How to Read a Speedtype Result Card Without Chasing One Number

Read the card as one record

The largest number on a Speedtype result card is net WPM. It is prominent because a speed adjusted for unresolved errors is more informative than raw activity alone. It should still be read with gross WPM, accuracy, errors, keystrokes, and time.

A result is a record of one session under one passage, language, difficulty, duration, keyboard, browser, and input method. It is not a general ability certificate.

Net speed

The main Speed value is net WPM. Speedtype takes attempted keystrokes, subtracts unresolved errors, divides by five, and divides by elapsed minutes.

If a 60-second run has 95 attempts and 76 unresolved errors:

Correct attempted characters = 95 - 76 = 19
Net WPM = 19 / 5 / 1 = 3.8

The large value should therefore be 3.8 WPM, not 19.0 WPM. If a screenshot displays a different result for those exact fields, it needs investigation.

The four worked results provide calculator checks for clean and error-heavy runs.

Gross speed

Gross WPM appears as smaller supporting text. It uses all attempted keystrokes before unresolved errors are deducted.

In the same example:

Gross WPM = 95 / 5 / 1 = 19.0

Gross WPM answers "how much accepted typing activity occurred at this pace?" It does not answer how much of the resulting passage is correct.

A large gap between gross and net points to unresolved accuracy rather than lack of input activity. A small gap means few unresolved errors relative to attempts, although corrected attempts can still be present.

Accuracy

Accuracy is the percentage of attempted characters that remain correct after unresolved errors are deducted:

Accuracy = (attempts - unresolved errors) / attempts * 100

The interface rounds accuracy to a whole percent. A displayed 99% may come from 98.57%, as in 69 correct attempts out of 70.

Use accuracy to add context, not as a pass/fail label. Short runs can change by several displayed percentage points from one character. Longer runs can hide where errors cluster. The passage marking and notes remain useful.

Errors

Errors are unresolved character differences in completed words and the current entered portion. They are not a permanent count of every wrong key that was corrected.

Completed words use a character edit-distance comparison. The current word counts mismatches already entered but not the suffix still waiting to be typed. The total is clamped so it cannot exceed the currently typed text length.

Read what Backspace changes before comparing someone who corrects immediately with someone who leaves mistakes unresolved.

Keystrokes

Keystrokes on the card refers to accepted character attempts tracked by the session. Replacement characters after a correction can add to this total even when the final input no longer contains the original mistake.

The value does not count every hardware event. Modifier keys, navigation, operating-system shortcuts, and intermediate IME behavior are not all equivalent to accepted passage characters. For Marathi, the input method can make physical-key interpretation especially complex.

Time

Time is the measured elapsed duration from the first accepted keystroke to the session end. A timed run normally ends at its selected limit; a completed passage or race can end under its mode's own finish rule.

WPM uses elapsed minutes rather than assuming every displayed mode lasted exactly one minute. Before comparing results, confirm that the time and mode match.

A 30-second and 60-second run can both be converted to WPM, but they do not expose the same endurance or passage variation. Keep duration stable for a baseline.

The colored tier banner

Practice currently groups net WPM into three interface tiers and displays an encouraging banner. These tiers are product presentation cues, not validated professional or educational standards.

Do not use Gold, Silver, Bronze, Tier 1, Tier 2, or Tier 3 as certification. The thresholds do not adjust for language, age, device, accessibility needs, passage, or input method.

The numerical fields and documented conditions are the maintainable part of the result. The banner can make the dialog easier to scan, but it should not replace the measurements.

Download and Share

Download creates a PNG result card in the browser. It includes Speedtype branding, net WPM, gross WPM, and up to four supporting metrics. It does not download JSON.

The image is generated locally from the displayed result using a canvas, then saved with a .png filename. It is suitable for a personal record, but the image does not contain a cryptographic signature or independent verification. It can be edited after download.

Share uses the browser's native share support when available. Otherwise it tries to copy summary text and the site address. Review the text before posting and avoid exposing account or personal information.

History and account context

Unsigned sessions can be stored in browser-local history. Signed-in features can save configured results to the user's cloud account. Clearing browser storage, changing browser profiles, or using another device can affect local history.

A downloaded image is separate from history. Downloading does not prove that a cloud save succeeded, and closing the dialog does not publish the result publicly.

See Privacy for current storage and service disclosures.

Decide the next run from a pattern

Use this simple reading order:

1. Confirm language, difficulty, mode, and time.
2. Read net WPM and gross WPM together.
3. Check accuracy and unresolved errors.
4. Check attempted keystrokes.
5. Note the passage opening and any interruption.
6. Compare with several runs under matching conditions.
7. Choose one question for the next run.

Examples of useful next questions:

• Can the gross-to-net gap narrow at the same duration?
• Does one capitalization pattern create repeated errors?
• Does changing only difficulty alter accuracy?
• Does the same Marathi input method behave consistently?
• Does a race target make accuracy less stable than practice?

Avoid repeating the test until one high number appears. A planned set is more informative than a selected personal best. The seven-session baseline provides a table and comparison rules.

When a result is internally inconsistent

A result deserves a defect report when its fields cannot satisfy the published formulas. Include the mode, elapsed time, attempts, errors, gross WPM, net WPM, accuracy, language, and passage opening.

For example, 95 attempts, 76 errors, and 60.0 seconds should calculate to 19.0 gross WPM, 3.8 net WPM, and 20% accuracy. A 0.0 net result with those unchanged values would be inconsistent.

Submit the screenshot and fields through Feedback and Support. Do not include credentials. For direct verification, use the scoring methodology rather than relying on the tier label or a social image.