Standing reference, updated with any material changeData Methodology
Every statistic we publish comes from this pipeline. This page is the single reference for how the numbers are made, what they can support, and what they cannot.
01What is collected
A score enters the dataset only when a player explicitly chooses "Post anonymous score" after a run. The stored record contains the game, mode, duration, score, broad input class (touch, mouse, keyboard, mixed, or unknown, as reported by the browser), the run's generated seed, the scoring rules version, and a submission timestamp. No name, email, age, location, or free-text field exists in the dataset.
02What the numbers contain
Scores are browser measurements. They include display refresh, browser scheduling, and input-device delay on top of the human response, so they run slower than laboratory measurements and differ between devices. That is why every published statistic is split by input class and never pooled across devices.
03Exclusions and validation
Submissions pass a server-issued run token, elapsed-time checks, per-game score ceilings, duplicate prevention, and rate limits. Scores that fail validation are rejected or marked and excluded from statistics. Accepted scores remain labeled provisional because the browser still computes the result.
04Repeat players
A random install identifier, stored only as a one-way hash, limits abuse but does not identify people. The same person can appear multiple times, from practice runs or from different browsers. Published statistics therefore describe posted runs, not unique humans, and we say so wherever the numbers appear.
05Known biases
This data comes from visitors of a browser games site who chose to play and then chose to post. That skews the sample toward practiced, competitive, younger-leaning players and toward whichever regions and devices our traffic comes from. None of our statistics should be read as describing the general population.
06Reporting thresholds
Medians appear once a group has at least 30 posted scores. Percentile tables and formal reports wait until the group passes 300, and we do not publish percentiles beyond the 90th on samples of this size. Averages are avoided because the distributions are skewed.
07Not a cognitive or medical measure
These are game scores. They do not measure intelligence, brain health, or clinical reaction capacity, and we never present them that way. Anyone citing this data should avoid those framings too.
08Citing this data
Aggregated statistics and charts from this site may be reused with attribution to "Play by Offline Observer" and a link to the page you cite. Aggregate downloads we publish are provided under a CC BY 4.0 license. We never publish individual records.
See the live numbers
The current sample sizes and medians are always visible.