Data & Methodology
Several of our articles publish benchmarks from real players, such as the median reaction time or the share of players who clear 3-back. This page explains where those numbers come from, how we clean them, and where their limits are.
What we record
When you finish a ranked game, your browser sends the result to our leaderboard. Each entry contains:
- a random ID created in your browser, not linked to any account
- an automatically generated nickname, such as Swift_Owl_1234
- the game, the mode, your score, and the time it was submitted
- for some games, a few details of the round, such as accuracy or level
We do not ask for an account, email, or real name, and the published benchmarks only contain totals and percentiles, never individual entries.
One result per player
- The leaderboard keeps each player's best result, so every benchmark counts each player once.
- A "player" is one browser. The same person on two devices counts twice, and clearing browser data creates a new ID.
- Because we keep the best result, benchmarks describe how well people can do on a good attempt, not their typical attempt.
How we filter results
Before publishing, we remove results that are almost certainly not genuine or not complete. The server already rejects impossible scores, such as a reaction time average under 100 ms. Each benchmark then uses its own limits:
| Test | Score | Excluded | Article |
|---|---|---|---|
| Reaction time | Average of 5 rounds (ms) | Under 120 ms or over 1,000 ms | Read |
| Schulte table 5×5 | Time plus 2 s per wrong tap | Under 8 s or over 120 s | Read |
| CPS test (5 s) | Clicks per second | Under 1 or over 16 CPS | Read |
| 10 Second Challenge | Distance from 10.000 s | 3 s or more away | Read |
| Dual n-back | Fastest clear and its level | 10-back and early 1-back records | Read |
| Digit span (forward) | Longest correct sequence | Backward version, too few players | Read |
| Trail making | Part A and Part B time | Part A over 180 s or Part B over 300 s | Read |
What these numbers can and cannot tell you
- Players choose to take the tests, so they are not a random sample of the population.
- Phones, mice, screens and browsers all add their own delay, which matters most for reaction time.
- Tests with only a few hundred players, such as digit span and trail making, are early data and may shift as more people play.
- Our online versions differ from clinical tests, so the results are not norms and cannot diagnose anything.
Updates
Each article states the date range of its data. When we refresh the numbers, we rerun the same queries and update the date range and the "Updated" date on the article.
Citing our data
You are welcome to quote these benchmarks. Please link to the article the numbers come from, and mention the date range. For questions or aggregated data, email contact@freefocusgames.com.
Example: FreeFocusGames (2026). What Is a Good Reaction Time? Benchmarks From 8,791 Players.
Open source
The games and the leaderboard that collects these results are open source. You can read the code on GitHub.