A Seven-Session Typing Baseline You Can Reproduce

A baseline is a personal reference, not a ranking

A typing baseline answers a narrow question: what result range do you produce under a defined set of conditions right now?

It does not answer whether your speed is good for your age, profession, language, or education. Speedtype does not maintain a validated population benchmark, so this protocol avoids borrowed average-WPM claims and guaranteed improvement targets.

The goal is repeatability. If the conditions and recording method are clear, you can compare a later set with less risk of confusing a new keyboard, different passage language, or longer duration with a skill change.

Choose the conditions once

Before Session 1, record:

Condition	Your choice
Device
Keyboard or input surface
Browser
Language
Input method or layout
Difficulty
Duration
Time window

Use the same choices for all seven measured sessions. If a browser update, injury, keyboard change, or input-method change occurs, note it. A changed condition does not make the result invalid, but it can make comparison ambiguous.

For Marathi, record the exact input method. Two methods can require different composition sequences even when they produce the same visible text.

Why seven sessions

Seven is not a magic training dose. It is simply a manageable odd-numbered set that gives a clear middle observation after sorting.

One unusually high or low run can occur because of passage familiarity, interruption, momentary focus, or an awkward phrase. Selecting only the best run exaggerates that noise. Taking the middle of seven reduces the influence of one extreme without requiring statistical software.

This is a project practice protocol, not a scientific validation study.

The session structure

Use one session per measurement period. If you complete multiple sessions on the same day, keep the rest interval similar and note that they are not independent days.

For each session:

1. Check that language, difficulty, and duration match the plan.
2. Complete one unrecorded warm-up run if you need it.
3. Reset before the measured run.
4. Type normally; do not disable corrections unless that is part of every session.
5. At the result card, record all fields before starting another run.
6. Note the passage opening and any interruption.
7. Stop after the planned measured run rather than repeating until a preferred score appears.

The warm-up is optional, but its use must be consistent. Warming up in only the later sessions could look like improvement when it is a procedure change.

Seven-session record

Session	Net WPM	Gross WPM	Accuracy	Errors	Keystrokes	Passage opening	Notes
1
2
3
4
5
6
7

Use the result-card guide if any field is unclear. Gross and net WPM should be recorded separately.

Calculate the middle result

Write the seven net WPM values from lowest to highest. The fourth value is the median, or middle, net WPM for this set.

Example values:

18.8, 20.1, 20.4, 21.0, 21.3, 22.5, 24.2

The middle net WPM is 21.0.

Do the same for accuracy if you want a compact second reference. Do not combine the fourth net WPM and fourth accuracy as if they necessarily came from the same session. They are separate summaries of the set.

Also keep the full table. The range and notes can reveal information that a single median hides.

Read patterns before setting a goal

Look for observable patterns:

• Is gross WPM stable while net WPM changes with errors?
• Do low results share the same interruption or passage feature?
• Does capitalization account for repeated mismatches?
• Does one input method create composition problems?
• Does accuracy fall only when you chase a higher pace?
• Are later sessions better only because they use a familiar passage?

Turn one pattern into one next-step question. For example: "Can I keep the same net WPM while reducing unresolved case errors?" That is more testable than "type much faster."

The accuracy-first protocol offers a structured practice cycle after the baseline.

Repeat the baseline fairly

After a practice period, repeat seven measured sessions with the original conditions. Compare:

• the two median net WPM values;
• the two median accuracy values;
• the distribution of gross and net WPM;
• recurring error notes;
• any changed equipment or software.

Do not compare your original English set with a later Marathi set or a one-minute set with a later five-minute set. Those can each be useful results, but they answer different questions.

If the passage library changed between sets, record the date. The current passage process is documented in how passages are reviewed.

Avoid common measurement traps

Repeating until the number looks good: This selects a favorable outlier. Decide the number of measured runs before starting.

Changing two things at once: A new keyboard plus a new duration makes it hard to identify the cause of a difference.

Ignoring corrected attempts: A clean final text can still involve extra attempted keystrokes. Read what corrections change.

Treating rounded accuracy as exact: Displayed accuracy is rounded to a whole percent. Two equal labels can hide slightly different underlying ratios.

Using Race as the baseline: Race adds a moving bot and variable bot pace. It is useful for pacing practice, but the race implementation is not a fixed measurement opponent.

Comparing only the maximum: The highest result can be real and still be unrepresentative of the seven-run set.

What would count as evidence of change

A later median moving in one direction is an observation, not proof of cause. Confidence increases when:

• conditions were preserved;
• multiple later sessions shift rather than one outlier;
• accuracy does not collapse while net WPM rises;
• the same pattern appears in another controlled set;
• notes do not show a clear external explanation.

Even then, report the result as your own under those conditions. Do not turn one person's change into a universal claim.

Open Practice when the table is ready. If a result field behaves differently from the published scoring methodology, preserve the smallest reproducible example and report it before using the set as a baseline.