Top 3% of what, exactly?
When the result screen says “top 12%,” you naturally read it as: “out of 100 people who took this here, I am around 12th.”
It does not mean that. This guide is about where that number actually comes from — described exactly as our scoring code works.
There are four kinds of number on screen
Mixing them up keeps people confused. Let us separate them.
1. The raw value. What was actually measured: 273ms of reaction time, 9 digits, 412 keystrokes per minute. This is the measurement itself.
2. The value with penalties folded in. Some tests do not use the raw value. Hard Brake takes your average time on green and adds penalties in proportion to the reds you pressed and the greens you missed. So Hard Brake’s “286ms” is not a reaction time. Never line it up next to 286ms from the reaction test.
3. The estimated top % and the points. This is the heart of this guide — covered below.
4. Your position in the Hall of Fame ranking. That one is a real rank: actual submitted records, sorted. It is a completely different kind of number from the first three.
How the top % is produced
Our code holds a table of anchors for each test — a list of [measured value, top %] pairs. With made-up numbers, for illustration:
| Measured value | Top % |
|---|---|
| 200ms | 20% |
| 240ms | 40% |
If your value is 220ms, it is not in the table, so we draw a straight line between the two points and read off the result. Exactly halfway, so 30%. That is the whole calculation.
(The real tables are much denser, and their values live only in each test’s meta.ts. Copying them here would leave this article stale the moment a table changes.)
Here is the part you must know. This calculation never looks at anyone else’s records. It does not query a distribution of visitors, and it does not update in real time. It just finds your spot in a table written in advance.
Points are then a rescaling of that top %: top 0.1% = 1000 points, 99.9% = 0 points, in a straight line. So points carry the same information as the top %, on a different dial. There is nothing new in them.
Where those anchors came from
It differs by test, and that matters.
- For things like reaction time, digit span and typing, where public statistics or widely known distributions exist, the tables were built with reference to those.
- For games where we made up the rules ourselves — Hard Brake, Color Trap, working memory — the tables are initial estimates calculated from those rules. The stimuli, trial counts and conditions differ from academic instruments, so we did not import their numbers.
That is why the note under your result says the anchors are initial values. Once enough records accumulate we intend to replace them with measured ones, and after that the same raw value will map to a different top %. (Attempts already saved are never re-scored.)
Put plainly: the top % means “assuming this distribution, roughly around here.” It is not a percentile of a real population.
So when is this number useful?
For two purposes it is genuinely useful.
First, comparing you with you. With the same anchors, same device and same conditions, yesterday’s you and today’s you compare exactly. However accurate the table is or is not, you used the same ruler both times.
Second, a rough sense of position. Top 5% and top 60% are clearly different stories. Small gaps like top 12% versus 15%, on the other hand, deserve caution: we have not computed how accurate the tables are (they are initial estimates, so there is nothing yet to compute it from), and your day-to-day condition and hardware error are folded in as well.
And the useless purposes, stated plainly: this number is not an exact rank against other people, and not a grade of ability. If you want that, look at the Hall of Fame ranking — real submitted records, actually sorted.
Tiers (the animals) are another layer again
Your animal tier comes from your series score: your personal best points across the tests in that series, added up, then taken as a share of the maximum (number of tests × 1000).
So doing brilliantly in one test barely moves your tier. Tiers only make sense once you have done the whole series. Conversely, any test you have not taken counts as zero, which drags your tier below where it belongs.
In short
- The top % is a linear interpolation from a table written in advance. It does not look at other people’s records, and it has no error bars yet.
- Points are the same information on a different dial. No new information.
- Values from tests like Hard Brake have penalties folded in. Do not compare them to other tests’ milliseconds.
- If you want a real rank, use the Hall of Fame ranking.
- Tiers only work properly once you finish the whole series.