In one study a V-shape looks more favourable; in another, a line or more complex formation performs better. Each picture shows an orderly group of runners, yet the percentages do not form one certain ranking. Differences often begin with the question itself: the geometry, body, airflow or objective studied was not the same.

A formation’s name is only a short label. The actual system includes body spacing, size and pose, the designated target runner’s position and how flow is represented. We examine how to compare these results meaningfully. Alongside original geometry and percentage examples, a published scientific debate shows why methods are more interesting than the largest savings figure alone.

The name V-shape does not fix the spatial arrangement

Two drawings may both show a V, but one has much longer arms and a differently positioned target runner. The angle can match while every distance doubles. The letters look similar on paper, but the flow problem differs: space behind and between bodies occupies different positions. A name cannot replace coordinates or a clearly dimensioned diagram.

The V’s orientation is not self-evident either. One drawing’s point faces the running direction, the other’s points backwards, while text may briefly call both a V. The target runner may be at the point, between the arms or at a separate position farther back. Find the airflow input direction and forward-motion arrow together. Imagining a shape from its name can easily reverse the picture.

Diamond and single file can likewise cover several versions. The number of helpers, whether someone is behind the target runner and whether lateral participants share the same crosswise row all matter. Group size alone is not protection: more bodies can create more interactions. Preserve the full spatial arrangement for comparable results.

The objective changes what we call optimal

If a formation is designed to protect one main runner, others’ drag may not be the primary objective. If everyone wants the most favourable shared progress, the whole group’s functioning matters. The two tasks can yield different best arrangements. Find the objective function: whose drag or power demand were researchers trying to reduce?

The 2025 model by Fernandes and colleagues included pacers’ requirements alongside the designated runner. Formation changes and helper availability became part of an estimated race strategy. The study linked calculations to earlier racing situations and energetic assumptions; not every alternative was run directly as a controlled marathon experiment.

Minimising drag at a given speed is different from determining how long helpers can stay in the task. The main runner’s most favourable instantaneous position is not necessarily the best solution overall if others’ demands change. A model needs some way to handle this. A result can still be useful if it does not, but answers a narrower question.

A runner completes a controlled track section while a coach observes the agreed pacing practice.Enlarge image
An editorial illustration created using AI, about focused race-pace practice; not a documentary photograph or a performance prediction.

The body and the flow between runners shape results together

The CFD study by Schickhofer and Hanson compared a specific female runner model and several formations at defined speeds. Side-positioned helpers could create unfavourable local flow for the main runner in some configurations. This importantly illustrates that a group is not simply a wall against air. Flow changes between bodies, and geometry determines shelter.

The work by Beaumont and colleagues on line positions pursued another arrangement question: changing the target runner’s place within a line. One study sought favourable conditions between formations, the other between positions in a specified line. Their shared word drafting does not make the questions identical. Results need to connect with their own controls.

Body shape and pose also belong to a model. Shoulder width and trunk and limb positions can present different geometry to the air. Replacing the target-runner model changes not merely the simulation’s ‘packaging’ but a geometric boundary condition. Do not rank percentages obtained with different bodies as though the same runner had been placed in several identically defined arrangements.

Wind tunnels and CFD approach the same question differently

CFD solves flow equations on a specified spatial mesh with particular settings. A wind tunnel measures real air interacting with a physical model. Both require checks: mesh and model choices in computation, for example, or scale, measurement arrangement and equipment effects in experiments. A method’s name alone does not remove the need to examine accuracy.

The 2023 experiment by Marro and colleagues used one-tenth-scale runner mannequins with adjustable postures to investigate simple and complex formations in a wind tunnel. Different formations produced different drag reductions, while possible time benefits were estimated using further relationships. Matching a small model to a full person is a separate methodological question: flow similarity needs examination alongside geometric similarity.

An original scaling example helps explain this. One flow-similarity parameter contains speed multiplied by characteristic size, divided by the same fluid’s viscosity. Reducing body size to one tenth would in principle require ten times the speed for the same parameter in an unchanged fluid. This does not say every experiment must operate exactly that way; it shows why simply making a smaller mannequin is insufficient.

Real matching is more complex. Researchers need to justify how sensitive the phenomenon is to changes in the similarity parameter within the relevant range. Mannequin supports, test-section size and proximity to the ground can add conditions. A good study reports these rather than showing only the best formation’s photograph. Understandable methods help readers judge transferability.

A real debate: criticism and the authors’ reply

Regarding the 2023 wind-tunnel work, Blocken published a scientific commentary in 2024. Among other issues, it questioned how body models and the experimental arrangement were matched and how differing formation results were interpreted. This was not a new complete physiological running experiment, but a critique of the original methods and conclusions. Read its claims in that genre.

In their invited response in 2025, the authors addressed the points, discussing model geometry, static-mannequin limitations and the possible role of positioning differences. They also called for further investigation. The debate shows that differing results under the same formation name involve specific methodological questions; our article does not declare a definitive universal winner from either side.

This is not a reason for readers to dismiss all flow results as uncertain. Scientific debate creates more precise questions: how was the body defined, how sensitive is the result to small position changes and which control was used? Disagreement becomes useful when it identifies the next check. New data help when they answer those questions too.

Read percentages alongside their own reference points

In our fictional example, one model gives a solo runner five newtons of drag and a formation value of two: a sixty per cent reduction. Another has a ten-newton solo control and a formation value of three: seventy per cent. The second percentage is larger, yet so is its resulting absolute force. This cannot directly decide the personal benefit of the two shapes, because their starting models differ.

A percentage describes change relative to its own control. Different bodies, airspeeds or force calculations also change the denominator. Those would first need a common basis for a genuinely shared comparison. ‘Seventy is greater than sixty’ is mathematically true, but scientifically insufficient for a universal formation ranking.

Energy and time outcomes add another layer. Separate models may connect reduced drag to metabolic cost and then possible speed. Two studies can differ here even with similar force reductions. One may hold estimated power constant, another use a different constraint, producing different time gains. Name the transfer relationship as carefully as the original flow geometry.

Trail runners look at a map at the shared meeting point between short loops and discuss the next section.Enlarge image
An editorial illustration created using AI, about a trail day organised around several short loops; not a documentary photograph.

Sensitivity is often more interesting than one best point

A formation’s most favourable state may be sensitive to small lateral or lengthwise shifts. If analysis gives only one exact coordinate set, we do not yet know how results change around it. Small positional changes occur continually while running. Look for results showing neighbouring settings’ behaviour as well as the minimum.

In our conceptual example, two solutions give the same minimum drag. For one, small deviations sharply worsen the figure; for the other, it stays similar across a broader range. Minimum values hide this distinction, yet usability may depend on it. This does not assign such properties to specific running formations; it illustrates a further question for an optimised result.

Airflow direction, runner movement and changing group size create more sensitivity questions. A result assuming fixed posture and constant headwind does not automatically examine them all. Conclusions remain precise when they preserve these conditions. A real training team must adapt movement to available space, its shared task and predictability.

Six questions for a useful comparison

First: is the same target runner and number of helpers used? Second: do distances, lateral offsets and directions match? Third: which body models and treatment of movement were selected? Fourth: how was CFD or the wind-tunnel arrangement checked? Fifth: what is the percentage’s control? Sixth: are we reading direct force or an estimate further derived through energy and time?

These are not questions only for researchers. They help readers of running-portal diagrams or product promises spot where an inference jumps too far. A drawing containing only a formation name and large percentage lacks necessary information. Clearly stated conditions make the result narrower but much more meaningful.

V, diamond and single-file models can therefore produce differing results because their names conceal different physical systems and research goals. Geometry, body model, flow conditions and conversions jointly determine the figure. The best reader’s choice is comparing conditions rather than automatically declaring the largest percentage the winner. Formation research then provides real information about running without promising ready-made universal tactics.

Sources

Helpful? Pass it on.