Starting the next key before releasing the last. It correlates +0.73 with typing speed across 136 million keystrokes — second only to the gap between keys itself. Slow typists roll 8% of their keystrokes; fast typists roll 40 to 70%. Formal touch-typing instruction barely moves it, and no mainstream tutor measures it. This sheet checks whether we can.
Source — Dhakal, Feit, Kristensson & Oulasvirta, CHI 2018 · n = 168,960 · 136,857,600 keystrokes
Free practice, measured. If you want to be told what to work on instead, today’s session is ten minutes and ends by itself. And if you are starting from zero — still looking down to find the keys — the method begins at Hands, not here.
Mode
Specimen
Same-hand inward rolls. Let each group become one gesture — the release of the first key should happen while the next is already travelling.
One click per keystroke. Set it below your comfortable speed — chasing a fast click moves the variance rather than removing it.
FIG. 1—Rollover ratio
r = +0.73 with wpm
––%
0
08SLOW
25MEAN
40FAST
70
100
AWAITING INPUT
Mean overlap—
Population ≈ 30 ms when rollover occurs
Rolled—
Keystrokes begun before the previous key was released
FIG. 2—Keystroke trace
500 ms / div · keydown→keyup
no signal
rolleddiscrete
TAB. 1—Measurements
no run
Net wpm—
Gross wpm—
Mean iki—
Speed restated: wpm = 12000 / iki
Consistency—
Coefficient of variation · target ≤ 0.45
Accuracy—
Floor 97%
Kspc—
0 corrections · fast typists 1.10, population 1.17
Hold time—
≈116 ms for fast and slow alike — not a lever
Alt benefit—
Same-hand minus alternating · positive means hands work in parallel
Keys held0
Hold several keys at once. If this stalls at two or three, the keyboard is dropping simultaneous presses and caps measured rollover regardless of technique.
Paused—
Gaps over three seconds stop counting against your speed. Reading ahead is not slow typing, so it is not charged as such.
Run complete—
Finish a run and it is recorded for the diagnosis.
IntegrityClean
Runs are flagged, never silently voided — you decide what a marked result is worth.