Methodology

Everything runs in your browser

Each challenge page arrives as ordinary HTML plus one script: the challenge itself. Once the page has loaded, playing sends nothing to our server. The timing, the scoring and the puzzle logic all run on your device, and your result is calculated there.

That keeps challenges fast and private, and it means no network delay can creep into a measurement. It also means your device's own speed is part of every timed result, which is what most of this page is about.

The clock

All timing uses performance.now(), the browser's high-resolution clock. Unlike the time of day, it only ever moves forward at a steady rate, so it isn't thrown off if your computer adjusts its clock mid-test. Browsers deliberately round it slightly for security reasons, by up to about a millisecond depending on the browser, which is why we report reaction times in whole milliseconds and nothing finer.

When the clock starts

In the visual reaction tests, the moment the signal changes, we tell the browser to show it straight away, and then start the clock in the next animation frame: the step in which the browser paints that change. Starting the clock any earlier would count the time the browser spends preparing the screen as part of your reaction.

In the Audio Reaction Time Test, the beep is scheduled ahead of time on the browser's audio clock, and we convert that moment to the page clock using the browser's report of what the speakers are playing at that instant.

When the clock stops

Every press, tap or key comes with a timestamp the browser records when the input arrives. We use that timestamp rather than reading the clock when our code gets round to handling the press, which could be a little later if the page is busy.

We listen for the moment a button, key or finger goes down, not for a completed click. A click only fires when you let go, which would add the time your finger takes to come back up.

What the browser can't see

A browser can time things accurately between the moment it draws something and the moment it receives your response. It can't see the delays outside that window, and every device has some:

• The screen. A display shows a new picture only when it refreshes: every 16.7 ms at 60 Hz, every 8.3 ms at 120 Hz. A change can wait up to one refresh to appear. Some screens, especially TVs, add processing time on top.
• Mice and keyboards. They report to the computer at fixed intervals. A common USB rate is 125 times a second, every 8 ms; many gaming devices report 1,000 times a second. Wireless devices can add more.
• Touchscreens. They check for touches at their own rate, and phones often spend time working out whether a touch is a tap or the start of a scroll.
• Bluetooth audio. Wireless headphones and speakers buffer sound before playing it, which can delay a beep noticeably, and the browser doesn't always know by how much.

These delays are part of every timed result, so results from different devices aren't directly comparable. The fairest comparison is always with yourself, on the same device, responding the same way. The Time Perception Test is the exception: it times the gap between two of your own presses, so device delays affect both ends equally and mostly cancel out.

Background tabs and busy devices

Browsers slow down timers in tabs you aren't looking at, to save battery. If you switch away in the middle of a timed round in a reaction test, that round is dropped rather than scored, because its timing can't be trusted. In Multiple Object Tracking the motion pauses while the tab is hidden and carries on when you come back.

A device that is busy with other work can be slower to draw changes and to notice input. For the steadiest results, close heavy tabs and plug a laptop in so it isn't saving power.

Randomness

Waits, which square lights up, tile patterns, shapes, dot positions, number puzzles, Stroop words and inks, where the T hides, Lights Out boards, Go/No-Go orders, target positions and flicker pictures are all chosen at random by your browser each time. Random waits are drawn evenly from a stated range, such as 2 to 5 seconds in the Reaction Time Test, so the timing can't be learned. Where a challenge needs a balance, it is built in rather than left to chance: exactly half the Mental Rotation pairs are mirror images, and exactly half the Stroop words match their ink.

How each result is calculated

Every challenge page goes into more detail. One line each:

Where the ratings come from

Some results come with a rating, such as Excellent, on a five-step scale. We still don't store anyone's results, so the ratings never come from other visitors to this site. They are shown only where they can be backed up in one of two ways:

• A published typical result. Where a public source reports typical results for a test built the same way as ours, the five bands are our own classification, set so that the source's typical result falls in the middle, Average, band. They are not an official, medical or percentile scale.
• A proven best answer. Puzzles with a fewest possible number of moves are rated on how close you came to it: exactly the fewest is Perfect, up to 15% more is Excellent, up to 35% more is Good, and up to 80% more is Fair. No population data is involved.

Every other challenge shows no rating, because we found no public data for a test built exactly like ours, and a rating without data would be invented. Several still quote a published average for a similar test as a rough guide; those are listed in the next section. We never show percentiles such as "top 10%": those are precise statistical claims we can't make. Sources were last checked on 2026-10-12.

Where the averages in our FAQs come from

Many challenge pages answer "what is the average score?" with a published figure for a similar test. Each one is named and dated on the page, and listed here with a link; all were checked on 13 October 2026. The tests behind them differ from ours in some way, such as paper instead of a screen, or different rules for mistakes, so they are a rough guide, not a rating.

• Verbal Memory Test: About 41 points (middle score). Human Benchmark, Verbal Memory statistics chart
• Visual Memory Test: Level 11 (middle score). Human Benchmark, Visual Memory statistics chart
• Sequence Memory Test: Corsi span 5.0 (mean, ages 21–89). Facchin and colleagues (2024), Neurological Sciences
• Mouse Accuracy Test: 450–500 ms per target (middle score). Human Benchmark, Aim Trainer statistics chart
• Stroop Test: 83 ms Stroop effect at ages 21–34 (mean). Forte and colleagues (2024), BMC Psychology
• Flanker Test: 44 ms flanker effect, young adults (mean). Cheng and colleagues (2021), Frontiers in Behavioral Neuroscience
• Trail Making Test: Part A 24 s, Part B 51 s at ages 25–34 (mean, paper test). Tombaugh (2004), Archives of Clinical Neuropsychology
• Digit Symbol Substitution Test: 57 correct in 2 minutes at ages 60–69 (mean, paper test). CDC, National Health Statistics Reports no. 126 (2019)
• Schulte Table: 30–50 s per 5×5 table counted as good. ClinicalTrials.gov registration NCT06613282 (2024)
• Face Memory Test: 57.9 of 72 on the Cambridge Face Memory Test (mean). Duchaine and Nakayama (2006), Neuropsychologia
• Audio Reaction Time Test: About 140–160 ms to a sound in the lab. Jain and colleagues (2015), Int. J. of Applied and Basic Medical Research
• Peripheral Vision Test: About 90° to the side, 50° up and inward, 60° down. Spector (1990), Visual Fields, in Clinical Methods

What results are not

The challenges are for fun and curiosity. Their results aren't medical, clinical or psychological assessments, they don't measure intelligence, and they can't tell you whether you are fit to drive or to do any other task. If you are worried about your memory, attention, hearing or reaction times, talk to a doctor.