How to Improve Your Reaction Time: What Actually Works

The first time I tried to “train” my reaction time, I went at it like a gym bro adding plates. More reps, more clicks, watch the number drop. It didn’t really work the way I expected. The number moved a little, then stalled, and I spent the next year reading studies trying to figure out why. So here’s the honest version of how to improve reaction time, ranked by what actually moves the needle, with the stuff that doesn’t work flagged so you don’t waste your time.

Quick reality check before we start. Per Human Benchmark’s public statistics, a large self-reported online aggregate of 81M+ recorded clicks, the typical online reaction time sits around a 273 ms median (284 ms mean). In a clean lab a simple visual reaction is more like 200-250 ms. Most of that quarter-second isn’t your reflexes being lazy. It’s light hitting your retina, getting relayed to your visual cortex (~50-100 ms right there), your brain deciding, then a motor command traveling back out to your finger. You’re trimming one slice of that pie. The decision slice. That’s it. Keep that in your head and the rest makes sense.

sleep, first and by a lot

If you change one thing, change this. Sleep deprivation doesn’t just make you slower. It makes you inconsistent, which is arguably worse. Your fast reactions stay sort of fast and then you randomly drop a 400 ms response because your attention blinked. A tired brain lapses. Reaction tests punish lapses hard.

I notice it on myself constantly. Bad night, and my scores on the main visual test get spiky and unreliable. Good week of sleep and the whole distribution tightens up. No supplement, no gaming chair, no fancy mouse fixes a sleep deficit. Fix the sleep first. Everything below is a rounding error by comparison.

caffeine, used on purpose

Caffeine works, but understand what it’s doing. The effect is modest and variable, and it mostly buys back attention. There’s research (McLellan et al. 2014) showing caffeine around 5 mg/kg partly offsets the slowing you get from sleep loss, and the mechanism is attention, not some magic speeding-up of your nerves. So it’s a patch, not an upgrade.

For me that’s about 350 mg if I’m 70 kg. I don’t take that casually; that’s a real dose. Time it 30-45 minutes before you actually need to perform, don’t stack it on top of an already-good night and expect double the effect, and don’t use it to paper over the sleep problem from the last section. It’s a tool. Treat it like one.

warm up, then practice (the realistic kind)

Cold, you are slow. A short warm-up (a few minutes of clicking) pulls your decision step into gear and you’ll see a quick early gain. People mistake that warm-up bump for “training,” post a screenshot, and assume the trend continues forever. It doesn’t.

Practice genuinely helps, but it sharpens the decision and the stimulus-recognition part, the trainable chunk. It doesn’t speed up phototransduction or the nerve conduction in your arm. Where practice pays off most is choice reactions, because Hick’s law (RT ≈ a + b·log₂N) tells you the cost grows with the number of options. The constants (a and b, often quoted around 200 ms and 150 ms/bit) are illustrative and vary a lot across studies, not settled values. Simple reactions run ~150-300 ms; choice reactions blow out to 300-700 ms once you’re picking between things. The picking is what you can get better at. That’s why an aim trainer or a choice reaction test feels more transferable to gaming than mashing a one-button test. You’re rehearsing the actual decision, not just the twitch.

vision, consistency, and using the right modality

Two underrated levers. First, consistency beats peak. A reliable 250 ms is better for almost everything than a 210 ms that’s surrounded by 380 ms misfires. Train for a tight spread, not a one-off lucky click.

Second, modality matters and you can exploit it. Sound is faster than light: auditory reactions come in roughly 25-50 ms ahead of visual (one study had ~284 ms auditory vs ~331 ms visual) because sound transduces faster than light’s slower phototransduction. If your task can be cued by audio, lean on it. Worth feeling the difference yourself on the audio reaction test.

fitness and removing device lag

Cardiovascular fitness helps reaction time, modestly. Not nothing, not dramatic. File it under “good for you anyway, small RT bonus.”

The lag one is sneaky. Web tests add roughly 10-50 ms of display and input lag on top of your biological time. A slow monitor, wireless input, vsync: these inflate your number without your reflexes being any slower. Clean that up and you’ll “improve” instantly. It’s not really you getting faster, it’s the measurement getting honest. I’d rather know my true number, which is part of why I trust our methodology.

the part nobody wants to hear

There’s a floor, and you will hit it. Around 100 ms is the practical floor for a simple human reaction. World Athletics treats any sprint start under 0.100 s as a false start by rule, and Pain & Hibbs (2007) measured elite sprint-start auditory reactions below 100 ms in the lab, with the neuromuscular component under 85 ms. So no, you are not training your way to a 100 ms click test. If you see sub-100 on a click test, that’s anticipation (you guessed the timing), not reaction.

For context, F1 drivers average around 200 ms off race-start lights (elite 150-220 ms), Usain Bolt reacted ~0.165 s in the Beijing 2008 100 m final, actually one of the slowest starters in that field, and he won anyway. Esports pros land ~170-190 ms, regular gamers ~200 ms. Elite, full-time, genetically gifted people live in the 150-200 ms range. That viral “0.04 s Bottas” number floating around? Anticipation, not a verified reaction. Don’t measure yourself against a myth.

Age is the other immovable thing. The MindCrowd study (Talboom et al., npj Aging 2021, n=75,666) found reaction slowing on the order of 7 ms per year, though the curve decelerates rather than adding up linearly, leaving the 70s roughly 40-60% slower than the early-20s peak. You can offset some of that with sleep and fitness. You can’t reverse the calendar.

so what do I actually do

Sleep like it’s your job. Caffeine on the days that matter. Warm up, then practice the decision with choice-based drills. Tighten your consistency, fix your gear, stay reasonably fit. Realistic outcome: you shave a chunk off the soft, attention-driven part of your time and you stop throwing away clicks to lapses. Modest, real, and honestly enough to win a duel.

If you want the deeper why behind every number above, I broke it all down in the factors affecting reaction time. Otherwise, go warm up and run the test. Your first cold score doesn’t count.

Sources

Written by Lokesh Rathore, Founder of ReactScore. Lokesh Rathore is the founder of ReactScore and the person behind everything on it. Reflexes and reaction time are a genuine fascination of his, so he built the test he wanted to use himself: one that measures honestly, corrects for the lag your screen and mouse add, and explains the numbers instead of hiding them. He keeps every figure on the site grounded in public datasets and published research. More about the author · how we measure →