What would Pogačar’s Rider Model look like?

This experiment runs two connected calculations around one rider whose state persists through the entire performance.

The mechanical model calculates the power required to move the rider and bicycle through the current section under the declared physical assumptions. The physiological model checks that demand against the rider’s current state and determines how much power is actually available.

Applied power returns to the mechanical calculation to resolve speed. Speed advances time and distance, while the physiological cost becomes the starting state for the next step. A power curve describes what a rider could produce from a defined starting state. A race keeps changing that state.

The reconstruction uses quasi-steady-state mechanics: at each replay step, it resolves speed from the applied power and the current section of road without reconstructing acceleration or inertia. Physiological state, elapsed time and distance advance in fixed 0.1 s steps.

Requested power and applied power are deliberately different quantities. A pacing plan can ask for an effort. The rider model answers using the rider’s current W′ balance and effective CP. If the request is no longer feasible, the engine applies the feasible power and records what limited it. Physics never receives imaginary watts simply because a target demanded them.

Consider one ordinary race transition. A teammate finishes a pull and the rider becomes exposed. The rider is not recreated. Only the aerodynamic context changes. At the same speed, required power rises. The model evaluates that new demand against the state carried out of the draft. If the effort remains feasible, W′ drains faster; if it does not, applied power and therefore speed must change. That result becomes the starting point for the next 0.1 s step.

This is why the experiment is more than five independent climb-time estimates. One CP/W′ candidate has to produce coherent state trajectories through five different combinations of duration, gradient, shelter, wind and race context. The fit is looking for a rider that survives all of those executions, not five unrelated average-watt answers.

We started with the parts of each performance that could be tied to public evidence: the route boundary, elapsed time, GPX geometry and the broad race phases visible in footage or described in contemporary reports. Those phases tell the engine whether Pogačar was setting the pace, following a wheel or moving inside a compact group. They are reconstruction boundaries, not second-by-second tracking data.

Race-day body mass was not measured. The frozen scenarios used 64.5–66 kg for the Grand Tour efforts, 65–67 kg for the Classics, and 66.5–69 kg for the December Coll de Rates ride. Weather, equipment and the exact benefit of a wheel also remained uncertain, so the experiment carried them as declared ranges rather than facts about Pogačar.

Weather was carried as an observation-specific environment scenario: reanalysis where the timing supported it, and a declared neutral-wind case where it did not. The engine resolved air density along the elevation profile and projected only the longitudinal wind component onto the local direction of each section. Crosswind and yaw-dependent CdA are not modeled in this experiment.

A regional weather cell can report atmospheric wind without telling us how much of it reached the rider behind buildings, vegetation and terrain. Gritmo therefore keeps environmental exposure and shelter from other riders as separate inputs. One cannot silently compensate for uncertainty in the other.

The research below supports the underlying concepts. It does not specify or validate Gritmo’s exact state representation or the reconstructed position of Pogačar inside these races.

The route is divided into sections with their own distance, gradient, direction, weather and aerodynamic context. For each section, Gritmo resolves the road-load forces.

This road-load structure follows established cycling mechanics. Martin and colleagues compared a mathematical cycling-power model with measured road power, including aerodynamic drag, air velocity, gradient and rolling resistance.

The reconstruction does not recover acceleration or inertial transients from the public observations. Gritmo’s full race simulation uses the same road-load equations inside a time-step calculation that also follows acceleration and inertia.

The drafting policy uses 1.00 for exposed riding, 0.68 as the lower aerodynamic-drag bound for one close direct wheel, and 0.62 as the absolute floor when additional riders provide the maximum shelter allowed by the model.

These are scenario bounds. They were not measured for Pogačar, and they were not copied as exact coefficients from one paper.

At 40 km/h in still air, with CdA 0.30 m² and air density 1.2 kg/m³, the exposed aerodynamic wheel-power term is approximately 247 W. Applying the 0.68 direct-wheel scenario reduces that aerodynamic term to approximately 168 W. Gravity and rolling resistance do not receive the same reduction.

Van Druenen and Blocken combined wind-tunnel measurements with CFD for uphill pacelines. In their tested configuration, following riders experienced large reductions in aerodynamic drag; at 7.5% and 6 m/s, those reductions translated into roughly 7–10% less total required power, depending on position. The study supports the plausibility and scale of uphill drafting. It does not reveal Pogačar’s actual multiplier in any of these performances.

Each candidate carried the same two capability parameters through every relevant transition: Critical Power, the modeled boundary above which sustained work draws on a finite reserve, and W′, the finite modeled work available above CP.

Normally the engine runs forward: start with a Rider Model and calculate the performance it can produce. Here the observed performance sits on the other side of the problem. The experiment compares the simulation with the public result, then repeats with another candidate Rider Model. Depending on the performance, elapsed time and independently supported checkpoints remain outputs to explain or explicit constraints declared in advance.

The search changes the candidate rider. It does not quietly move the route, the clock or the physical assumptions until a preferred answer appears.

Published estimates of Pogačar’s watts stayed outside the fit while CP and W′ were narrowed. They became independent context only after the Rider Model family was frozen.

The representative result, 436 W CP / 38 kJ W′, sits inside the surviving family. It is not a direct physiological measurement. The experiment supports a bounded family—approximately 432 W / 39 kJ to 439 W / 37 kJ under the frozen assumptions—not one magic answer.

From a wheel to exposed air, through an attack and into another race, the physical demand changes and the physiological state evolves. The Rider Model does not reset at the transition.

One candidate therefore has to remain coherent through all five performances. That continuity is what the experiment tests.