Peloton drafting explained, and why the bunch rides as it does

A road cyclist at racing speed is not really fighting the road or the hill. On the flat, the great majority of the power going into the pedals is spent pushing air out of the way, and any account of peloton drafting has to start there, because that single fact is what gives the bunch its shape, its tactics and its accident rate.

Move the air problem and you move everything. A rider who can spend a hundred kilometres letting somebody else do that work arrives at the finish with a completely different set of legs, and every formation you see in a bike race, from the tidy line to the diagonal fan to the shapeless blob, is a solution to the same question of who is in the wind and for how long.

The rest of this is the mechanism: what is happening in the air, why the saving falls off so quickly with distance, why it is larger deep in a crowd, and what the arrangement costs in broken collarbones.

• 3Exponent linking power to speed against air
• 3Points of contact a rider may take support from
• 3Maximum depth to width ratio for a frame tube
• 6.8UCI minimum bicycle weight, in kilograms

Drag wins at racing speed because it grows faster than everything else

Three forces take a rider's power on flat ground: rolling resistance between tyre and road, friction in the drivetrain, and aerodynamic drag.

The first two are close to constant with speed. Rolling resistance depends mostly on the tyre, the pressure and the surface, and while it does rise a little as you go faster it does not rise dramatically. Drivetrain losses are a small percentage of whatever is going through the chain.

Drag is different in kind. The force from air resistance rises with the square of the speed through the air, and power is force multiplied by speed, so the power required rises with the cube. Double the speed and the power to overcome drag goes up eightfold.

That is why the balance flips so sharply. At walking pace a bicycle is a rolling-resistance problem. At the speed a professional bunch travels on the flat, it is an air problem with a small rolling-resistance correction attached. The full arithmetic of that relationship, and what a given percentage of drag reduction is actually worth, is set out in the piece on where a cyclist's watts really go, and it is the foundation under everything below.

The consequence for a group is immediate. If nearly all your power is going into the air, then anything that changes the air you are riding into changes almost all of your power demand. There is no other lever in the sport with that leverage. Losing weight helps on climbs. Better tyres help everywhere by a little. Sitting behind somebody changes the dominant term.

What is actually happening behind the wheel in front

A cyclist is what an aerodynamicist calls a bluff body: not streamlined, with the flow separating rather than following the shape all the way round. Behind a bluff body sits a wake, and the wake is where the drafting benefit lives.

Two things are true of that region and they are often confused with each other.

The air in the wake is moving. Flow separating off the rider's back and shoulders drags air along with it, so the air immediately behind is travelling forwards relative to the ground. A follower riding into it meets a lower relative wind speed than a rider in still air at the same road speed. Since drag depends on the square of the relative air speed, even a modest reduction in that speed is a substantial reduction in force.

The air in the wake is at lower pressure. Most of a cyclist's drag is pressure drag: high pressure on the front, low pressure in the separated region behind, and the difference between the two acts backwards on the body. A follower sitting in that low-pressure region has less pressure on their own front than they would in undisturbed air, which reduces their pressure drag directly.

The two effects work together and they are not the same mechanism, which matters when you start asking why the saving changes with position. The velocity deficit recovers with distance as the wake mixes with the surrounding air. The pressure effect is strongest very close to the rider ahead and falls away with a different profile. A follower is riding a combination of both, and the combination is what makes the benefit so sensitive to a few centimetres.

The saving falls away with distance, and it falls away steeply near the wheel

Ask any rider what happens when a gap opens and they will describe a cliff rather than a slope. That description is aerodynamically accurate.

The wake is not a tunnel of still air with sharp walls. It is a turbulent region that is continuously exchanging momentum with the air around it, so the deficit it carries decays as it travels backwards. Very close to the rider ahead, the following rider is in the most disturbed, slowest-moving, lowest-pressure part of it. A metre back, the wake has already mixed with a good deal of surrounding air and recovered some of its speed. Several metres back, it is measurably there and no longer worth much.

The decay is not linear, and the steepest part is at the near end. That produces the behaviour every cyclist knows: the difference between sitting on the wheel and sitting half a metre off it is far larger than the difference between half a metre and a metre. It also produces the corresponding tactical fact, which is that a rider who cannot hold the wheel is not slightly worse off than one who can. They have fallen off the steep part of the curve and their power demand has risen sharply at the exact moment their legs told them it could not.

Published figures for how much a single following rider saves vary widely between studies, and the variation is real rather than sloppy. The gap tested, the riders' positions, the equipment, whether the measurement was a wind tunnel, a track test or a simulation: all of it moves the answer. Anyone quoting a single number for "the drafting saving" is quoting one experiment's conditions as though they were a constant.

Why the middle of a big bunch is the cheapest place in the sport

Here is where the intuition built on two riders breaks down completely.

If drafting worked purely as one rider sheltering the next, then the tenth rider in a line would be in the wake of the ninth and would save roughly what the second rider saves behind the first. That is not what happens in a real bunch, and the reason is that a peloton is not a line. It is a wide, deep, densely packed crowd, and past a certain size it stops behaving like a set of individual riders and starts behaving like one large bluff body with people inside it.

Three things compound.

Rows ahead accumulate. The rider in row eight is not in the wake of row seven. They are in a flow field that has already been worked over by seven rows of riders, each of which has removed momentum from it. The deficits do not simply add, but they do build.

Neighbours block the sides. A rider in the interior of a bunch has riders to the left and right as well as in front. Air that would otherwise be flowing past their flanks is being disturbed by somebody else's shoulders. In a single line that shelter does not exist.

The group acquires its own boundary. Once the bunch is large enough, the front row and the outer edges take the pressure and the interior sits in a region that is partly enclosed. That is a completely different situation from being behind one person.

The most substantial attempt to quantify this is the CFD and wind tunnel work by Blocken and colleagues on a full-size peloton, which reported that drag on riders well inside the group can fall to a small fraction of what an isolated rider experiences, a reduction of up to roughly ninety-five per cent.

That figure is startling and it is easy to misread, so it is worth stating what it does and does not mean. It does not mean a rider in the bunch is doing five per cent of the work of a rider off the front, because rolling resistance, drivetrain losses and the fact that the bunch is going quicker than a lone rider would all remain. It does mean that the aerodynamic term, the term that dominates everything on the flat, can very nearly disappear for a rider who is well positioned in a large group.

That is the honest explanation for something viewers find baffling: how a breakaway of five committed riders working perfectly together can be reeled in by a bunch that appears to be doing nothing. The bunch is not doing nothing. It is doing very little, which is different, and the arithmetic of who catches whom is worked through properly in the tactics piece.

Where the published measurements disagree, and why both sides are right

The literature on drafting does not agree with itself, and the disagreement is instructive rather than embarrassing.

The older experimental work, going back to Kyle's wind tunnel and track studies, established the effect and produced conservative savings for riders in lines, measured at realistic gaps with the equipment of the time. Those figures are the ones that made their way into coaching manuals, and they are considerably smaller than the peloton simulation results.

The modern CFD work models arrangements the older experiments could not physically build. You cannot put a hundred and twenty riders in a wind tunnel. What you can do is simulate them, validate the simulation against wind tunnel measurements of a smaller group, and then extrapolate to the full bunch.

So the two bodies of evidence are answering different questions. One asks what a rider saves sitting on a wheel. The other asks what a rider saves buried in a crowd. Both answers are correct for their own arrangement, and the gap between them is not a contradiction but a measurement of how much the crowd itself contributes.

Two cautions apply to all of it. Simulation results depend on the assumed geometry, and a modelled peloton is tidier and more static than a real one, which shuffles constantly. Wind tunnel results depend on the model riders' positions and on the absence of a moving ground plane in some setups. Anyone treating either number as a physical constant has stopped reading the method section.

The rider on the front gains slightly, which almost nobody believes

Ask a group of cyclists whether the rider on the front gets anything out of having a wheel behind them and most will say no, and a few will say it makes them slower. Both are wrong.

The mechanism is the pressure side of the drafting effect, running backwards. A leading rider's drag comes substantially from the pressure difference between the high-pressure region on their front and the low-pressure separated region behind them. Put another body immediately into that low-pressure region and the region cannot expand and organise itself as freely. The base pressure behind the leader rises slightly. A higher pressure behind means a smaller front-to-back difference, and a smaller difference means less pressure drag.

The effect is small. It is nothing like what the follower is gaining, it does not make leading cheap, and no rider has ever asked somebody to sit on their wheel in order to go faster. But it is real, it appears consistently in the measurements, and it is a good corrective to the mental model of drafting as one rider taking something from another. Nothing is being taken. The two bodies together have a different flow field from either alone, and both are slightly better off than they would be in isolation. One of them is very much more better off.

Crosswinds move the shelter sideways, and the tarmac runs out

Everything so far has assumed the air arrives head on. It usually does not.

A moving rider experiences apparent wind, which is the vector sum of the wind over the ground and the headwind generated by their own motion. In still air the apparent wind comes straight at you. Add a wind from the side and the resultant swings round towards it, arriving at an angle to the direction of travel. That angle is the yaw angle, and it is the single number that determines where the sheltered pocket sits.

The wake goes downwind of the rider producing it, not straight backwards. So the sheltered position moves off the centre line, round to the downwind side, behind and to one shoulder of the rider ahead.

A group riding efficiently in a crosswind therefore does not form a line. It forms a diagonal, each rider stepped sideways from the one in front, fanning out across the road towards the downwind verge.

And then it hits the verge, which is where a physics problem becomes a bike race.

The echelon's capacity is set by the width of the road

A diagonal has a finite length before it runs out of tarmac. That is the whole of it. Once the fan reaches the downwind edge of the usable road, there is no sheltered position left, and every rider who has not made it into the diagonal is sitting in the gutter, in the full wind, in single file, behind a group that is sharing the work between all of them.

The arithmetic below is a constructed example with invented numbers, put in to show the shape of the constraint rather than to describe any real road.

Suppose each rider in an echelon sits about 0.7 metres to the side of the rider ahead, and that a metre of the road's width is unusable because of the verge, the camber and a rider's natural unwillingness to ride on the white line. On a road that measures six metres of tarmac, that leaves five metres of workable width, which at 0.7 metres per step accommodates seven steps, so eight riders including the one on the front. Add two metres of road and you fit three more.

5m of road6riders

6m of road8riders

7m of road9riders

8m of road11riders

A constructed illustration, not a measurement. It assumes each rider sits 0.7m laterally from the one ahead and that one metre of the road's width is unusable. Real echelons vary with camber, surface, wind angle and how brave everybody is feeling.

Show the numbers

Worked example: how many riders fit in one echelon

Item	Value
5m of road	6riders
6m of road	8riders
7m of road	9riders
8m of road	11riders

A professional bunch contains many times those numbers. So the moment a wide road turns into a crosswind, a group of a hundred and eighty riders is being asked to fit into a formation that holds perhaps a dozen, and the mathematics of who gets in is settled by who was near the front when the road turned.

Second and third echelons form behind, each of them a fresh diagonal, and each of them chasing into the same wind with fewer riders to share it. That is why the gaps rarely close. It is not a question of willingness. A group of eight sharing the wind is aerodynamically better resourced than a group of five, and the smaller group is behind by definition.

The narrower the road, the more vicious the effect, which is why teams study the width of the tarmac on exposed sections as carefully as they study the gradient of climbs.

Through and off: the rotating paceline as a machine

A line of riders taking long turns on the front is the crude version of sharing the work. The refined version is the rotating paceline, called through and off, and it is one of the most elegant things in sport.

Two lines run side by side. One is advancing, the other is drifting back. A rider reaching the front of the advancing line moves across to the head of the retreating line, eases fractionally, and lets the whole group filter past them until they reach the tail, at which point they slot back onto the advancing line and begin working forward again.

Nobody is on the front for more than a couple of seconds. That is the point. Instead of six riders each doing a minute of very hard work followed by five minutes of recovery, everybody does a continuous stream of tiny efforts separated by tiny recoveries, and the group holds a speed none of them could hold alone.

Which side the riders peel off to is not a matter of taste. The retreating line goes on the windward side and the advancing line runs in its shelter. Get it the wrong way round and the riders doing the work of moving forward are the ones in the wind, which is exactly backwards, and the rotation will fall apart within a kilometre.

1. 1Arriving at the front of the advancing lineThe rider is now in clear air and taking the full drag. This is the only genuinely expensive part of the cycle and it lasts a second or two.
2. 2Moving across to the windward sideThe rider steps sideways into the head of the retreating line without slowing much. Braking here is the most common way a rotation breaks down.
3. 3Easing, not stoppingThe rider drops their effort by a modest amount and lets the advancing line come past. Cutting the power too hard opens a gap and forces everybody behind to sprint back on.
4. 4Drifting back down the retreating lineRiding in the windward line at a slightly lower speed. This is partial recovery rather than real rest, because the retreating line is the less sheltered of the two.
5. 5Reaching the tailThe last rider of the advancing line comes past and the rider swings in behind them. Judging this moment badly leaves them behind the group instead of in it.
6. 6Working forward in the sheltered lineThe cheapest part of the cycle. The rider is in the wake of the whole advancing line and recovering as much as this format allows.
7. 7Back at the frontThe loop closes, and it closes faster than most people expect. In a fast rotation a rider is on the front again within half a minute.

The whole loop takes a few seconds. Its efficiency depends on the transitions, because moving across and easing are the moments a rider is least sheltered.

The pull length is the variable everybody argues about. Short pulls mean more transitions, and every transition costs something because moving across the group is itself a poorly sheltered position and an acceleration. Long pulls mean fewer transitions and more time spent in clear air. There is an optimum somewhere in between and it moves with the wind, the road and how tired everybody is, which is why a rotation that is running beautifully at kilometre ten is ragged at kilometre eighty.

The failure mode is always the same. One rider takes a pull that is a fraction too hard, the rider behind has to accelerate to hold the wheel, that acceleration propagates, and within a lap of the rotation everybody is riding harder than the group's average pace requires. The rotation stops being a machine for sharing work and becomes a machine for wasting it.

Why a team time trial cannot rotate for ever

If through and off is so efficient, the obvious question is why a team time trial does not simply rotate at full gas from the start to the finish and produce an unbounded speed.

Three limits close in, and they close at different rates.

The turns are above threshold and the recovery is not below it. A rider on the front of a fast rotation is producing power well above what they could sustain, and the recovery in the sheltered line is not full recovery. Each cycle therefore leaves a small deficit, and the deficits accumulate. The rotation does not fail suddenly. It fails as a slow accumulation that arrives all at once in one rider's legs.

The group can only travel at the pace of the rider currently on the front. A rotation that includes somebody who is struggling either slows to their pull or drops them. There is no third option, and this is why teams shed riders deliberately rather than accidentally. The team time trial's arithmetic and its own set of rules turn on the result being taken on a specified rider crossing the line, which means the correct plan is often to burn several riders completely and finish with the minimum number required.

Transitions cost more as everybody tires. A tired rider brakes when moving across, eases too much, and comes back on late. The rotation loses smoothness exactly when smoothness is what it is running on, and once the accelerations creep in, the whole thing gets more expensive per kilometre travelled.

So the correct plan is not a constant rotation. It is a rotation whose shape changes as the ride goes on: even turns early, shortening turns as fatigue builds, riders removing themselves from the rotation before they blow rather than after, and the strongest few taking longer pulls at the end when there is nobody left to hide behind. Judging when a rider should pull out of the rotation for good is the hardest call in the discipline, and getting it wrong by thirty seconds costs more than any equipment choice.

Drafting is why a bunch crashes the way it does

The same closeness that makes the bunch cheap makes it lethal, and the mechanism is worth understanding properly because the popular explanation, that riders are reckless, explains almost none of it.

Start with the fundamental problem: a rider in a bunch is sitting far closer to the wheel in front than their own stopping distance. That is not carelessness, it is the entire point. The saving lives in the near part of the wake, so a rider who leaves themselves room to stop has given up most of the benefit of being there. Every rider in every professional bunch is knowingly riding inside their own braking distance, all day, and the sport is built on it.

Then add that they cannot see. A rider in the middle of a bunch has no view of the road surface, no view of the corner, and no view of anything except the backs and wheels immediately around them. Information about a pothole, a slowing rider or a change of direction reaches them through the movements of other people rather than through their own eyes.

Now the amplification. When the front of the bunch slows, the news travels backwards as a wave, and it does not travel cleanly. Each rider reacts a fraction of a second late, so each has slightly less room than the rider in front had, and so each brakes slightly harder. The disturbance grows as it propagates. A gentle deceleration at the front of a hundred and eighty riders can arrive at the back as a genuine emergency stop, which is why crashes so often happen well behind whatever caused them and why the riders involved never saw the cause.

Then the overlap. In a packed bunch, riders' front wheels are frequently alongside the rear wheels of riders beside and ahead of them. A front wheel that touches the side of a rear wheel is almost always unrecoverable: the contact deflects the front wheel sideways, the steering goes with it, and the rider is on the ground before any correction is possible. The rider ahead usually stays upright and often does not know it happened. This is why the phrase "he took him down" is so frequently unfair.

And finally the crosswind case, which combines all of it. Riders in the gutter fighting for a wheel are travelling at high speed, with no shelter, on the edge of the road, in a line, at the limit of their capacity, watching the wheel ahead rather than the surface. Every ingredient is present at once.

None of this is fixable while the sport keeps drafting, because the closeness is the sport. What is adjustable is the behaviour at the margins, which is where the governing body has been spending its rule-making effort, including a card-based sanctioning system for dangerous riding brought in across professional road racing from the 2025 season and aimed squarely at riders who deviate from their line in a sprint or push others towards the barriers.

The rules that stop aerodynamics being pushed any further

If sitting in still air is worth this much, the obvious engineering answer is to make riders more aerodynamic and let them get closer. The UCI has spent a century preventing exactly that, and the position rules are where the prevention is most visible.

The framework rests on a few principles. A rider must be supported at three points and no more: the feet on the pedals, the hands on the handlebars, and the seat on the saddle. Fairings and anything whose purpose is to reduce air resistance by adding non-structural material are prohibited. Frame tubes must fall within a depth-to-width ratio, which caps how aerofoil-shaped a bicycle can become. A complete bike must weigh at least 6.8 kilograms. There are limits on saddle position relative to the bottom bracket and on how far forward the extensions of a time trial bar may reach.

Then there are the positions that were banned outright on safety grounds rather than for equality of equipment. From April 2021 the UCI prohibited the supertuck, in which a rider sits on the top tube with their body folded over the front of the bike, and the practice of resting the forearms on the handlebars while riding in a road bunch. Both are noticeably faster. Both put the rider's weight in a place from which the brakes cannot properly be reached and from which a small disturbance is uncorrectable.

That pair of bans is the clearest statement the sport has made about the trade it is willing to accept. The most aerodynamic positions available to a rider on a road bike are also the ones from which they are least able to control the bicycle, and a bunch is the environment in which the ability to control a bicycle matters most. Given a straight choice between faster and steerable, the rules have chosen steerable, and the equipment regulations that grew out of the same instinct run to a great many pages for the same reason.

The rules are not neutral in their effects. They protect the visual identity of the sport, they hold back manufacturers who could produce something quicker, and they push the remaining gains towards clothing, helmets and rider position within the permitted envelope. What they do not do, and cannot do, is change the underlying fact that the cheapest place to be on a bicycle is behind somebody else.

What to watch for in a bike race

Watch the shape of the bunch, not the front of it. A single line means high speed, wind, or somebody making it hard. A wide blob means the pace is easy and everybody is finding shelter without effort. A diagonal fan means a crosswind, and the race is probably about to break.

Watch which side riders peel off to in a rotation. They should be pulling off into the wind. If the rotation is going the wrong way round for the conditions it will not last, and a team that gets it right in a crosswind will gain time on one that does not.

Watch the gap between the last rider of an echelon and the road's edge. That gap is the number of places left. When it closes, the race has just been decided for everybody behind, and it will be several minutes before the television realises.

Watch what happens after a bunch slows. The crash is usually not at the front. The wave that started there arrives twenty places back, magnified, several seconds later, which is the reason a rider who was nowhere near the incident ends up in the ditch.

Watch how long the turns are getting. Shortening pulls in a break or a team time trial are the most reliable sign of fatigue in the sport, and they show up long before anybody's face does. The power numbers behind them are readable after the fact, and the shortening turns tell you the same story live.

The rest of the archive on how bike racing works, and on the physics it is built out of, sits in the cycling section. Almost everything in it comes back to the same sentence: the air is the opponent, and a bike race is a two hundred kilometre argument about who has to face it.