Symbol Typing Test

What you will actually be typing

Short clusters of punctuation, generated fresh each run. Below is a captured run at each level, exactly as the generator produced it, with what that level measures underneath.

These are not fragments of code and they are not meant to be. They are random draws from a character set, which is the point: with no syntax to recognise, nothing carries you through a cluster and every key has to be found on its own. What that costs you in realism, it buys back in isolating the reach.

Who this test is for

Programmers first, who type more punctuation than prose and pay for every hesitation on a brace. Then anyone whose text is dense with operators or delimiters: people writing spreadsheet formulas, building queries, editing configuration files, working in Markdown, or setting and retyping passwords full of special characters. And anyone who has noticed that their typing speed is respectable right up until a bracket appears.

It suits a specific diagnostic use too. Because it strips out letters and digits entirely, a poor score here is unambiguous — it is a Shift and reach problem, not a vocabulary or rhythm problem, and it points at a small number of keys you can fix directly.

The difficulty levels are not a ladder

On every other drill here, each level is a harder version of the one before it. On this one that is not true, and the measurements say so plainly. Two of the three things that make symbols hard actually decrease as you move from Easy to Hard.

The explanation is in what each level draws from. Easy uses a curated list of ten common symbols, nine of which need Shift and nine of which sit on the number row — so it is almost pure Shift-chord practice on a single row, held at arm’s length from home. Medium and Hard draw on wider sets that include unshifted characters like -, =, /, ; and the square brackets, which dilutes the Shift load and, in Hard’s case, pulls characters down onto the top, home and bottom rows.

So Hard is harder in a different currency. Its clusters are longer, and its characters are scattered: the row-change rate between adjacent keystrokes climbs from 18% to 62.8%, the highest of any text on this site. Easy trains the chord. Hard trains the travel. Neither contains the other, and treating Easy as a warm-up for Hard will leave the Shift work under-practised.

Every punctuation key, and what it is for

Symbols are the only part of the keyboard where people genuinely do not agree on the names. The column below gives each character’s normative Unicode name alongside what people actually call it, because the two often differ and the official one is frequently a surprise — the forward slash is a solidus, the underscore is a low line, and the full stop is not called a period by the standard at all.

This assumes a US QWERTY layout. The characters are the same on every layout; the keys that produce them are not. Which Shift to hold is covered on the capitals and punctuation stage, which teaches the habit rather than listing it — the column here is a lookup for when you already know the rule and just need the key.

Read down the “slips into” column and a pattern emerges: almost every common symbol error is a Shift failure on a key you found correctly. You hit the right key and released Shift a fraction early, so ! came out as 1, : as ;, ? as /. That is a timing problem rather than a location problem, and it responds to slowing down by a few percent rather than to more repetition. The exceptions are the genuinely confused pairs — backtick against apostrophe, backslash against forward slash — where the fix is knowing which is which.

What this drill leaves out

The generator draws from a fixed list of 22 characters, which means ten of the thirty-two punctuation keys never appear. That is worth stating rather than hiding, because the omissions are not random and several of them matter a great deal.

We can put a figure on it. Measured across every coding corpus on this site — 320,341 characters of real source, which runs 17.69 punctuation marks per 100 characters — the symbols this drill can produce account for 69.5% of the punctuation that code is actually made of. The rest is missing here.

Together those characters carry 29.9% of the punctuation in that corpus, and the double quote alone is one of its most common characters of any kind. A drill that never produces a quotation mark is not a complete preparation for typing code, and we would rather say so here than let you find out later.

One limit on all of this: the corpus being measured is our own, eight languages of passages written for this site, and it leans towards markup and typed JavaScript. That is why the angle brackets rank so highly. It describes the code we wrote, not all code — but it is real code, measured rather than guessed, and the omissions it exposes would show up against any corpus.

How this is scored

The site formula applies unchanged: five keystrokes make a standard word, and a cluster earns credit only if every character in it is right. Two things about symbol text interact with that in ways worth knowing.

First, Shift is free. The score counts characters produced, not keys pressed, so a shifted symbol costs you two key presses and earns you one character. On Easy, where 90.1% of characters are shifted, you are doing close to twice the physical work per credited character that a prose test asks for. That alone explains much of the gap between your symbol speed and your normal speed, and it is not something to correct for — every typing assessment scores this way.

Second, the clusters are short. An Easy cluster averages 1.6 characters, so with the space that commits it, one attempt is 2.6 keystrokes against the five in a standard word — which makes a wrong cluster cost more than it contributed. That arithmetic is set out in detail over on the alpha-numeric drill, which runs into the same wall; the upshot here is that an Easy score is dominated by accuracy and tells you very little about pace.

Where to go next

This drill is narrow on purpose. It fixes individual keys, and it is deliberately not the place to build the patterns those keys appear in.

If you only do one thing after a run, read the heatmap and check whether your worst characters are shifted ones. If they are, the problem is the chord timing and Easy is where to work on it. If they are unshifted characters like the brackets and the slashes, it is location, and Hard is the level that drills it.

Frequently asked questions

Sources and method

The character names in the reference table are the normative Unicode names, read directly from the Unicode Character Database rather than from any secondary list. Where a character has a widely used everyday name that differs, both are given — the standard is authoritative about what a character is, not about what people call it in conversation.

The keystroke and finger columns describe a US QWERTY layout and the finger assignments conventionally taught for it. No standards body publishes those assignments, and other layouts place several of these characters elsewhere, so treat that pair of columns as a well-established convention rather than a specification. Uses and common slips are drawn from where these characters appear in programming, markup and everyday text; they are editorial judgements, not measurements, and are presented as such.

Everything numeric is measured. The drill’s Shift load, cluster length, row placement and row-change rates come from scripts/token-stats.mjs, which reproduces the engine’s generator and samples 500,000 clusters per level from a fixed seed at build time. The coverage figures compare that character set against punctuation counted across our own coding corpora — 320,341 characters of source we wrote for this site. That corpus is ours and leans towards markup and typed JavaScript, which is stated in the section itself rather than buried here.

1. Unicode Character Database — UnicodeData.txt, the normative character names for U+0021 to U+007EThe Unicode Consortium · link checked 4 Sept 2026