Curve Sprint Test for Soccer: Protocol, Radius, and Deficit

A winger clocks 3.10s over 20m in preseason testing, near the top of the squad, and then spends the first competitive match arriving half a step late on every underlapping run around the fullback. Nobody's arithmetic is wrong. The test just measured the wrong shape of speed. Almost none of the high-velocity running in a soccer match happens in a straight line — players arc around opponents, bend their run to stay onside, curve off the ball to open a passing lane — yet nearly every standard speed test still uses a 10m, 20m, or 30m straight sprint. Fílter and colleagues built a curve sprint test specifically to close that gap, using the one curved line that already exists on every marked pitch: the penalty arc. What follows is the full protocol, the radius options the research has actually tested, how to score a curve sprint deficit, and the mistakes that quietly invalidate the number.

The curve sprint test asks an athlete to sprint maximally along a fixed-radius arc rather than a straight line, then compares that time against a linear sprint over the same distance. Fílter, Olivares-Jabalera, Santalla, Nakamura, Loturco, and Requena (2020) formalized the modern version using the soccer penalty arc itself — a 9.15m radius from the penalty spot, producing a 17m curved path — tested to both the right and left, alongside a matched 17m linear sprint. Forty experienced soccer players performed three trials of each condition across two separate testing days, timed with photocells.

Reliability held up well for a field test built around a continuously curving line rather than fixed gates: inter-session ICCs of 0.93 for the right-side curve and 0.89 for the left, with intra-session coefficients of variation of 0.87% and 1.15%. The more interesting number is what the curve sprint did not share with the linear sprint — the two overlapped by only ~35% of shared variance (R²), meaning straight-line speed explains barely a third of how fast a player moves along a bend. The rest is something the 20m dash simply isn't built to see.

Fílter's group had a practical reason for building this test: an earlier kinematic analysis of match footage found that roughly 85% of maximal-velocity actions in a professional soccer league were curvilinear, not straight. Their companion kinematic analysis (Fílter et al., 2020, International Journal of Sports Medicine) found the mechanics genuinely change on a bend — inside and outside legs take on different roles, with asymmetric ground contact and step-width adjustments needed to generate centripetal force without losing forward propulsion. That's why a player can be fast in a straight line and still look mechanically unfamiliar the first time you ask them to sprint hard around a tight arc.

Radius matters too. Altmann, Ruf, Fílter, and colleagues (2024, Science and Medicine in Football) tested 19 elite youth male players across three curve radii — 11.15m (wide), 9.15m (medium), and 7.15m (narrow) — each over 17m, both sides, against a matched linear sprint. All three curved conditions were significantly slower than the linear sprint, with mean differences of 0.08–0.16s and large effect sizes (0.83–1.49). The widest and narrowest radii differed by a further 0.04s (effect size 0.47), and most players were significantly slower on one side than the other at the medium radius — side dominance is a real, measurable pattern, not noise.

One caveat: a separate study in elite female players (17 professionals, curve velocity rather than time) found a near-perfect link between linear and curve sprint velocity (r > 0.9) — very different from the ~35% shared variance above. Sex, sample size, and whether you compare velocity or raw time all seem to move that relationship around, so no single correlation should be read as final.

The published protocol leans on a line that's already painted on every regulation pitch, which is most of its appeal as a field test.

• Primary arc: The penalty arc itself, 9.15m radius from the penalty spot, giving a 17m curved path. Off a marked pitch, replicate it with a rope anchored at a fixed center point and cones every 2–3m so the line is visible at speed.
• Linear comparison sprint: A straight 17m sprint, matching the curve distance exactly — the deficit calculation depends on it, so don't substitute a 20m or 10m line for convenience.
• Both directions: Mark the arc to run curving right and curving left from the same start line. Given the side-dominance findings above, testing only one direction misses real information.
• Timing: Photocell gates at start and finish. A stopwatch introduces enough reaction-time error to blur the 0.04–0.16s differences this test is built to detect.
• Optional radius kit: For athletes already familiar with the standard test, add 11.15m and 7.15m arcs (same center point, different rope lengths) to profile performance as the turn tightens, following Altmann et al.'s design.
• Surface and footwear: Standardize both — cleats that grip well on a dry, tight arc can behave very differently on damp grass, and that shows up more in curved running than straight sprinting.

Warm-up (10–12 minutes): Progressive jogging and dynamic mobility, then 3–4 straight build-up strides to roughly 80% effort. Add two or three submaximal practice reps on the arc itself — most athletes run their first curved rep conservatively, hugging a line inside the marked radius, and a couple of low-intensity reps to learn where the arc sits prevents that from contaminating the maximal trials.

1. Athletes start from a standing position behind the line, front foot at the start marker.
2. On the tester's signal, sprint maximally along the marked arc to the finish line, staying as close to the marked path as the cones allow. Cutting inside the arc shortens the distance and invalidates the trial; drifting wide does the opposite.
3. Record time from the start photocell to the finish photocell.
4. Allow 2–3 minutes of full recovery, then repeat for three trials per side (right-curving and left-curving).
5. Randomize or alternate the order of right, left, and linear conditions across athletes rather than always testing the same sequence, which controls for fatigue or pacing-learning effects building across the session.
6. Finish with three maximal trials of the matched 17m linear sprint, same rest interval.

Take the fastest of the three trials as the score for each condition — right curve, left curve, and linear. Record all three individually rather than an average; the gap between them is where the coaching information lives.

Curve sprint deficit (CSD) is calculated the way change-of-direction deficit is for cutting tests: CSD = curve sprint time − linear sprint time, using the fastest trial of each over the same 17m distance. A larger deficit means more time is lost specifically to running on a bend, independent of straight-line speed. Since the test also yields a right- and left-side score, side-to-side asymmetry can be calculated too: (slower side − faster side) / faster side × 100.

Across all three radii, curved sprints were 0.08–0.16s slower than the matched linear sprint (effect sizes 0.83–1.49, large), and the widest radius beat the narrowest by about 0.04s (effect size 0.47). There's no agreed set of normative deficit bands yet the way there is for something like the 30-15 IFT — this is a newer test, focused so far on reliability and mechanics rather than large-sample benchmarking — so the most useful comparison right now is an athlete against their own baseline and left-right split, not a fixed cutoff from a single study.

Because curve and linear sprint speed share only about a third of their variance, a program built entirely around straight-line acceleration and top-speed work trains a skill that doesn't fully transfer to match running. Curved sprint exposure has to be programmed on its own terms.

Start technique work at the widest comfortable radius (11.15m) before progressing to 9.15m and eventually 7.15m as arc-running mechanics and single-leg strength improve — the tighter the turn, the more centripetal force the outside leg produces while the inside leg manages a shorter, compressed step. Program both directions every session rather than favoring an athlete's stronger side; given how consistently side dominance showed up in the research, deliberately loading the weaker direction is usually the higher-value decision. A typical max-velocity curved session mirrors linear structure: 4–6 reps per side at a given radius, 90–120 seconds full recovery so quality doesn't degrade into fatigue-driven technique breakdown.

Pair curved sprint work with unilateral strength and hip-stability training — the outside leg produces more lateral force while the inside leg shortens its stride, so hip abduction/adduction strength and single-leg deceleration control give the mechanics something to stand on. Retest deficit every 4–6 weeks where the schedule allows, using the same radius and timing setup each time.

• Freehand arcs instead of a fixed radius. Cones placed by eye without a rope anchored at a true center point drift in radius session to session — and since Altmann et al. found a 0.04s difference between just two radii, an inconsistent arc can silently swap a real fitness change for a measurement artifact.
• Testing only one direction. Given how often side dominance showed up in the research, a single-direction test misses exactly the asymmetry this test is positioned to surface.
• Skipping the linear comparison sprint. A raw curve time without a matched 17m linear time to subtract gives you a number with no baseline — you can't tell whether an athlete is generally slow or specifically weak at curved running.
• No familiarization reps. Athletes new to maximal curved sprinting tend to run conservative lines on their first attempt, inflating the apparent deficit for reasons unrelated to physical capacity. Two or three submaximal practice reps fix this cheaply.
• Confusing this with a change-of-direction test. The 505 and pro-agility tests measure a discrete braking-and-cutting action at a fixed angle. The curve sprint test measures continuous running along a bend with no stop, plant, or reacceleration — a different physical quality, and results from one shouldn't be used to infer the other.