Methodology
How Polytaipe measures typing
Every number Polytaipe shows you (WPM, accuracy, consistency, rollover, per-key latency) has an exact definition. This page documents those definitions and the research each one is built on, so you can trust the numbers or check our work.
Exact definitions
What is WPM (words per minute)?
Polytaipe computes WPM the standard way: characters typed divided by 5, per minute. A "word" is defined as 5 characters, not a dictionary word. We report two numbers: raw WPM counts every keystroke, including ones you later correct; net WPM only counts correct characters, so an error costs you speed, not just accuracy.
What is accuracy?
Accuracy is the percentage of keystrokes that matched the expected character: correct characters divided by total keystrokes typed, expressed as 0–100%. It's a simple ratio, but it's the single input the research most consistently says to optimize first, and accuracy-prioritizing typists improve speed roughly 19% faster than speed-first ones.
What is consistency?
Consistency measures how steady your pace is within a session, not just your average speed. We compute the coefficient of variation (CoV) of your per-second raw WPM across the session, then invert it into a 0–100 score. A high score means your rhythm barely wavered; a low one means bursts and stalls.
What is rollover?
Rollover is the percentage of keystrokes where you pressed the next key before releasing the previous one, measured from key-down and key-up timestamps, not just key-down order. Dhakal et al. (CHI 2018) found the fastest typists in a 136-million-keystroke dataset rolled over on 40–70% of keystrokes; Polytaipe tracks it per session.
What is the error taxonomy?
Every mistake gets classified into one of four types, following Soukoreff & MacKenzie's CHI 2003/04 framework: substitution (wrong character), insertion (extra character), omission (skipped character), and transposition (two characters swapped). We also split errors into corrected (fixed before the session ended) and uncorrected, the harsher signal.
What is per-key latency?
Per-key latency is how long, in milliseconds, you take between key presses, tracked individually for every physical key, not just averaged across the whole keyboard. It's what lets Polytaipe tell you that your left pinky on Q is genuinely slow, instead of just telling you your overall WPM is a bit low.
What does the research say about improving typing skill?
Three studies anchor Polytaipe’s design: a 136-million-keystroke analysis of what separates fast typists from slow ones, a study of what actually predicts typing skill beyond finger count, and a decades-old distributed-practice experiment on how training time should be spread out. None of it is folk wisdom. All of it is cited below.
Dhakal et al., CHI 2018: 136 million keystrokes, 168,000 typists
The largest keystroke dataset ever analyzed found an average net speed of 52 WPM, that the fastest typists roll over 40–70% of their keystrokes, that consistent inter-key timing beats raw bursts, and that accuracy-prioritizing typists improve their speed roughly 19% faster than speed-first typists. Cross-hand letter pairs were the strongest predictor of skill.
Feit et al., CHI 2016: "How We Type"
Ten fingers aren't required to type well. This study found that anchored hands, a consistent finger assigned to each key, and not looking down at the keyboard predicted typing skill better than adherence to strict ten-finger touch-typing form, which is why Polytaipe doesn't force a single "correct" finger dogma.
Baddeley & Longman, 1978: distributed practice
This classic training-schedule experiment found that 1 hour of daily practice produced better typing-skill gains than 2- or 4-hour sessions at the same total volume. Spaced, shorter sessions beat marathons. That's one of the reasons Polytaipe is built around 15–30 minute loops instead of long grinding sessions.
Why the design stays out of the way
Polytaipe ships one tuned look, ink and paper plus a single functional hue, instead of a theme store, because how a practice line is presented measurably affects how quickly you perceive its letters. Stray color competes for attention, clutter increases crowding, inconsistent fonts slow letter recognition, and unexpected motion pulls your eye off the task. None of this is folk wisdom either. It’s the same standard of evidence as the typing-skill research above, applied to the interface itself.
Theeuwes, 1992: attentional capture by color singletons
A single out-of-place color measurably slows a visual task you're focused on, because the visual system orients to it involuntarily. In the trainer, color appears only where it is the information: rust means "fix this" (mistyped characters, weak keys, hot zones on the heatmap), and nothing else on screen carries a hue.
Pelli & Tillman, 2008, Nature Neuroscience: crowding
Letters get harder to recognize when flanked by clutter; visual crowding is often a bigger limit on letter recognition than raw sharpness. So the practice line stays centered, spacing stays generous, and nothing sits in a side panel competing with it while you type.
Sanocki, 1987/1988; Sanocki & Dyson, 2012: font tuning
Your visual system tunes to a font's specific letterforms over repeated exposure, and breaking that consistency slows letter identification. Practice text in Polytaipe is always the same monospaced face at a stable size. There is no font picker, because switching fonts would reset that tuning every time.
Zhai et al., 2023, Applied Sciences; Snyder, Logan & Yamaguchi, 2015: typing-specific evidence
Direct research on typing interfaces is limited, but what exists points the same way: legibly sized characters help, and heavy color-coding can slow you down by adding a "find the color, then read it" step before you can act. Clean, immediate feedback, not decorative feedback, is what matters for catching errors as they happen.
Abrams & Christ, 2003: attentional capture by motion onset
New motion in the visual field pulls attention automatically, even when you're trying to ignore it. So while a session is live, the only thing that moves on screen is the caret: no confetti, no shakes, no pulsing stat, nothing sliding in from the side.
Rey, 2012; Sundararajan & Adesope, 2020: the seductive-details effect
Two meta-analyses of the learning-science literature converge on the same finding: decorative-but-irrelevant additions to an instructional interface reliably hurt learning outcomes, because they consume limited working memory that would otherwise go toward the task. A typing trainer is a skill-acquisition interface first, and the same logic applies.
Light or dark? We stay honest.
For fine text in a lit room, dark-on-light has a small repeatable edge in proofreading and acuity, largely because a brighter screen constricts the pupil (Piepenbrock, Mayr & Buchner, 2013–2014). Dark screens reduce glare in dim rooms and some readers prefer them (some with astigmatism find bright-on-dark harder). Neither is universally better, so the reading pages are paper, the trainer is ink, and we claim no speed magic for either.
Dozens of themes are fun to use. Monkeytype does them brilliantly, and we say so without irony. We traded that fun for signal on purpose. Personalizing the practice surface itself (theme packs, custom fonts, color pickers on the typing line) is a stated non-goal, not a missing feature waiting on the roadmap. The one planned exception is accessibility: contrast and size adjustments, because access is not decoration.
Why we built this
We didn’t start from a feature list. We started from the complaints. Typists across Reddit, Hacker News, and app-store reviews describe the same failures: plateaus nobody diagnoses, drills that don’t teach, practice that doesn’t transfer to real writing or code, and, for anyone learning a new layout, abandonment once a fixed lesson set runs dry. Polytaipe exists to close that loop honestly, without ever claiming a layout switch will make you faster when the evidence says it won’t.