What actually makes you faster
Most of what the internet sells for reaction time does nothing. Most of what works is free, and you already know what it is.
Reaction time feels like it should be trainable the way strength is: put in the reps, watch the number fall. The honest answer from a century of measurement is less exciting and more useful. The number has a floor set by biology, a daily wobble set by how you live, and a small trainable slice at the top. Here is where the milliseconds actually are.
The part you cannot move
Nerve conduction speed, synaptic delays and muscle activation set a hard floor around 150 to 200 milliseconds for responding to a visual signal, and no training touches it. Age moves the floor slowly against you: simple reaction time is quickest in the early twenties and drifts upward by fractions of a millisecond per year after, so gently that habits, health and practice swamp it for decades. If someone advertises a method to react in 80 milliseconds, they are selling a way to guess early, and the reaction test here already discards those.
The big levers are boring
Sleep is the largest lever most people hold. Restricting sleep reliably slows simple reaction time by tens of milliseconds, and adds something worse than slowness: lapses, the half-second-plus responses where attention simply went away. One short night shows up in the number; a week of short nights shows up like alcohol.
Caffeine genuinely works: a coffee’s worth reliably buys back a small double-digit slice of milliseconds, more if you were tired, with diminishing returns after the first cup. It is the one supplement with a large, replicated effect. Everything else marketed for reaction speed lives somewhere between “small and fragile” and “no effect when tested properly.”
Exercise and arousal matter in both directions. Moderate physical arousal, a warmed-up body, standing rather than slumping, speeds responses; exhaustion slows them. Regular aerobic fitness correlates with faster processing at every age, most visibly past middle age.
Alcohol is the anti-caffeine, worth tens of milliseconds in the wrong direction at doses people call “fine to drive on.” The lapses grow before the average does.
What practice really buys
Practice on a reaction task does improve your score, but the gain is mostly not nerve speed. Early runs carry overhead: hesitation about the task itself, wasted false starts, inefficient key grips. A few sessions strip that away, typically 10 to 30 milliseconds on a simple test, and then the curve flattens hard. What keeps improving longer is consistency, the ± beside your average shrinking as lapses and guesses disappear.
Two trainable skills do keep growing, and both are measurable on this site. Choice reaction improves as the mapping from signal to key becomes automatic; the decision lag genuinely shrinks with practice, because you are automating the choosing, not the seeing. And anticipation improves substantially with exposure, because it is a forecasting skill: gamers and athletes are not faster wires, they are better predictors, acting on what is about to happen. When a streamer seems inhumanly quick, you are watching anticipation and pattern knowledge, on top of hardware chosen for speed.
The transfer question matters: getting faster at one reaction game makes you faster at that game. The evidence that it generalizes to driving, sport or another task is weak, which is also the honest verdict on commercial “brain training.” Train the thing you want to be fast at.
The cheapest 30 milliseconds
No lifestyle change buys milliseconds as cheaply as hardware does. Moving from a 60 Hz screen to a 144 Hz one, or from a wireless touchpad to a wired mouse, removes real, physical delay from the measurement loop; the difference between a slow laptop setup and a fast desktop one is commonly 20 to 40 milliseconds, which is more than caffeine will ever give you. The lag your device adds itemizes where it hides. It only improves the number, not you, but if you are comparing scores with a friend, make sure you are not comparing monitors.
An honest protocol
If you want your genuine best on the test: sleep normally, take it in the late morning or early evening rather than at 3 a.m., have your usual caffeine, sit up, use the fastest screen and input device in the house, run it twice to burn off the task overhead, and keep the third run. That is your number. It will be a little worse on a phone, a little worse tired, a little better next month if you play often, and none of those changes will mean your nervous system changed.
The number’s percentile depends on a curve that is still partly an assumption, declared test by test on the Method page. The comparison this site stands behind is you against you, same setup, across days; everything in this guide moves that comparison, and nothing in a supplement aisle does.