What your typing score measures.
A useful score needs a clear definition. TypeScore Lab counts text output, keeps incorrect attempts in accuracy after correction, and compares sessions with matching settings.
WPM and raw speed
One standard word is five text units, including spaces and punctuation. A text unit here is a Unicode grapheme cluster, segmented by the browser after NFC normalization. It can contain more than one code point. The reported WPM is correct text units divided by five and then by elapsed minutes, rounded to the nearest whole number.
For example, 300 correct units in 60 seconds gives 300 ÷ 5 ÷ 1 = 60 WPM. If the current text contains 310 units but only 300 match their target positions, raw speed is 62 WPM and correct speed remains 60 WPM. Raw speed counts the current text, not every physical key ever pressed.
Accuracy and corrections
Accuracy is correct committed attempts divided by all counted committed attempts, expressed as a percentage. Incorrect attempts remain counted after you fix them. If the target is abc, entering x, deleting it, and then entering abc gives three correct attempts out of four: 75% accuracy. The final text is correct, but the attempt history records the mistake.
Partial combining sequences can remain pending until completed, abandoned, or the test ends. Corrections count removed or replaced text units, so they are not always identical to the number of Backspace presses. An IME replacement can change several units in one operation.
IME composition and elapsed time
Unfinished composition is not graded as if it were final text. Your system IME controls candidate selection, and committed output is compared with the prompt. Once a test starts, time continues while you compose, select candidates, pause, or correct a word. This measures the completed task, including those decisions.
Why languages should not share one ranking
A Hangul syllable, a Latin accented letter, and an Indic consonant cluster can each take different numbers of keystrokes. Chinese candidate selection also involves work that a direct letter layout does not. Five grapheme clusters are a consistent reporting unit here, but not equal physical effort or five dictionary letters in every script.
Compare the same language, input method, input source, mode, and test length. Results using the current scoring version are kept separate from older results when calculating comparable progress. A new layout deserves its own baseline.
What key and finger reports can tell you
Committed text does not reveal every physical key used by a native IME. System-input reports should be read as character performance. Where a selected direct layout supports a single-key mapping, finger information is a layout-based estimate rather than a recording of which finger you used.
A practical way to track improvement
- Use the same language, keyboard, mode, and duration for a small set of tests.
- Review accuracy and corrections alongside speed.
- Choose one recurring error and practise it in the relevant language lesson.
- Repeat the comparison after several practice sessions. One unusually fast result is less informative than a steady pattern.
If the displayed result does not fit these definitions, report it with the test settings and a short example.