Typing race — personal ghosts, accuracy & weak-pair practice

Correct-character WPM counts currently matching target characters, divides by five and by elapsed minutes. Gross WPM counts all inserted characters. Accuracy is correct insertions divided by all insertions, so an error remains in that denominator after you correct it. Each metric is labeled explicitly.

This release uses matching replay files, not live rooms. Share the challenge link and exchange a completed ghost JSON yourself. Both runs must use the exact same passage and settings. Imported records are unverified local practice data.

No. It is clearly labeled as a generated 40 WPM pacing guide. You can replace it with your previous matching run or an imported matching record. No fabricated player identity or leaderboard is shown.

The current matching-character count is recalculated, but the original wrong insertion remains in accuracy and mistake counts. Expected keys and adjacent expected pairs involved in errors are recorded. This separates final text quality from the effort needed to correct it.

The tool selects the most-missed expected two-character pairs and finds words from the current original passage containing them. It builds a local repeated practice passage. That drill has a different passage fingerprint, so it cannot be compared as the original daily race.

No. The game state and mistake analysis stay in the tab. Save the replay JSON and mistake CSV explicitly if you want to keep them. The editor is excluded from capture, and the product uses only the site’s generic usage events rather than raw typing telemetry.