight.

The distinction matters because the ratio drifts with speed. Double from a low speed and you get well under eight, because rolling resistance is still a large share of a small number. Double from an already-fast speed and you approach eight, because the air term has crowded everything else out. The honest general statement is that the exponent you experience is somewhere between one and three, and it climbs as you go faster — which is also why the last mile per hour of a time trial is so much more expensive than the first, and why marginal gains are hunted in drag area rather than anywhere else.

The headwind asymmetry, proved

Every rider knows the out-and-back where the wind was meant to cancel out and did not. The reason is in the equation: the drag term uses your speed through the air, but the power term multiplies by your speed over the ground. A headwind therefore inflates a squared quantity while leaving the multiplier alone, and a tailwind deflates the same squared quantity by less than you would hope.

Put a letter on it and it becomes provable. Let k be the wind as a fraction of your ground speed. Into the wind the air term scales by (1+k)², with it by (1-k)². So the extra you pay is (1+k)² - 1 = 2k + k², and the saving you get is 1 - (1-k)² = 2k - k². Their ratio is (2+k)/(2-k), which is greater than one for any wind at all and grows with the wind. At k = 0.5 — a wind of half your riding speed — it is exactly 2.5 ÷ 1.5 = 5/3. The headwind costs you five-thirds of what the tailwind returns, and no amount of pedalling changes the arithmetic.

Now the version you can feel. Take our rider holding 180.5 W at the pedals, which is 20 mph on a still day, and give the day a 10 mph wind. Into it, that same 180.5 W drives only 14.39 mph. With it, the same power gives 26.61 mph. Ride 10 miles out and 10 miles back: 41.7 minutes into the wind, 22.6 minutes home, 64.2 minutes total against 60.0 minutes on a calm day. You have lost 4.2 minutes and your average is 18.68 mph instead of 20 — even though the wind was exactly balanced. Notice the trap in the raw speeds: they average 20.5 mph, above your calm speed. It is the time you spend in each that decides, and you spend far longer in the headwind.

Drafting is the cheapest speed in cycling

If most of your power buys air, then anything that reduces the air you have to move is worth more than anything else you can do. Sitting on a wheel typically cuts drag by a quarter to a third; deeper in a bunch the reduction is larger still, and wind-tunnel and CFD work on full pelotons has measured riders in the sheltered middle experiencing only a small fraction of the drag of a lone rider. Take a 30% drag reduction at 25 mph on our model rider: 327.2 W alone becomes 242.5 W in the wheel, a saving of 26% of total power for doing nothing but holding a line.

Read it the other way and it is more striking. Keep the power constant instead and that same 30% drag reduction takes you from 25 mph to 27.98 mph — nearly 3 mph for free. Halve the drag, as a rider deep in a large bunch might, and the power to hold 25 mph falls from 327.2 W to 186.1 W, a 43% cut. No position change, no equipment and no training produces anything of that size, which is why bike racing is a tactical sport rather than a time trial with a crowd.

Where the flat model stops: hills, and what CdA really is

Put a gradient under the same rider and the hierarchy inverts. At 12 mph up a 5% climb, gravity takes 209.7 W, rolling resistance 21.0 W and air only 30.3 W — 12% of the total. That is the whole reason climbing is about power-to-weight while flat riding is about power-to-drag: a term that was zero on the flat now dominates, and the aero term you spent the last section obsessing over has been squeezed into a corner. It also explains why a heavier aero wheelset is a good trade on a rolling road and a bad one on a mountain.

The last honesty is about CdA itself. It is not a specification you can look up; it is a measured property of you, your bike, your clothing and your position on a given day, and it changes when you move your hands, lift your head or put on a flapping jacket. Published figures span a wide band, from roughly 0.4 m² sitting up on the tops to well under 0.25 m² for a well-drilled rider in an aero position, and the difference between the ends of that range is worth more than any component. Every number in this article assumed 0.32; the shape of every conclusion holds regardless, but your personal crossover speed and your personal watt figures will not match ours until you have measured your own.

Frequently asked questions

Does a lighter bike make me faster on the flat?

Barely. On flat ground mass appears only in the rolling term, which is already the small one. Take 2 lb off a 176 lb rider-and-bike and the rolling power at 20 mph falls from 35.0 W to 34.6 W — a saving of about 0.4 W out of 180. The same 2 lb is worth vastly more on a climb, where mass enters the gravity term, and worth something every time you accelerate. Flat speed is bought in drag area, not in grams.

Why is my speed lower on a hot day even though I feel fine?

Warm air is thinner, so it should make you faster, not slower — that is the counterintuitive part. Air density falls by roughly 0.3% per degree Celsius of warming, so a 20-degree swing changes your aero power by around 7%. If you are slower in heat despite that gift, the cause is you rather than the air: cardiac drift, dehydration and the blood diverted to your skin for cooling all cost more than the density gains. The same logic works in reverse on a cold winter morning, when dense air quietly makes every ride harder.

How much does a crosswind cost?

More than the head-on component alone suggests, and the model in this article does not capture it. A crosswind combines with your own speed into an apparent wind that arrives at an angle, and a bicycle's drag at a yaw angle is not a simple projection — deep wheels and some frames actually produce a small forward force at moderate yaw, while a body sideways-on presents more area. The practical field observation is simpler than the physics: a crosswind is expensive because it forces you out of a straight line and off the shelter of the wheel in front.

Is the crossover speed really the same for everyone?

No, and the formula tells you exactly how it moves. It is √(Crr·m·g ÷ (½·ρ·CdA)), so it rises with the square root of mass and of rolling coefficient, and falls with the square root of CdA. A heavy touring bike on wide, soft tyres has a crossover well above ours; a light rider tucked low on fast tyres has one below it. Nobody's crossover is high enough to matter for the conclusion, though. Even doubling the rolling coefficient only multiplies the crossover by √2, and it is still a speed you pass in the first minute of a ride.

If I ride the same loop, why is my time worse on windy days even when the average wind is zero?

For the reason worked out above, plus one more. The asymmetry ratio (2+k)/(2-k) is greater than one for every non-zero wind, so any wind at all costs you on a closed loop even when its directions cancel. On top of that, the time you spend in each direction is unequal — you are in the headwind section for far longer than the tailwind section — which weights the expensive part of the loop more heavily. Both effects push the same way, which is why a windy day is slow even when the wind is fair for half of it.