Aim Trainer

This aim trainer asks you to click drifting targets for 30 seconds, then scores the run the way a human-computer interaction lab would: not just as a hit count, but as throughput measured in bits per second. That single number fuses your speed and your accuracy, and it stays comparable whether you played on big slow targets or small fast ones — which is exactly what a raw "milliseconds per target" score can never do.

Why Not Just Count Hits?

Almost every aim trainer reports two numbers: average time per target, and accuracy. Both are easy to read and neither means much on its own.

• Time per target depends on the targets, not on you. Make them bigger or closer and everyone gets faster. A 400 ms score on huge targets is worse aim than an 800 ms score on tiny ones, but the leaderboard cannot tell.
• Accuracy and speed trade against each other. Any player can push accuracy to 99% by slowing down, or halve their time by spraying. Reporting them separately lets you flatter whichever one you are chasing.
• Neither is comparable across sites. Two trainers with different target sizes produce numbers that simply cannot be put side by side.

Throughput fixes all three at once, which is why ISO 9241-9 — the international standard for evaluating pointing devices — is built on it rather than on raw times.

How Fitts's Law Works

In 1954 Paul Fitts showed that the time to point at a target does not depend on distance and width separately, but on their ratio, and only logarithmically. In the Shannon formulation used here, the difficulty of a movement is:

\( ID = \log_2\!\left(\frac{2D}{W} + 1\right) \) bits

where \(D\) is the distance your cursor travels and \(W\) is the target's width. The units are genuinely bits, in the information-theory sense: hitting a small target far away delivers more information than slapping a big one nearby, in the same way that picking one option out of sixteen conveys more than picking one out of two.

Movement time then rises linearly with difficulty, \(MT = a + b \times ID\), so dividing difficulty by time gives a rate that is independent of the targets themselves:

\( TP = \dfrac{ID}{MT} \) bits per second

This is the property that makes the whole tool work. Our three modes differ by more than a full bit of difficulty per target, yet a player of a given skill scores almost the same throughput in all three — Casual, Standard and Precision all land within 0.01 bits/s of each other for the same ability. If your own score swings a lot between modes, it usually means one of them is pushing you off your natural speed-accuracy balance.

How Accuracy Gets Folded In

Here is the elegant part, and the reason accuracy is not shown as a separate grade. Scattering your clicks around a target is mathematically indistinguishable from aiming carefully at a bigger target. So ISO 9241-9 throws away the target's nominal width and substitutes the effective width \(W_e\), computed from where your clicks actually landed:

\( W_e = 4.133 \times SD_x \)

where \(SD_x\) is the standard deviation of your click endpoints along the direction you were moving. The constant 4.133 is chosen so that \(W_e = W\) exactly when 4% of your clicks miss — the error rate Fitts's original task was calibrated to.

The consequence is that speed and accuracy stop being a tradeoff you can game. Spray, and your endpoints scatter, \(W_e\) grows, and each target delivers fewer bits. Slow down to never miss, and \(W_e\) shrinks but \(MT\) grows faster. One number, no dial to fiddle with.

When you play the arena above, this is measured from your real click endpoints. When you type a score in by hand there are no endpoints to measure, so the tool recovers an equivalent spread from your error rate instead, using Crossman's correction \(W_e = W \times 2.0665 / z\) with \(z = \Phi^{-1}(1 - e/2)\).

What Counts as a Good Score

The median mouse user lands near 4.0 bits/s here. That is a little under the ~4.5 bits/s that published ISO 9241-9 studies report for a mouse, because those use stationary targets — ours drift while you move to them, so your correction sub-movements cost roughly 12% extra time.

Is It Better to Be Fast or Accurate?

Mostly, it does not matter — and that is the design working rather than a shrug. Sweep a single player from a 4% error rate all the way out to 30% and their throughput moves by under half a bit, because the speed they gain is paid straight back in effective width. Throughput measures your pointing capacity, not the strategy you happened to pick on the day.

There is one real exception: over-aiming genuinely costs you. At close to 100% accuracy you are creeping onto every target, and the time that costs grows much faster than the accuracy repays. The same player easing back to roughly a 4% error rate typically gains a full bit per second. If you never miss, you are being too careful.

How to Actually Improve

• Kill the correction phase. Most wasted time is not the flick — it is the overshoot and the creep back. Aim to land on the target once rather than converging on it.
• Lead moving targets. Aim where the target is going, not where it is. This is worth more on Precision than any raw speed.
• Move from the elbow, not the fingertips, for anything more than a short distance. Fingertip flicks are precise but slow over range.
• Fix your sensitivity and leave it alone. Fitts's constant \(b\) is essentially a learned motor calibration; changing sensitivity resets it.
• Play in a large window with a wired mouse. A small arena compresses \(D\) and \(W\) together, and wireless adds latency you pay on every single target.
• Track throughput, not time per target — it is the number you cannot fake by choosing easier targets.

Frequently Asked Questions

Measured as throughput, the typical mouse user scores around 4.0 bits/s on this moving-target task. Above 4.4 beats most people, 5.0 to 5.8 is the range trained esports players reach, and below 3.2 usually means a trackpad or an unfamiliar setup. Raw time per target is a poor way to compare scores, because it depends entirely on how big and how far away the targets were.

Because milliseconds are not comparable. Big, close targets produce fast times that say nothing about aim, and you can always trade accuracy for speed to flatter whichever number you are chasing. Throughput divides the difficulty of the movement, in bits, by the time it took. The difficulty cancels out, so the same player scores roughly the same on easy and hard targets — which is exactly why ISO 9241-9 standardised it for comparing pointing devices.

Published in 1954, Fitts's Law says the time to point at a target rises with the logarithm of the distance divided by the target width. In the Shannon form used here, \(ID = \log_2(2D/W + 1)\), measured in bits. A target twice as far away, or half as wide, is exactly one bit harder. Movement time rises linearly with \(ID\), so dividing \(ID\) by movement time gives a size-independent rate: throughput.

It is folded into the difficulty rather than scored separately. Scattering your clicks is mathematically identical to aiming at a bigger target, so the real width is replaced by an effective width taken from where your clicks actually landed. Missing more widens that effective width, which lowers the bits each target is worth and therefore your throughput. Speed and accuracy end up as one number that cannot be gamed against itself.

Barely — and that is the whole point. Casual, Standard and Precision differ by more than a bit of difficulty per target, but throughput divides by that difficulty so it cancels out. A player who scores 4.0 bits/s on Casual should score close to 4.0 on Precision. A large swing between modes usually means one of them is pushing you off your comfortable speed-accuracy balance.

Published ISO 9241-9 studies put a mouse near 4.5 bits/s, but those use static targets. Ours drift while you move to them, so your correction sub-movements cost extra time and typical throughput lands about 12% lower, near 4.0. Display refresh, wireless lag, a trackpad instead of a mouse, and a small browser window each cost measurable bits too.

Yes. Type its average time per target and its accuracy into the form and pick the difficulty mode whose targets are closest in size. The tool will reconstruct the geometry and score it. It will be less precise than playing here — a typed score has no click endpoints to measure the effective width from, so the tool falls back to estimating the spread from your error rate — but it is enough to place you on the curve.

Yes, considerably more than reaction time responds to training. Most of the gain comes from cutting the correction phase: landing on the target in one movement rather than overshooting and creeping back, and leading moving targets rather than chasing them. A consistent sensitivity, a wired mouse and a larger window all help. Track throughput rather than time per target, because throughput is the number that cannot be faked by picking easier targets.