| What the number is | What the number is not |
|---|---|
| Seconds you kept the red block clear of the blue ones | A reaction time in milliseconds. That is a different measurement, and this page does not take it |
| A score for this task, on this device, with the control you used | A universal score you can set against somebody else’s |
| Where that score falls in a field modelled from two simulated reference players | A place on a real leaderboard. Nothing about your run leaves your device |
Move one red square. Nine blue rectangles bounce around a track and accelerate as the clock runs. Let a rectangle touch you, or touch the dark ring around the field, and the run is over. Most people last a few seconds on their first attempt. Ten seconds is a strong run, and the block below the field will tell you exactly where yours landed.
Before you spend an evening on it, here is what it is actually worth. This is not a reaction time test, and there is no good evidence that playing it makes you faster at anything outside this page. It measures something narrower and more real: tracking several moving objects at once and getting out of the way in time. That improves quickly at first, mostly because the controls stop needing thought. Every run ends with a chart showing where it sits.
How to play, and what ends a run
On a computer, move your mouse inside the field. The red square follows the pointer, and there is nothing to click. On a phone or tablet there is a control strip below the field: slide a finger across it and the square moves with you. That keeps your hand off the part of the screen you are supposed to be watching. Pressing the field itself works too, if that is where your thumb goes first.
Two things end a run, and they are both easy to forget under pressure. A blue rectangle touching you ends it. So does the edge: the dark ring around the play area is not a frame, it is the boundary, and it is lethal. Your best score is kept on your own device and nowhere else.
The rectangles are not identical. Each is drawn at random, so they differ in size from run to run, and they get faster the longer you survive. The readout beside the clock shows the current multiplier: one at the start, 1.50 once you pass ten seconds, 1.75 at twenty, and levelling off just under two.
What makes it hard
Almost nothing about this game is forgiving, and knowing why is most of the way to playing it well.
The edge is a liability, not a shelter. In most dodging games a wall is somewhere to hide behind. Here it is the thing that kills you. Hugging the border costs you twice: you lose the room to escape, and you spend the run watching one direction instead of all of them. Keeping to the middle is worth roughly double the survival time.
There is nowhere to camp. Nine rectangles move at once, they bounce off the walls at slightly unpredictable angles, and their paths are not something you can learn and recite. Any position that looks safe now will be crossed by something within a second or two.
And the clock is not on your side. Every ten seconds the rectangles close half the remaining gap to twice their starting speed. A long run is disproportionately hard, because the last few seconds are played at a speed the first few never reached. That is what stops a lucky start from turning into a lucky minute.
What counts as a good score
Ten seconds is a solid first session. Twenty is genuinely good. Anything past thirty puts you in the far tail of the field, and a small number of players get past a minute.
The chart under the field turns your run into a percentage. Here is the same curve as a table, so you can see where the lines fall before you play:
| You lasted | You beat | What that looks like |
|---|---|---|
| Under 3 seconds | 14th percentile | The blocks won before you had the controls figured out. Almost everyone’s first two runs land here |
| 5 seconds | 34th percentile | You are moving, but reacting to what is already next to you |
| 10 seconds | 67th percentile | A strong first session, and where most people settle after a dozen attempts |
| 20 seconds | 91st percentile | You are reading the gaps rather than the blocks and picking escape routes early |
| 30 seconds | 97th percentile | Top few percent. The rectangles are moving at close to double speed by now |
| 60 seconds | 99th percentile | About as far as the model can usefully separate anyone. The curve reaches 99.6% here and the panel caps the display at 99 |
It is a model, not a tally. Nothing about your run leaves your device, so there is no leaderboard and no collected data to rank you against. What the chart does instead is place your score on a curve. The curve is fitted to two simulated players measured on this exact field. One never touches the controls, and lasts a median of 2.5 seconds. The other dodges near-optimally, and lasts a median of 21.6 seconds. Treating those as roughly the 10th and 92nd percentile of the field fixes the shape, and its middle sits at about seven seconds.
Those two reference marks are the only measured points in the table above. Everything between them is an assumption, and it lives in one place in the code so it can be replaced the day there is real data to fit. The percentage is capped at 99 deliberately: rounding the top of the curve would print a clean 100, which would tell you that you beat everyone.
Six things that actually move your score
These are the levers that matter, in rough order of how much difference they make.
- Stay off the edge. The single biggest change for most players. Central territory means more directions to escape into. If you keep dying near a wall, that is the fix.
- Watch the gaps, not the block. You need the space around your square, not the square itself. Players who stare at their own block die to something arriving from out of frame.
- Move early and less. Small corrections made a beat early beat large ones made at the last moment, because a rectangle’s path is easier to predict before it is close.
- Use a mouse if you can. A pointer is more precise than a thumb, and on a phone a larger screen gives you more room to work with. Scores are not comparable across the two.
- Take three or four runs before judging yourself. Runs one and two are spent learning the controls. The measurement starts after that, and your best score will keep climbing for a while.
- Sleep and caffeine are real, and they show up here. Sleep loss measurably slows reaction time, which how much sleep you actually need covers in more detail, and caffeine moves the number the other way for a while, which caffeine against a nap compares. Neither turns this into a medical test. Both are effects you can watch in your own scores across a week.
The skill it actually tests
Visual tracking under time pressure. You are watching several objects, predicting roughly where they will be a fraction of a second from now, and moving one of them out of the way. It leans on anticipation and on the hand or thumb speed to act on a decision you have already made.
It is not a reaction time test, and the difference matters. A reaction time test measures one response to one stimulus and reports it in milliseconds. The usual figure quoted for an adult responding to something they see is about 250 milliseconds, and it varies with the sense involved and with how alert you are. This game reports seconds of continuous survival, which blends several abilities and cannot be reduced to one of them. If you want a number in milliseconds, this is the wrong page, and saying so is more useful than letting you assume otherwise.
It is also not a measure of intelligence, eyesight, fitness or reflexes in any clinical sense. It measures how you did at this task, on this screen, with this control.
Does it make you quicker at anything else?
There is a real body of research on action games and reaction speed, and it is more contested than the headlines suggest. The honest version has three layers.
The correlation is solid. People who play a lot of action games respond faster than people who do not, and consistently so. The largest meta-analysis of the area was published in Psychological Bulletin, pooling the studies from 2000 to 2015. Players came out about half a standard deviation ahead on perceptual, attentional and cognitive measures, with faster responses and no loss of accuracy.
The causal claim is much weaker. The same analysis looked separately at studies that trained people rather than compared them, and found a considerably smaller average effect, about a third of a standard deviation. It also flagged publication bias, estimating that effects in the published literature run about 30% larger than in the full literature. A later re-analysis in Collabra: Psychology went further. Applying stricter corrections to the same set of studies, its authors found minimal evidence that the training transfers at all, and argued that some of the original effect came from the same participants being written up across several papers, a practice they call salami slicing.
And none of it is about this game. Every study above used complex action or strategy titles: shooters, real-time strategy, a football game. Moving one square away from other squares is not that. Gains that transfer in the literature appear after tens of hours of play, and even that is disputed. One 2021 study found twenty minutes of an action game produced no measurable change in cognitive control. A 2025 study did find faster responses after a single hour, but its sample was small, it tested immediately afterwards, and it could not separate real learning from arousal or simple practice.
The one thing that is certainly true is the least exciting: you get better at this specific task by doing it. That is practice, not transfer. Your third run will beat your first because you have learned the controls and started reading where the rectangles are going. Dressing that up as anything more would be dishonest, and this page has no reason to.
Where the “18 seconds” and the Air Force claim came from
Almost every copy of this game carries some version of two lines: that surviving eighteen seconds makes you exceptional, and that United States Air Force pilots are required to last two minutes. Neither has a source, and the second one is worth taking apart.
The text is old and has been copied forward for two decades. The earliest pages hedged it themselves: the wording was “it has been said that the US Air Force uses this for fighter pilots”, with no citation offered and none apparently sought. From there it spread through game portals, forum posts and app store listings without anyone adding a reference. The threshold moves as it travels, too: eighteen seconds on some sites, twenty-two or fifty on others, whichever number the copywriter liked.
The people best placed to check it did, in 2007. A naval aviation forum thread titled “Internet Myth on AF Fighter Pilot Skills” took the claim apart, and the pilots posting their own attempts landed at 16, 17.985, 18.2 and 23.6 seconds. If the game were a screening tool with a two-minute bar, the people it was screening would not be scoring seventeen.
What the Air Force actually uses for pilot selection is a computerised battery called the Test of Basic Aviation Skills: nine subtests taking about seventy-five minutes, covering psychomotor coordination, spatial ability and multitasking, alongside a separate academic aptitude score. There is no dodge-the-blocks component and no two-minute threshold. The published figure for fighter pilots’ visual reaction time is around 230 milliseconds, three orders of magnitude away from two minutes, because the two are different quantities entirely.
The accurate statement is that the claim has no verifiable source. Not that anybody disproved it: there was never a document to rebut. It is chain-letter copy repeated often enough that it now reads as fact. Which is also why the percentile on this page is a stated model rather than a threshold dressed up as a fact.
Reading the number honestly
Your score is not comparable with anyone else’s, and barely comparable with your own across devices. Four things move it: whether you used a mouse or a thumb, how large the field rendered on your screen, the refresh rate and latency of your device, and how familiar you already were with the game.
Two runs of your own are comparable, though, which is the point of the ramp being tied to elapsed time rather than to how well you are doing. Every attempt reaches 1.50 at the ten-second mark, so the same number always describes the same conditions. A longer run is genuinely harder rather than merely longer.
The one published reference point worth quoting is the world record kept by the site hosting the most-played English version: 123.1 seconds, with country averages they collect themselves. Treat it as an interesting number rather than a target, and remember that they are collecting scores and this page is not.
If you want tools that measure something you can act on, the protein intake calculator and the cost per use calculator both do that. This one is here because it is a good test of a small skill, and because it is a claim worth being accurate about. The rest of the tools are worth a look too.
Questions people actually ask about this test
Is 18 seconds a good score?
It is a good score, and it is nowhere near a threshold. Eighteen seconds sits around the 88th percentile of the modelled field, so it is better than most attempts and worse than a fair number. The “18 seconds” figure comes from copy that has been repeated for twenty years without a source, covered above.
Is this a reaction time test?
No. It reports seconds of survival in a tracking task, not milliseconds of response to a single stimulus. If you want the reaction time number, you need a test built for it: one stimulus, one response, timed. The average for an adult reacting to something visual is around 250 milliseconds.
Will playing this make my reaction time faster?
Almost certainly not in any way you would carry elsewhere. It will make you better at this game. Evidence that action games speed up general responses is mixed, the effects are small, and the studies used far more complex games over far longer periods.
How do you actually improve reaction time?
Sleep is the largest lever you control, because sleep loss slows responses measurably, and how much sleep you actually need covers how much you are probably short of. Practice the specific task you care about, since almost all of the gain is specific rather than general. Caffeine helps for a while, as caffeine against a nap compares. Nothing here is a shortcut past those three.
What is the highest score anyone has managed?
The site hosting the most-played English version lists a world record of 123.1 seconds. On this page’s curve that is the 99.98th percentile, which is why the panel shows a capped 99 rather than pretending the model can rank the very top.
The Bottom Line
This is a tracking and avoidance test, not a reaction time test, and it will not make you faster at anything else. What it will do is improve, quickly and visibly, at the one skill it measures. Ten seconds is a strong first session. The chart after each run tells you where you stand, and the basis for that chart is printed on the panel rather than hidden.
Next step: play four runs in a row and watch the third and fourth. If your best comes from the middle of the field rather than the edges, you have already found the biggest lever there is.
This content is for informational purposes only. It is not a medical or psychological assessment, and it is not a substitute for advice from a qualified professional.