Reaction Time Test

Simple visual reaction time follows a shallow curve across the lifespan: it sharpens through the teens, peaks somewhere in the early-to-mid twenties, then slows gradually — a few milliseconds per decade through midlife, steepening after about 60. The ranges below are typical medians on a click-based visual test like this one, not diagnostics.

Here is the part most age charts bury: for one person, the gap between a rested day and a wrecked one is routinely larger than the gap between two adjacent decades. A 25-year-old on four hours of sleep will lose to their own rested score by more than a decade of aging would cost them. Which is why an age percentile is entertainment, and your own baseline is information. See the fuller breakdown in average reaction time by age.

A visual reaction test measures the whole chain: the screen has to draw the stimulus, your eyes and brain have to process it, your hand has to move, and the input device has to report it. Only the middle part is you. The rest is equipment, and it is worth real milliseconds.

The display numbers are just arithmetic: a 60 Hz monitor draws a new frame every 16.7 ms, so on average the flare waits about half a frame before it physically exists. Stack a 60 Hz panel, a cheap polling rate and a phone touchscreen against a tuned desktop and the equipment alone can account for something like 20–40 ms — bigger than most of the differences people argue about online.

Two practical consequences. First, comparing your number to a stranger’s is close to meaningless unless you happen to share a setup. Second, comparing your number to your own number on the same device is clean, because the hardware delay is a constant that cancels out. That is the entire reason this site is built around baselines rather than leaderboards.

If you want the fairest possible score: use a wired or modern low-latency wireless mouse on a desktop, close background tabs, disable any frame-smoothing your GPU driver is applying, and run it on the same machine every time.

Simple reaction time is the most state-sensitive measurement in cognitive science. That is a weakness if you want a stable IQ-style number, and a strength if you want to know how you are doing right now — which is the actual question before a ranked queue.

• Sleep loss — the big one. A single skipped night costs roughly +49 ms and about five times more attention lapses in lab conditions. Chronic short sleep accumulates a similar debt without the subjective feeling of being wrecked, which is the dangerous part.
• Time of day. Everyone has a circadian trough — for most people somewhere in the small hours, with a smaller dip in the early afternoon. Same person, same setup, several tens of milliseconds apart.
• Caffeine. Reliably improves vigilance performance, with the catch that it is mostly restoring you rather than upgrading you, and it borrows against the sleep that protects tomorrow’s score.
• Alcohol. Slows responses and inflates lapses at doses well below where people feel impaired.
• Time on task. Vigilance decays within a session. Your twentieth minute of testing is slower than your second — which is also true of your fifth game in a row.
• Warm-up. The first trial or two of any session tends to run slow. This test discards nothing, but the median absorbs it, and the daily check compares like-for-like.
• Illness and dehydration. Both drag vigilance measurably, often before you notice symptoms.

Notice that most of that list is under your control the night before, not in the moment. The honest use of a reaction test is not to grade your talent — it is to catch the days when your equipment (you) is running below spec.

This page measures simple reaction time: one stimulus, one response, no decision. The flare appears, you press. That purity is exactly why it is the best available probe of your current state — with the decision stripped out, what is left moves almost entirely with alertness.

It is also the least game-like measurement on this site. Real gameplay almost never gives you one signal and one button. It gives you a wind-up that might be a hook or might be a bait, and two or three things you could do about it. Adding even a two-way choice slows everyone by roughly 50–150 ms, and the size of that penalty varies far more between people than simple reaction time does.

So use them together: this test tells you how sharp you are today; the reflex test adds the decision step and tells you how well you choose under time pressure. If your simple reaction time is normal but your choice reaction time has fallen off, the problem is not your hands — it is processing, and that is a much better argument for not queueing.

The instrument is a psychomotor vigilance task — the paradigm sleep researchers reach for when they need a measurement that is sensitive to state and resistant to practice effects. The implementation details matter more than most sites admit, so here they are.

• Random foreperiod. The delay before each flare varies unpredictably. A fixed delay would let you time the rhythm and measure your counting rather than your reacting.
• False starts repeat the trial. Press before the flare and that trial is thrown out and re-run. Without this, the optimal strategy on any reaction test is to guess early and keep the lucky ones — which is how people end up with 120 ms scores that mean nothing.
• Eight trials, and we report the median. Not the mean, and not your best. One lapse can drag a mean by 40 ms while telling you nothing about your typical response; your best single trial is mostly luck. The median is what labs use because it is the robust estimate of your central tendency.
• Timing uses performance.now(). A high-resolution monotonic clock, not wall-clock time, so nothing drifts or jumps mid-trial.
• Nothing here is recorded. Runs on this page are local to your browser. There is no leaderboard to farm and no reason to cheat the instrument, which keeps the measurement honest.

Same engine, same scoring as the slot inside the daily check — the only difference there is that your results are stored so a baseline can form.

Ranked honestly, from largest effect to smallest:

• Sleep properly. Nothing else on this list comes close. Sleep is not a way to get faster than your ceiling — it is the difference between playing at your ceiling and playing 50 ms under it, every night, for free.
• Fix your hardware once. Moving from a 60 Hz panel and a 125 Hz mouse to a high-refresh display and a 1000 Hz polling rate is a permanent gain of roughly 10–15 ms that requires no discipline whatsoever. It is the only shortcut on this page.
• Warm up before the games that count. The first minutes of any session run slow. A couple of minutes of anything responsive costs you nothing and moves your first game out of the warm-up penalty.
• Practise the task. Repeated practice does reliably improve scores on reaction tasks. The honest caveat: most of that gain is learning the specific task, and how much transfers to a moving champion in a teamfight varies a lot between people.
• Use caffeine deliberately. It works, it is well studied, and it is a loan. Taken late it costs you the sleep that was the top item on this list.
• Don’t queue tilted or exhausted. Not a training tip, but it protects more LP than any of the above. The cheapest winrate you will ever buy is the games you decided not to play.

What will not work: reaction-time supplements, and any product promising to raise your ceiling by a large margin. Your ceiling is fairly fixed. Your daily distance from it is not, and that gap is where all the recoverable performance actually lives. The longer guide is at how to increase your reaction time.

A mid-range skillshot gives you somewhere between about 250 ms and 1,200 ms to respond, depending on distance and projectile speed. At max range, nobody’s reaction time is the limiting factor — the dodge is decided by whether you were watching. Up close, in the 250–400 ms window, it is decided almost entirely by how sharp you are, and a 50 ms swing between your rested and tired self flips a meaningful share of those.

In a 516-player study of League players, reaction speed was the strongest cognitive correlate of rank on record. That is a correlation across players, not a promise about you — being fast does not make you Challenger, and the causation could run either way. What it does establish is that this measurement is pointed at something the game genuinely taxes.

The practical read: your reaction time is not a rank predictor, it is a readiness signal. It will not tell you whether you deserve to climb. It will tell you, with real sensitivity, whether today is a day where the close ones go your way — which is why it anchors the daily check alongside four other measurements.

Human Benchmark is the reference point for this whole category and deserves the traffic it gets — it made cognitive testing a normal thing to do on the internet. If you just want a number to text a friend, it is great and this page will not change your life.

The differences that actually matter are methodological. Their reaction test runs five trials and reports the average; this one runs eight and reports the median, which is markedly harder to distort with a single lapse or a single lucky click. False starts here repeat the trial rather than counting, so early-guessing does not pay. And their score is a one-shot population percentile, where this test is built to be repeated so a personal baseline can form.

That last one is the real split. A population percentile answers “am I fast compared to strangers on unknown hardware?” — a question with a lot of noise in it. A baseline answers “am I fast compared to me, on this machine, on my good days?” That question is answerable, and it is the one worth acting on before you queue. Both tests are free; this one is free without a signup, and the rest of the battery is too.