The science behind Smartypes

Short, daily practice

In a classic study, postal workers learning a keyboard skill trained on one of four schedules. Those who practiced one hour a day learned faster per hour of practice, made fewer errors, and remembered more months later than groups who practiced longer or more often each day, even though some of those groups practiced more hours in total (Baddeley & Longman, 1978). A later meta-analysis found the same direction across many studies, with the largest benefit of spacing for simple motor tasks (Donovan & Radosevich, 1999).

To make that choice checkable, Smartypes measures you the same way every session and gives you a progress report at 7, 14, 30, 60, and 100 practice days, comparing your first opening check with your latest. Badges count practice days rather than punishing a broken streak: in a study of how habits form, missing a single day did not materially affect the process (Lally, van Jaarsveld, Potts & Wardle, 2010).

The opening check: diagnosis from timing

Every session starts with about a minute of typing. The Coach records when each key goes down and comes up, and diagnoses from those measurements rather than from how you describe yourself. Keystroke research explains why the timing matters:

• Speed varies enormously. Across 168,960 people typing online, the average was about 52 words per minute; the slowest tenth typed under about 26 and the fastest tenth over about 78 (Dhakal, Feit, Kristensson & Oulasvirta, 2018).
• Rhythm is a mark of skill. Fast typists are far more consistent from one key to the next, not just faster on average (Dhakal et al., 2018). Smartypes reports a consistency score for this.
• Overlapping keys. Pressing the next key before releasing the last one was the strongest timing predictor of speed in that data set (Dhakal et al., 2018). Smartypes measures this rollover on a physical keyboard.
• Some errors matter more. Wrong keys and skipped letters were more strongly linked to lower skill than extra letters (Dhakal et al., 2018), so the Coach weights them more.
• Your fingers notice errors first. Typists slow down right after a real error, even when they do not consciously notice it (Logan & Crump, 2010).
• Recovery from a miss. Because typists slow down after an error, the time a miss costs is worth watching. Smartypes times, for each text, how long it takes from a wrong key to the next right one, and whether the key after that is still slow, always against your own usual keystroke and your own earlier texts (Logan & Crump, 2010). The Progress page shows the trend for your home row drills, with tips for a quicker recovery. Smartypes does not claim that its tips shorten recovery time: the numbers are there for you to watch, not a promise that any tip works.
• A mistake costs more than the wrong key. In one study of 18 skilled typists (about 80 words per minute), the key before a mistake took about 154 milliseconds, the Backspace that fixed it about 416 and the first correct key after it about 255 (Crump & Logan, 2013). When 800 typists could not use Backspace, the key right after a mistake was still slower, about 517 milliseconds against 180. The Progress page shows how much of your typing time goes on fixing mistakes, averaged over your last five opening checks.
• Letter pairs are not equal. Pairs typed with the same finger are naturally slower than pairs that alternate hands. The Coach judges each pair against what is expected for its finger pattern and how common it is in English before calling it slow (Salthouse, 1986; Behmer & Crump, 2016).
• The first letter of a word includes planning time, so it is not counted against the key itself.

From these, the Coach decides what limits you today: finding keys, accuracy, rhythm, letter pairs, speed, endurance, or technique (such as pressing Shift with the same hand as the letter). A level check, the full version of the opening check, places new learners by what they demonstrate, and you can take one any time you feel you have outgrown your practice.

How the plan is built

• The right difficulty changes as you improve. Practice that is too easy or too hard slows learning, and the best level of challenge shifts with skill (Guadagnoli & Lee, 2004). The Coach sets targets just beyond your measured level and pushes speed only while accuracy holds.
• New keys first, mixing later. Switching between different drills can help retention, but the benefit is smaller outside the lab and close to zero for beginners (Shea & Morgan, 1979; Brady, 2004). So beginners practice new keys on their own, and drills are interleaved with real text only past that stage.
• Guidance fades. Constant feedback helps while it is there but can get in the way of building an automatic skill (Salmoni, Schmidt & Walter, 1984; Winstein & Schmidt, 1990). The on-screen key guide lights a key only while you are still learning it, or when you hesitate or miss, and live numbers are hidden while you type unless speed is the point.
• Eyes on the screen. Learners drift back to looking at their hands because it pays off in the moment, even though touch typing pays off later; an on-screen task that rewarded keeping eyes on the screen helped (Yechiam, Erev, Yehene & Gopher, 2003). The Eyes up block works on the same idea.
• Technique is a means, not a test. Typists who did not use classic ten-finger technique could still be fast and accurate; consistent finger-to-key mapping and looking ahead mattered more (Feit, Weir & Oulasvirta, 2016). Smartypes teaches touch typing to beginners but rates everyone by what they achieve.

Careful practice, and accuracy

Careful, methodical practice is a sound way to work on accuracy and control, and how you practise, and how fast, is always your choice. Three findings sit behind that. A mistake costs more than the wrong key, as above. A consistent finger for each key was one of the things that went with high typing performance in a study of 30 everyday typists, along with looking ahead and moving the hands as little as possible (Feit, Weir & Oulasvirta, 2016). And in laboratory studies of a movement task, practice moved the whole trade-off between speed and accuracy, mostly by making movements more consistent from one attempt to the next (Shmuelof, Krakauer & Mazzoni, 2012). That is why the Progress page looks at time lost to mistakes and at consistency, not only at speed. Careful practice is offered for accuracy and control, which are worth having for their own sake; Smartypes does not claim it is a faster route to speed. In the one direct comparison we read, slow, medium and fast practice on a movement task gave the same gain, and we found no typing study that compares them.

What "expert" means here

Smartypes defines Expert as 80 or more words per minute at 97% or better accuracy on unfamiliar text with capitals and punctuation. In the large online study above, that speed would place someone in roughly the fastest tenth. Tiers below it (Steady, Capable, Proficient, Advanced) mark the way. Progress in skills like this is usually fast at first and slower later, which is normal (Newell & Rosenbloom, 1981). Typing is a consistent, self-paced skill, the kind of task where structured practice has shown its most reliable payoff (Macnamara, Hambrick & Oswald, 2014).

Above Expert is the Upper Echelon: Master at 100, Maestro at 125, Virtuoso at 150, Titan at 175 and Fabled at 200 words per minute, at the same 97% accuracy. Each rank is defined by those two bars alone, and Smartypes makes no claim about how many typists reach them. At these speeds a step of 25 words per minute can take a long time, so Home shows how far you are between one rank and the next, along with your best opening check.

Learning words while you type

The vocabulary block teaches words in a sequence built from memory research: read the word, its meaning, and an example; type it in two different sentences; then type it from memory later, from a definition or a sentence with a blank; and meet it again after growing gaps of days.

• Producing a word, including typing it, helps you remember it better than reading it silently (MacLeod, Gopie, Hourihan, Neary & Ozubko, 2010; Forrin, MacLeod & Ozubko, 2012).
• Testing yourself beats re-reading for long-term memory, even though re-reading feels more effective at the time (Roediger & Karpicke, 2006; Karpicke & Roediger, 2008).
• Filling in a blank helps more than copying (Slamecka & Graf, 1978).
• Spreading reviews over days and weeks keeps words far longer than cramming (Cepeda, Vul, Rohrer, Wixted & Pashler, 2008; Bahrick, Bahrick, Bahrick & Bahrick, 1993).
• Seeing a word in several different sentences builds its meaning better than one sentence (Bolger, Balass, Landen & Perfetti, 2008).
• Word parts (roots, prefixes, suffixes) help with learning words, especially for younger readers (Bowers, Kirby & Deacon, 2010; Goodwin & Ahn, 2013).

New words are read first, with no typing, so the effort of typing does not crowd out the meaning. That is a design precaution drawn from research on divided attention, not a tested result. In the one study we found that compared typing with handwriting for learning new words, people remembered them about equally well either way (Ihara et al., 2021).

The Coach

The Coach is a small language model and a set of rules that run entirely in your browser. The language model learned word patterns from the sentences and word list written for Smartypes, and it uses them to judge how hard a key sequence is and to make practice drills for new keys. It does not write sentences: every sentence you type was drafted with AI help and approved through a review process that includes human review (see How Smartypes was made, below), and the Coach chooses which ones to show. It only shows sentences you can type with the keys you have unlocked, it leans toward the keys and letter pairs you need, and it avoids repeating itself. It is not a general chatbot, and nothing you type is sent anywhere. Every adjustment it makes moves a fixed fraction of the way toward what it measured, and level changes need repeated evidence, so your practice settles instead of swinging back and forth.

How it is designed

The interface and the Coach share one theory of action: Don Norman's seven stages of action and Norman's account of human error (Norman, 2013). Each keystroke, text, block, session, and week follows the same loop: a goal and a plan, the action, then feedback you can interpret and compare with the goal. Norman's study of action slips drew partly on typing errors (Norman, 1981; Rumelhart & Norman, 1982), and the Coach uses the same distinction: a slip is a known key hit badly, and a gap is a key not learned yet. They call for different practice.

How Smartypes was made

Smartypes is an independent project by Steven Quinn Singleton, built with Claude, an AI assistant made by Anthropic. Claude did much of the drafting: the software, most of the practice sentences, the word lists and the explanations on this site. The direction is human.

That human part is where the value for you comes from. Steven decides what Smartypes should teach, in what order, and what it should never promise. Steven practices with it personally, so the problems that get fixed are ones a learner would feel, not only ones a measurement would show. Practice sentences are read by hand, sampling the pool and correcting what is found, because whether a sentence is natural to type, clear and fair to a learner is a judgment an automated check cannot make. The corrections made by hand are the ones in use. And because one named person stands behind the site, there is someone to tell when something is wrong.

Checks work alongside that judgment: mechanical checks for typing slips, a review by separate AI reviewers, and a check of factual claims against cited sources. As of September 2026 that source check covers most of the factual sentences and is continuing. Sentences that a check contradicted were corrected, and sentences it could not confirm were taken out. Errors can still get through, so if you find one, please say so on the feedback page and it will reach the person who can fix it.

The research summaries on this page were prepared the same way. Each cited paper was checked to exist, but a summary can still misstate a paper, so follow the links to the originals for the full findings.

What Smartypes does not claim

• No guaranteed speed or timeline. People improve at different rates; Smartypes is a practice tool, not a promise. Your own progress reports show what your practice is producing.
• That 15 to 30 minutes a day is a scientifically proven dose. It follows the direction of the research, below the range that was tested.
• That any single accuracy percentage is the scientifically optimal level of difficulty.
• That making the cursor wait for the right key works better than free typing, or the reverse. We found no direct evidence either way; Smartypes uses the waiting cursor so measurements stay clean, and offers free typing as a setting.
• That an on-screen keyboard or blank keycaps are proven to speed learning. No controlled study was found either way.
• That typing beats handwriting (or the reverse) for learning vocabulary.
• That rest breaks, sleep, or a sense of control make you learn faster. Recent, more rigorous studies have not supported the strong versions of these claims; breaks between blocks are there for comfort.
• That the recovery tips shorten your recovery time. Recovery is measured so you can see your own trend; Smartypes does not promise that any tip changes it.

Research is summarized in plain language; follow the links to the original papers for the full findings. Nothing on this site is medical advice. If typing causes pain, stop and consult a health professional.