What the Zwift Draft Is Worth, and Where It Stops Paying

Everyone on Zwift knows the draft saves watts. Almost nobody can say how many, and the numbers that get repeated in forums are usually somebody's guess wearing a lab coat.

Zwift has never published the equations behind its draft. There is no released drag coefficient, no documented distance curve, no group-size multiplier you can look up. So this article does the one thing that is still honest: it works the problem from physics that does not depend on Zwift's source code. Air resistance in Zwift behaves like air resistance on a road in the two ways this article leans on. Going faster costs steeply more, and sitting behind someone costs less. Both are visible from the saddle without knowing a single line of the game's code.

What follows is reasoning and arithmetic, not telemetry. Every number below is either a published fact about a route, a value from this site's own Alpe du Zwift fit, or the output of a physics model whose inputs I state in full so you can redo it yourself. Where I use somebody else's figure, I say whose it is.

What the draft actually does

A rider moving at 40 km/h is pushing a column of air out of the way. That air does not simply step aside and stop. It gets dragged along behind the rider in a turbulent wake, still moving forwards at some fraction of the rider's own speed.

Put a second rider in that wake and something useful happens. The air they are riding into is already travelling in their direction of travel, so the relative wind hitting them is slower than their ground speed. Drag depends on the speed of the air relative to the rider, not the speed of the rider over the ground. Slower relative air means a smaller pressure difference between the front and back of the rider, and a smaller pressure difference means less force pushing them backwards.

In the language of the standard cycling power model, the follower behaves as though they have a smaller drag area. Their real body has not changed shape. The air around it has. That is the whole mechanism, and it is why the benefit collapses the moment you drop out of the wake: get far enough back and the disturbed air has slowed, spread and mixed back into still air.

Why speed decides everything

Aerodynamic drag force follows the classic equation:

The force rises with the square of speed. The power needed to overcome it rises with the cube. That single exponent is the reason drafting matters enormously in a flat race and barely registers on the Alpe, and it explains more about Zwift tactics than any amount of pack-position folklore.

Doubling your speed multiplies your aerodynamic power demand by eight. Adding 10% to your speed costs 33% more aerodynamic power. Every acceleration in a fast group is expensive out of all proportion to how big it looks on the screen.

The model I am using

To turn shares into watts I need numbers. Here are mine, stated openly so you can disagree with them and rerun the arithmetic:

• Rider plus bike plus kit: 78 kg
• CdA: 0.32 m², a typical value for a rider on the hoods and the kind of figure standard cycling power models use
• Rolling resistance coefficient: 0.004
• Air density: 1.225 kg/m³
• Drivetrain losses: ignored

These are conventional assumptions, not measurements of your setup or of Zwift's internals. Zwift's own rolling resistance and drag values are not published, so treat every total below as a ballpark. The ratios are the point, and ratios are far more robust than absolute watts.

What 25% looks like in watts

Take flat ground at 40 km/h, which is 11.11 m/s. With the numbers above:

• Fighting the air: 0.5 × 1.225 × 0.32 × 11.11³ = about 269 W
• Rolling resistance: 0.004 × 78 × 9.81 × 11.11 = about 34 W
• Total: about 303 W, of which roughly 89% is going into the air

Now apply the community's 25%. Holding 40 km/h on a wheel would cost about 227 W instead of 303 W, a saving of about 76 W. Rolling resistance does not change, because your weight and speed have not changed, so the entire saving has to come out of the aerodynamic term: 269 W falls to about 193 W. That is a 28% cut in the aerodynamic term.

Hold on to that 28%. It is the community's 25% translated into the only place a draft can possibly act, and it is what I apply to every gradient below.

Flip the question round and it gets more vivid. Keep pushing the full 303 W while sitting in that draft and the same model puts you at roughly 44.5 km/h instead of 40. Four and a half kilometres per hour, for free, for doing nothing except staying on a wheel.

The gradient ladder: where the draft stops paying

Gravity does not care how fast the air is moving. The power you spend lifting yourself up a hill is mass times gravity times gradient times speed, and no rider in front of you can reduce any of those four terms. So as the road tilts up, two things happen at once: the gravity bill grows, and your speed falls, which shrinks the aerodynamic bill by the cube.

Here is the same model at four points on the same bike, with the 28% aerodynamic reduction applied at each one:

Read that ladder from top to bottom and the entire tactical picture of Zwift racing falls out of it. On the flat the draft is the single largest lever available to you, bigger than any training block you could do in a month. On a serious climb it is a rounding error.

Why Alpe du Zwift is not a drafting race

The bottom rung of that ladder is not an arbitrary example. Alpe du Zwift is 12.2 km with 1,036 m of gain across its 21 hairpins, which is an average of roughly 8.5%. This site's own fit for the Alpe is:

At 3.2 W/kg that returns 59:58, which is where the famous sub-hour target comes from. Covering 12.2 km in an hour means an average of 12.2 km/h, or 3.39 m/s, and that is the speed I fed into the bottom rung above. At 3.39 m/s the cube law has done its work: the aerodynamic term has shrunk to around 8 W. Even a perfect draft, one that removed the air entirely, could not save you more than those 8 W.

Ven-Top tells the same story. The timed Ventoux KOM segment is about 19 km with 1,481 m of gain, an average of roughly 7.8%, and the full route is about 20.9 km and 1,534 m. Slightly shallower than the Alpe, slightly slower overall because it is longer, and still comfortably inside the zone where power-to-weight decides the result and the wheel in front of you is decoration.

One nuance worth having. Gravity power scales with your total mass while aerodynamic power does not, so a lighter rider climbing at the same speed spends a slightly larger share of their power on the air. Slightly larger than 3% is still tiny. The conclusion does not move.

What Zwift does not tell us, and what people invent to fill the gap

If you have read other drafting guides you have probably seen a tidy table of drag coefficients: 0.9 solo, 0.55 in the draft, a specific number for each distance behind the wheel. Those tables are fabrications. Zwift has not released a drag coefficient, and no one outside the company is in a position to read one off the game.

Here is what can be said honestly about Zwift's draft, separated by how confident anyone can reasonably be.

Things you can see for yourself

• The draft is modelled, and the game tells you when you are in it. The effect is not subtle at racing speed.
• The range is short. Let a bike length or two open up on the flat and the benefit falls away quickly.
• Power-ups change the aerodynamics for a fixed window. The aero helmet cuts your drag, the draft van boosts the draft you receive, and the burrito denies the draft to riders behind you. Zwift shows the duration on screen. I am deliberately not quoting percentages for these, because the numbers that circulate are community estimates rather than published values.
• The Drop Shop rates frames and wheels for aerodynamics as well as weight, so equipment choice does move your effective drag area in game. The underlying values are not published either.

Things riders report consistently

The most discussed one is the sticky draft: the sense that the game holds you on a wheel and that coming past a rider takes a burst of extra effort rather than a smooth overlap. Enough riders describe it in the same terms that I take it seriously as a description of how the game feels. I have no measurement of it, and neither does anyone quoting a percentage for it.

In my own riding, the thing I notice most is not the size of the saving but the sharpness of its edge. Sitting in feels almost free. Half a bike length too far back and the power needed to hold the same speed climbs immediately. That edge is what you are really managing in a Zwift group ride.

Things that simply are not in the game

Crosswind is the big one. Zwift routes do not have a wind direction blowing across the road, which means the echelon, the single most important formation in outdoor crosswind racing, has nothing to form against. Any guide telling you to sit at a 30 degree angle to the rider ahead in Zwift is describing a manoeuvre the game does not simulate. Riding alongside someone rather than behind them is not an echelon. It is just a worse draft.

Measure your own draft saving

I would rather hand you a method than a statistic I do not have. This takes about twenty minutes and gives you a number for your own weight, your own equipment and your own riding position in the pack.

Be honest about what this gives you. A group ride's speed wanders, your position in the wake wanders, and three minutes is a short sample. You will get a rough estimate for one particular situation, not a constant. That is precisely why I am not publishing a table of my own results as though it were research.

If you want to see what a properly bounded dataset looks like and what it can and cannot support, this site's regression methodology write-up explains the Alpe and Ven-Top models, and the climbing analysis covers what those finishing times do tell us about pacing.

Race strategy that follows from the arithmetic

Everything below is derived from the ladder above. None of it needs race telemetry, because the conclusions are forced by the physics.

Time on the front is the entire cost

At 40 km/h the rider on the front pays about 303 W and everyone tucked in behind pays about 227 W for the identical speed. Your average power for an hour is therefore decided almost entirely by what fraction of that hour you spend in clean air.

Between never pulling and taking a quarter of the work there is about 19 W, roughly 8% of your hour, for the same finishing time. That is the real trade. Nineteen watts is not a disaster, and the rider who refuses to ever take a turn tends to find the group's speed drops or the group's patience does.

The price of a five-second gap

This is the calculation that should change how you ride. At 40 km/h, five seconds of gap is about 55 metres of road. To close 55 metres in 30 seconds you need to average about 1.85 m/s more than the group, which is around 46.7 km/h. In clean air, with the same model, that costs about 465 W.

Try to shut it faster and it gets much worse, because halving the time doubles the speed you need to find on top of the group's and the power bill follows the cube. Closing the same 55 metres in 15 seconds needs roughly 53 km/h and about 680 W.

Compare that with the 227 W you would have been paying had the gap never opened, and note that you arrive back on the wheel with no recovery and a group that has not slowed down. Two or three of those efforts is a race. This is the mechanism behind getting dropped: not a lack of fitness, a lack of attention for four seconds on a corner exit or the top of a roller.

Attack where the draft is worth least

The ladder tells you where a gap is cheap to open and expensive to chase. On the flat, a chasing group that shares the work pays something near the 246 W of the quarter-turns line above, while the rider off the front pays the full 303 W to hold the same speed. That gap in the bill is why lone flat attacks against an organised group usually fail. On a 6 to 8% gradient the chasers have almost no aerodynamic term left to share, so a gap opened there has to be closed with raw power-to-weight and cooperation buys them very little.

The practical version: if you are the stronger climber, attack on the climb, where the group cannot cooperate its way back to you. If you are the stronger sprinter, stay in the draft, pay the 227 W instead of the 303 W all day, and spend the difference in the last 30 seconds.

Position before the climb, not on it

Because the draft nearly vanishes as the road goes up, the group stretches and splits within seconds of the gradient arriving. If you hit the bottom at the back, you have to close gaps at exactly the moment when closing gaps is most expensive per second, since there is no wheel to recover behind afterwards. Move up while the moving-up is cheap, which is while the group is still fast and you are still in shelter.

The mistakes that cost the most watts

Yo-yoing off the back

The single most expensive habit in Zwift group riding is drifting back a few metres, then surging to reattach, over and over. Each surge is a trip up the cube law: the reattachment speed is higher than the group speed, and the power needed climbs disproportionately.

Put numbers on it. Ten reattachments of ten seconds each at 100 W above your draft power is about 10 kJ of extra work. A steady two-minute turn on the front, at the 76 W above draft power the flat example gives, is about 9 kJ. The yo-yoing costs you slightly more total energy and delivers it in the form your legs cope with worst, in repeated hard spikes with no recovery between them. Smooth power that keeps the gap constant is nearly always cheaper than reactive power that keeps closing it.

Easing off the instant you get in the draft

The draft reduces the power required to hold a speed. It does not hold the speed for you. If you drop your power by more than the draft is worth, you decelerate relative to the group and the gap opens. On the flat you have roughly 76 W of room. Take 120 W off and you are going backwards.

Sitting on the front by accident

In a large Zwift group it is entirely possible to spend ten minutes on the front without deciding to. The cost is not dramatic in any single minute, but it is the difference between a 227 W hour and a 303 W hour, and it usually gets discovered at the worst possible moment. Check where you are as often as you check your power.

Chasing an aerodynamic advantage on a climb

Everything in the bottom rung of the ladder applies here. If the aerodynamic term is 3% of your power, then equipment choices, positioning and drafting are all fighting over a sliver of 8 W. On a climb, weight and sustained power are the whole game. Use the Alpe calculator to see how much a change in W/kg is worth by comparison, and you will never think about draft on the Alpe again.

Training that makes the draft cheap to hold

Holding a wheel in a fast group is not a steady-state skill. It is a sequence of small, quick power corrections, and the riders who do it cheaply are the ones who can make a correction early and small rather than late and enormous.

Practise the correction, not just the effort

Group rides are the obvious training ground, but ride them with a specific job rather than just hanging on. Pick one rider and try to keep your gap to them visually identical for ten minutes. Anticipate the accelerations that come at the bottom of every descent and the top of every roller, and start responding half a second before you have to rather than half a second after.

Turn ERG mode off for this one. ERG smooths out precisely the responsiveness you are trying to build, and it removes the small gear and cadence decisions that make a real surge cheap or expensive.

Equipment: what actually matters for holding a wheel

The bike in your living room has no aerodynamics as far as Zwift is concerned. Your drag in game comes from your avatar and whichever Drop Shop frame and wheels you have equipped, never from the machine you are actually sitting on. What your physical setup does affect is how quickly your avatar responds to you. A direct-drive trainer with a fast power response makes small corrections land quickly. A wheel-on trainer with more slip and more smoothing puts a delay between your legs and the screen, and a delay is what turns a small correction into a large one.

The same goes for power source. If your power reading is noisy, you cannot make small corrections, because you cannot see the effect of the last one. None of this changes the physics. It changes how precisely you can operate inside it.

What I still cannot tell you

It is worth being blunt about the edges of this article. This site's dataset is roughly 500 verified Alpe du Zwift finishing times and 300 verified Ven-Top finishing times from ZwiftPower. Each record is a power figure paired with a completion time. That is enough to fit the climbing models this site publishes and nothing else.

It contains no group rides, no pack positions, no distance-behind-wheel data and no race telemetry. So there are real questions here I cannot answer:

• How exactly the saving decays with distance behind the rider ahead in Zwift.
• How much extra a second, third or fourth rider in front of you is worth.
• Whether the sticky draft is a distinct mechanic or just how the draft curve feels near its edge.
• What the actual drag values behind Zwift's frames and wheels are.

Answering any of those properly would need controlled runs at fixed speeds with logged positions, repeated enough times to average out the noise. Until someone does that and publishes the method, the honest answer is that the physics gives you the shape of the answer and the exact coefficients are unknown.

The summary worth remembering

The draft is not a trick or a game mechanic to be exploited. It is one term in a power equation, and the size of that term is decided by how fast you are going.

• Aerodynamic power rises with the cube of speed, so the faster you are, the more the draft is worth.
• On flat ground at 40 km/h, roughly 89% of your power is going into the air, and the community's working figure is that a wheel saves about 25% of your total.
• At 4% and 20 km/h that saving is down to about 4% of your power. On the Alpe's 8.5% average it is under 1%.
• Gaps are astonishingly expensive to close. Five seconds at 40 km/h is a 465 W half-minute.
• Nobody outside Zwift knows the coefficients, and any guide that hands you a precise drag coefficient for the game has invented it.

Get the flat sections right by staying sheltered and never letting a gap open, and get the climbs right by improving the number the climbs actually care about. Those are two different problems with two different solutions, and the drag equation tells you exactly where the boundary between them sits.