How to keep a personal pixel practice log.
A total score tells you how far a set missed. A short manual log can help you decide what to change next. Note each axis's direction during every round's feedback, before continuing. After five rounds, add the total and average axis errors from the final summary. There is no account, public ranking, or population benchmark: the useful comparison is you, on the same display, over several short sessions.
Absolute error and signed bias answer different questions
Pixactly's round score uses absolute error. If a target is 180px wide, both 170px and 190px are 10px away. Absolute error answers, “How far did I miss?” It is the right number for the score because either direction is still a miss.
In a manual log, you can give the difference a direction using
drawn size − target size. A positive width value means the
rectangle was too wide; a negative value means it was too narrow. For
height, positive means too tall and negative means too short. The
overlay shows which rectangle extends farther, while the axis error
gives you the size of the miss. The final summary shows absolute
averages, not the round outlines or signed differences, so record
direction while each overlay is still visible.
A worked five-round example
The values below are illustrative, not results from another player. They show why direction matters even when the total score looks clear.
| Round | Target | Drawn | Width bias | Height bias |
|---|---|---|---|---|
| 1 | 40 × 25 | 45 × 23 | +5px | −2px |
| 2 | 80 × 65 | 78 × 72 | −2px | +7px |
| 3 | 135 × 110 | 143 × 105 | +8px | −5px |
| 4 | 210 × 165 | 218 × 160 | +8px | −5px |
| 5 | 300 × 260 | 309 × 252 | +9px | −8px |
| Average signed bias | +5.6px | −2.6px | ||
In this example, width usually overshoots by about 6px while height is about 3px short. The next useful instruction is not “try harder” or “make everything smaller.” It is narrower and more testable: reduce horizontal distance slightly while leaving the vertical guess alone.
Use a three-set calibration protocol
- Set one: establish a baseline. Draw honestly. During each round's feedback, note signed differences before continuing. At the end, add the average absolute errors from the summary and calculate signed averages from your own notes.
- Set two: change one axis. Inspect the larger-error axis. If its signed notes mostly point in one direction, test a small correction in the opposite direction. If they alternate, test a consistent endpoint instead of a blanket size change. Keep the other axis's approach steady.
- Set three: check whether the bias moved. A lower total is welcome, but the more useful question is whether the signed bias moved toward zero and absolute error also fell, without the other axis becoming worse. Opposite large misses can cancel in a signed average. Targets change between sets, so repeat the observation before treating it as a reliable improvement.
Compare on the same display
A CSS pixel does not have one fixed physical size. Pixel density, browser zoom, available canvas space, and viewing distance all change how a target feels. Compare nearby sets on the same device and at the same browser zoom when you want the history to mean something.
Switching from a phone to a large monitor is still interesting, but it is a new context rather than a clean continuation of the same test. The log should treat those observations as a new context rather than a universal rating of visual ability.
What to write down
A useful manual entry can stay very small: date, device, total score, larger average axis error, visible direction, and the single change you plan to test next. Keep it in your own notes or spreadsheet if you want to compare sessions.
Pixactly keeps your personal best in local storage and temporarily stores the latest final summary in session storage for restoration. It does not save your signed practice notes or create an account or public leaderboard. The privacy page describes analytics and consent controls separately.