A model of a fly, not a fly. Three states, flying straight, turning and resting, and a handful of numbers that say how long each lasts. They are all at the end.
Saccade
A fly does not curve. It flies a straight run, snaps through about ninety degrees in a twentieth of a second, and flies straight again. Inside a run it is predictable. The surprise is only ever when the next turn comes.
Wagner 1986, houseflies. Tammero & Dickinson 2002, fruit flies: 50 to 70 ms a turn.
The memoryless turn
Having flown straight for a while tells you nothing about whether it will keep doing so. The gap between turns is drawn from an exponential, the one distribution with exactly that property. That is why watching longer does not help.
Reaction time
Your swat lands where you aimed, 150 ms after you decide. That is human reaction time at the fast end: the quickest simple visual reaction ever verified is about 100 ms, elite athletes sit near 120 to 150, most people near 250. Once the muscles fire the swatter is close enough that the swing is as good as instant, so the game charges you the reaction and nothing more. With a lag of L against runs that average s, the fly has turned before the swat arrives 1 − e−L/s of the time: at 150 ms against a housefly, about a third. That one line is the whole game.
Not a dial. A longer lag would feel like a bug, with swats missing a fly that was under the ring.
It sees you coming
A resting fly shifts its legs to point its jump away from a looming swatter and is off the wall within a couple of hundred milliseconds, before your arm has moved. Here it notices a swat launched within thirty centimetres and leaves 150 ms later. Approach slowly, aim a little ahead of where it faces.
Card & Dickinson 2008.
The spook
A miss is a swish, and the fly goes frantic: a third faster and turning twice as often, for a couple of seconds, then back to its pattern. That one is from watching real ones, which take longer to settle than this. The trail turns warm while it lasts.
Rest spots
A fly in a room spends most of its time sitting, favours edges and corners, and goes back to the same ones. The oldest advice is about this, not about flight: wait for it to land. The escape reflex is what makes even that hard.
The mosquito
It turns less often and drifts at a third of the housefly's speed, so the same hand mostly wins the bet against it. That is why it is the easy one here, and why in life it is not: a mosquito is hard to see, not hard to hit.
The moth and the lamp
Away from a light a moth is erratic, all wander and no plan. Near one, a reflex that keeps its back to the brightest part of the sky, which for a few hundred million years was up, points it at the lamp instead, and the correction becomes an orbit. That is the room's own light, up on the right, and it catches nothing: it is simply where a moth will be, which makes it where to wait. The zapper you can buy works on the same reflex and does catch.
Fabian et al. 2024, which replaced the older guess about moths mistaking bulbs for the moon.
The catch
A net is wider and slower than a swatter. It forgives a rough aim, and a swatter aimed where the fly will be beats it. The insect is let go at the edge rather than left on the wall.
Until September 2026 the net was bigger, and it caught the mosquito every time: 50 swings out of 50 across everyone's play. It is now a real butterfly net's size and a tenth of a second slower, and a test checks both halves of the trade. There was a third method, a cup and card over a resting insect, removed in September 2026: a minute gives only five or six landings, which is too few to learn a press-and-hold on.
Flypaper
A strip hangs where you click and costs three swats, and it stays up for the rest of the round. Nothing is drawn to it: a fly that sees it coming usually swerves, one that reaches it sticks about one time in four, and a spooked fly is not looking where it goes, which is why a swat behind the fly drives it in. The paper has room for five, and every fly on it is a space the next one cannot use, so it gets less sticky as it fills and a full strip holds nothing at all.
It pays off below about a one-in-five hit rate and not above it, which is roughly where a first day sits. Every fly it catches is a hit: the first three pay back what it cost, and a fourth or a fifth is a hit on top of your twenty, which is the only way a round gets past a hundred per cent. It is also the chancy one: about half the time it comes to nothing. Hang it in the flight path, where it catches nearly three times what it does against a wall, and almost nothing in a corner.
The bug zapper
A zapper glows where you set it and costs five swats. It does not wait to be walked into: its light bends whatever flies near and slows it down, harder the closer it gets, so an insect that comes close spirals in rather than circling. Turning alone would not do it, because a fly at full speed cannot fly a circle tight enough to touch a grid that small however hard it is pulled, which is why it has to lose speed as well. It has nothing to use up and runs for the whole round.
A fly goes to ultraviolet, which is why the kitchen fly light is a UV tube over a glue board, and a moth goes further. A mosquito barely goes at all: a summer of garden zappers in Delaware caught nearly fourteen thousand insects, of which 31 were mosquitoes and biting gnats.
The generous one, at about two a round, though with nothing throttling it that is an average and not a promise: one round in seven it catches nothing at all. Past its five swats, every catch is a hit on top of your twenty. Where you set it is most of the rest. In the flight path it catches several times what it does against a wall, and in a corner almost nothing.
The swatter is not a point
A swat that lands within ten centimetres counts as a hit; that is the ring. The wall is two metres wide at 2.5 mm to a unit. The histogram below the wall is how far every swat missed by, yours in colour, everyone's in grey.
The daily
One round a day, the same for everyone: the date seeds the model, so every player's fly starts in the same place and would fly the same route if nobody swatted. The moment you swat, yours is yours, because a swat spooks it. One attempt, kept in this browser. The line of dots is your twenty swats in order and gives the fly away to nobody.
The crowd
Each swat goes to the wall as the insect, what you used on it, the lag, how far it missed by, whether it hit, and how long you waited before it. The daily adds a day and a score, and each round is noted when it starts and when it ends. An opaque id in this browser lets later swats update earlier ones. That, and a count of page opens, is everything this site stores: no cookies and no names. The one thing loaded from anyone else is Cloudflare's visitor count, described on the privacy page.
The last row is the bug zapper you can buy, not the room light: how hard its glow turns each insect. Its reach is about five centimetres for all three, and inside that it also takes about nine tenths of their speed. Speeds are for a two-metre wall. The housefly's is on the low side of what they manage in a room, because a faster one on a wall this small spends its life avoiding the edges. Turn angles are drawn from a bell curve around the stated value, a turn's timing from an exponential with the stated mean. The rest, the edge-seeking, the length of a rest, is a guess with a plausible shape.
Not here: depth. A real fly leaves the wall, and a swat from the side is a different problem from one arriving flat. Not here either: wind from the swatter, or a fly that learns you are there. A hit leaves a mark and the next insect comes in from the edge a beat later. A round is twenty swats: long enough to see the pattern, short enough that the page ends rather than you having to. A challenge link carries the insect, the mode, your score and a first name if you type one; nothing about it is stored, so faking one fools exactly one friend. The tests under tests/ run the model for two simulated minutes and check it against its own description, not against a fly.