The Aero Variables Simulation Can’t See: Skinsuits, Socks and Bottle Placement

A simulated position can be accurate to the degree and still miss the two things that decide whether a race day is fast: what you are wearing, and where the water is. AiRO and tools like it — a system we are currently looking into offering at UpHill — model your shape against a wind tunnel dataset well enough to guide a fit. They cannot see fabric, seams, or a bottle sitting in the wrong place.

That gap is not a criticism of simulation. It is a scope limit, and it is worth understanding before you spend money assuming a rendered position is the whole answer.

What simulation is actually modelling

AI-CFD tools take a photograph or a short video of your position and compare it against a large database of wind-tunnel-tested shapes. The output is a CdA estimate for that body position, and for rider position specifically it tracks close to real wind tunnel numbers. That part of the claim holds up.

What it is not doing is measuring your bike. It assumes a generic value for the bike itself — Josh puts the industry-standard assumption at roughly 0.03 CdA for a time trial or triathlon setup — which means your actual frame, wheels and cockpit are not in the number at all. Two riders in an identical position on completely different bikes get the same simulated result.

Kit is the same story. A simulation sees a rough outline, not fabric. It has no way to tell a smooth, well-fitted skinsuit from a baggy one, because both render as the same silhouette.

The variables nobody simulates

Four things that move real CdA, none of which show up in a photo-based model:

• Skinsuit fabric and seam placement. Two suits that look identical on a hanger can differ by more than a wheel upgrade. Seam location relative to airflow matters more than most riders expect.
• Socks and calf sleeves. Small surface area, disproportionate effect, because they sit in fast-moving air right where the leg is doing the most work.
• Overshoes. Change the airflow around the foot and ankle in ways that vary by shoe, sock and shoe cover combination.
• Bottle placement. Between the arms, behind the saddle, or on the down tube can each be faster depending on the rest of the setup. Moving a bottle has been worth more than a helmet upgrade in testing here. Sometimes riding with a bottle in the right spot is faster than riding without one at all, because the bottle is smoothing airflow that was previously separating badly off the frame.

Why this matters more than it sounds like it should

A rider who simulates a position, buys a matching skinsuit because it looked fast on someone else, and never checks bottle placement has done maybe half the available work. The position might be genuinely good. The kit and hydration setup are unverified guesses dressed up as decisions.

None of that is knowable from a photograph. It has to be measured on the actual rider, on the actual bike, wearing the actual kit, with the bottle where it will be on race day. That is the entire case for field testing existing alongside simulation rather than instead of it.

A sensible order of operations

The workflow that makes the most sense, and the one the aero testing page is written around, is simulate to narrow the search, then measure to be sure.

Start with a bike fit that establishes a position you can actually hold for the length of the effort. A fast position you abandon after twenty minutes is not fast. From there, simulation is a cheap way to explore variations — saddle height, torso angle, arm position — and narrow a wide field of options down to a shortlist without burning session time on shapes that were never going to work.

Field testing is where the shortlist gets settled, because it is the only step that measures the whole system: the position, the bike, the kit and the hydration setup, together, in real air. That is also where the skinsuit question, the bottle question and the overshoe question actually get answered, because those variables never entered the simulation in the first place.

This is the model we are currently looking into at UpHill: offering AiRO as an add-on to a regular bike fit, for riders who specifically want to optimise their aero position. The fit comes first, and identifies a handful of positions that are genuinely comfortable to hold for the length of the effort. AiRO would then run against those specific positions, to narrow down which ones are worth booking a dedicated aero field testing session for, rather than testing every option on the day. Pricing is still being finalised, but we are looking at roughly $150 for a set number of simulations.

What to bring if you are testing kit specifically

Every skinsuit or jersey you might race in, not just your favourite. Socks and calf sleeves, plain and aero versions if you own both. Overshoes. Every bottle position you are curious about — the marginal cost of testing one more configuration during a session you are already running is close to zero, so bring the option even if you think you already know the answer.

Riders are frequently wrong about which of their own kit choices is faster. That is not a criticism, it is just what happens when a decision was made by eye instead of measurement. The point of a session is not to confirm what you already believed. It is to find out.

Where this leaves simulation

Nowhere it should not be. Simulation is a genuinely useful early step, and for anyone exploring position options before committing testing budget to one or two shapes, it is worth using. It answers “what shape should I hold.” It has never claimed to answer “what is my whole setup costing me,” and it should not be read as though it does.

A worked example

Take a rider with a well-simulated position: torso angle, arm position and head position all checked off, all sitting near the wind tunnel benchmark for that shape. On paper the position work is done.

Put that same rider through field testing wearing their training jersey instead of a race skinsuit, bottle mounted between the arms out of habit rather than by design, and generic socks, and the number that comes back can easily be several watts worse than the simulated position implied. None of that is a fit problem. All of it is kit and setup, and none of it would have shown up in a photo-based model no matter how accurate the position estimate was.

This is the practical argument for testing the whole system rather than trusting a partial answer, however accurate that partial answer is within its own scope.

Want your actual setup measured, not modelled? Book an appointment at UpHill Performance.