Setting Up a Reactive Agility Test With Light Gates

A wing defender clocks 2.15 seconds on the 505, near the top of the roster, then gets beaten off the dribble in the next scrimmage by a teammate who ran two-tenths slower on the same test. The coach chalks it up to effort. The real explanation is simpler: the 505 measures how fast someone decelerates and re-accelerates once they already know which way they are turning. It says nothing about how fast they read a cue and pick the right direction under time pressure, which is the skill that actually decides most one-on-one defensive moments on a court or pitch.

That gap is why more testing setups now attach a stimulus, usually a light, to an otherwise standard change-of-direction course. The running mechanics stay identical to a 505 or a T-test. What changes is that the athlete does not know which way they are cutting until a light tells them mid-sprint. This guide covers building that setup with light gates, running the protocol so the decision component is actually isolated rather than buried inside one total time, and reading the numbers that come out the other end.

Pre-planned change-of-direction tests are good at exactly one thing: isolating physical qualities. Deceleration strength, hip mobility at the plant step, reactive strength off the push. None of that shifts when you tell an athlete in advance which cone to round. Sheppard, Young, Doyle, Sheppard, and Newton (2006) tested this directly in team-sport athletes, comparing a light-based reactive agility test against a standard pre-planned COD test, and the two barely tracked together at all, r = 0.21, not statistically significant. Less than five percent of the variance in reactive scores was explained by pre-planned COD speed. Knowing where an athlete ranked on one test told you almost nothing about where they would rank on the other.

What did predict reactive agility performance in that same study was a battery of visual and decision-making measures, how quickly the athlete picked up a cue and how efficiently they scanned the space in front of them. That is the part a stopwatch-and-cones setup cannot see, and it is exactly the part that separates a defender who reads the game from one who is only fast in a straight line.

The base layout borrows directly from a standard COD course, then adds one component a 505 does not have: a randomized decision point partway through the run. Athletes accelerate from a standing start, break a trigger beam a stride or two before the decision point, and a stimulus light fires on one of two sides. They redirect toward whichever side lit up and sprint through the matching finish gate.

Keep the decision point at 5m from the start rather than closer. Anything under about 3m catches the athlete before they have reached anywhere near top acceleration, and the test collapses into a simple reaction-time measure rather than a change-of-direction-under-uncertainty measure. Two finish gates sit at a 45-degree angle roughly 3m beyond the decision point, giving a total path of about 8m on either side.

If dedicated reactive light-gate systems are not in the budget, a set of standard dual-beam timing gates plus a separate randomized stimulus light, or even an assistant coach standing at the decision point who points left or right on a random cue heard through an earpiece, will do the job, provided the stimulus timing is genuinely unpredictable.

1. Warm-up (8-10 minutes): light jog, leg swings, lateral shuffles, and 3 build-up sprints through the full course length at increasing effort, finishing at match-speed.
2. Pre-planned baseline (4 trials): tell the athlete which side they are cutting toward before each run, 2 trials per side, no stimulus involved. This becomes their physical-only reference time.
3. Reactive familiarization (2-3 submaximal trials): confirm the athlete understands which light means which direction and that gates are detecting cleanly, at roughly 80% effort.
4. Reactive maximal trials (6-8 trials): full effort, genuinely random side each time, no more than two consecutive calls to the same side, 60-90 seconds of rest between trials since the cognitive load adds up faster than the physical load does.
5. Discard and rerun any trial where the athlete visibly commits to a direction before the light fires, or where total time comes in close to or faster than their fastest pre-planned trial on that side. Both are signs of a guess rather than a genuine reaction.
6. Scoring: average the legal reactive trials for the reactive score, then subtract the pre-planned average from it to get the reactive cost, the number that actually isolates decision speed.
7. Session log: record the stimulus delay range used, ambient lighting (direct sun on an infrared gate causes false triggers or missed breaks), and which color or side convention you used for the stimulus, since that convention needs to stay identical across sessions for the numbers to compare cleanly.

Young and Willey (2010) examined the reliability of a field-based reactive agility test and reported an ICC around 0.80 with a coefficient of variation near 5%, noticeably looser than the 0.90-plus reliability that well-run planned COD tests like the 505 typically produce. That gap is not a flaw in the concept; it reflects the fact that a decision adds a genuinely variable element to the trial that a pre-planned run does not have. Practically, it means you need more legal trials per session to average out the noise, and a slightly bigger change before calling it real progress rather than session variance. One limitation worth flagging: their stimulus was a human tester's body lean, which can develop subtle unconscious tells over repeated testing sessions that a randomized light removes.

Serpell, Young, and Ford (2011) went a step further and tested whether reactive agility itself can be trained independent of physical COD work. Over a 6-week block, a group given video-based decision training improved their reactive agility test times meaningfully more than a group given generic COD training alone, while pre-planned COD times stayed roughly flat in both groups. That is a moderate effect for a short intervention, and it is direct evidence that the decision component and the physical component respond to different training stimuli. The sample was small (n=24) and the stimulus was sport-specific to Australian football, so treat the exact effect size as suggestive rather than something to bank on for a different sport's decision cues.

The single most useful number this test produces is not the reactive time itself, it is the reactive cost: reactive average minus pre-planned average, measured on the identical physical course. That subtraction cancels out the athlete's raw COD ability and leaves you with something close to a pure decision-speed number.

The reference ranges below come from an 8m total path, 5m to the decision point plus 3m to the finish gate, and should be read as a general zone rather than a cutoff. Course length, stimulus type, and surface all shift the absolute numbers.

A wide reactive cost paired with an already-strong pre-planned time is the clearest signal this test gives a coach. It means the physical qualities are there and the bottleneck sits in perception and decision-making, which points training toward small-sided games with numerical uncertainty and video-based anticipation reps rather than more cone-and-gate volume. A narrow reactive cost paired with a slow pre-planned time flags the opposite problem, and the fix is standard COD strength and mechanics work.

Most bad reactive agility data traces back to the stimulus being more predictable than it looks, not to anything wrong with the athlete.

None of these look serious in the moment. Together, they are the difference between a test that measures decision speed and one that quietly measures how well an athlete memorized your setup.

Retest on the same 4-6 week cycle as any other agility marker, and track pre-planned time and reactive cost as two separate lines rather than one composite score.

If pre-planned time keeps improving but reactive cost holds steady or widens, more COD drilling is not the answer. Shift volume toward small-sided games with numbers-up or numbers-down scenarios and video-based anticipation reps that force a real decision under time pressure.

If reactive cost is already tight but pre-planned time has plateaued, the athlete is deciding fast enough already, the limiter is deceleration strength or plant-step mechanics, and standard COD strength work is the better use of the next training block.

A test that only reports total time will never tell you which of those two situations you are looking at. The light gate is what turns one number into an actual training decision.