This game needs WebGL2, which this browser did not provide.
Pick a hill, then press Space to start the inrun.
Attack
Ski V
Body–ski
A jumper is an aerofoil — six real FIS hills, at their certificate dimensions
This game needs WebGL2, which this browser did not provide.
Pick a hill, then press Space to start the inrun.
An independent, from-scratch reimplementation of ski jumping, built on the FIS International Ski Competition Rules, Book III (edition June 2026), the FIS Jumping Hills Construction Norm 2018, and the FIS Certificate of Jumping Hill of each of the six hills it contains. It is not affiliated with, endorsed by or connected to FIS or any member association, and no code, art or data from any other ski jumping game is used.
Ski jumping has been contested since Ole Rye’s 9.5 m at Eidsberg, Norway, in 1808, and at every Winter Olympics since the first, at Chamonix in 1924. The V-style that this app models was introduced by Jan Bokłöv of Sweden in the mid-1980s and swept the sport within a decade.
A ball, a javelin or a sled is a projectile or a body on a surface. A ski jumper is neither: the athlete and the two skis together are an aerofoil, and every metre after the takeoff is produced by lift the pilot is continuously controlling. So the forces here are not chosen, they are measured — from Walter Müller’s wind tunnel, whose fits he published in the open in his CERN lecture The physics of ski jumping (CERN‑2006‑014, pp. 269–278):
lift = ½ ρ v² · lA drag = ½ ρ v² · dA
where lA and dA are lift and drag areas in m² — the form the literature uses, because a jumper has no well-defined wing area. This app flies
lA(α) = 1.7073 sin α cos α dA(α) = 0.1091 + 1.7340 sin² α
whose three constants are obtained by least squares, in code at start-up, against Müller’s three measured angles of attack — 30°, 35.5° and 40°. The fit is within 0.50% on lift and 0.12% on drag. The shape is chosen (the cross-flow form of a separated bluff body) because Müller’s own polynomials return a negative lift area below 7.98° and a negative drag area below 2.49°, and a jump spends its first second in exactly that range.
Fitted to one curve, the model is then checked against others it was not shown. Feeding it the mean flight position of the best ten athletes at the 2002 Olympic competition (Müller Fig. 2 — the angle of attack, the body-ski angle, the hip angle and the ski opening, second by second) and comparing with Müller’s own published force history of a reference jump (his Fig. 6):
Where it does not agree is worth stating too. Its maximum lift-to-drag ratio of 1.90 at α = 13.7° is an extrapolation below every measured range, and it is too high: Ryu, Cho and Cho’s CFD (Trans. JSASS 58(4):203–212, 2015), validated against this same wind tunnel to 7.2%, gives 1.457 at 15°, and FIS’s own competition-measured coefficients peak at 1.33. Four independent sources agree the peak sits near 15–25°, and none of them places it at the 30–40° that is often assumed.
Each hill is rebuilt from nine numbers on its own FIS Certificate of Jumping Hill and the Construction Norm’s own construction: a cubic parabola from the base of the takeoff table to the P point (§4.2: “a cubic parabola will resolve this challenge nicely”), then the landing area as a single circular arc of radius rL through P, K and L (ICR Art. 411.3.2.2), then a parabola to the outrun. Nothing is drawn that the physics did not run on.
The reconstruction is checked against the certificate’s own independently printed P, K and L distances. On Holmenkollen all three land within 9 cm; on Oberstdorf within 4 cm. Two hills do not close, and this app says which and why: Bergisel’s printed rL contradicts its own l₁ and l₂ by a factor of two, so it is not shipped; and Vikersund is a flying hill, outside the design system entirely (HS = 1.2 w there, not w/0.9), where L comes out 7.5 m long.
Where the flight ends. Gasser defines the landing as the intersection of the trajectory with the terrain, with the trajectory starting at the takeoff edge. Standing the centre of mass 0.92 m off the snow instead — which looks more careful — costs eight metres of jump, because near the landing the trajectory closes on the hill at only six or seven degrees.
Which way the wind blows. FIS publishes a headwind factor of about 6 m per m/s on a large hill. Modelled as a horizontal wind, this engine says a 3 m/s headwind shortens a Holmenkollen jump by 7.4 m — the air carries the jumper backwards faster than the extra lift carries them forward. Modelled along the landing slope, which is where ICR Art. 415.3 puts the measuring instruments and what the results sheets mean by “TAN. WIND”, the same 3 m/s lengthens it by 11.4 m. The hill’s factors then come out at –, against FIS’s published 6.0 and 9.0 for a large hill — and the model produces a tail-to-head ratio of about 1.45 out of the physics, where FIS’s own published factors sit at exactly 1.5 on thirteen of thirteen current hills. Nothing here was fitted to that.
Not affiliated with, endorsed by or connected to FIS or any member association. CREDITS.txt lists every source by article and every reconstructed figure; llms.txt carries the machine notes.