Skip to main content

Typical scores on our Pymetrics practice games

How people actually score on CogniPrep's practice versions of the Pymetrics games, game by game. Everything here is counted from our own practice sessions, anonymised and added up. None of it comes from Pymetrics or Harver, and no employer sees it.

Snapshot taken 8 November 2026, from sessions played between 22 July 2026 and 8 November 2026.

606
completed practice sessions
314
players
1 of 12
games with enough data to show a spread
These are not Pymetrics results
Pymetrics does not give you a score, and says its games have no right or wrong answers. A 2021 audit of its system explains why: each employer's model is trained on how a group of its own high performers played, compared with a wider group, and candidates are placed in recommendation tiers against that model. What counts as a good pattern changes from one employer to the next, so nobody outside Pymetrics can turn a practice score into a real result, us included.

What the practice score is

Our Pymetrics practice games that measure an ability end with a score out of 100. It is our own measure of how you did on that run. On Keypress it is built from how many times you pressed, how fast and how evenly you kept pressing, and how much you slowed down in the second half. The real game shows you nothing, so read the score as feedback on your own rhythm rather than a target. Every game weighs things differently, so compare a score only with other scores on the same game.

After a game we also show a percentile: the share of saved practice sessions on that game that your score beat. The chart below shows the same kind of spread, counted more strictly (see how we counted, further down), so it will not match it exactly.

How scores spread on Keypress

Each bar is a band of 5 points and shows the share of sessions that scored in it. To place a score of your own, open the table under the chart: it gives the score at each tenth of the spread.

Keypress

Median score
85
Median first attempt
84

Tap or hover over a bar to read it

8 in 10 sessions scored between 66.9 and 96. Based on 390 sessions from 271 players since 29 August 2026.

Score at each percentile
Keypress: the score below which each share of sessions fell
PercentileScore
10th66.9
20th77
30th80
40th83
50th (median)85
60th87
70th90
80th93
90th96
Practise Keypress

Games without a spread

Balloon, Money Exchange I, Money Exchange II and Hard or Easy Task have no better or worse result. They record traits such as how you weigh risk, how you share money and how much effort you choose to put in, so our versions give a profile instead of a score, and we do not publish a spread that could be read as a target. Between them they have 17 counted sessions so far.

The games below have fewer than 200 sessions or 100 players so far. A spread from so few would mostly show who happened to play, so we leave it out until they have more.

Tower
7 sessions from 4 players
Digits Memory
36 sessions from 28 players
Stop
9 sessions from 2 players
Arrows
26 sessions from 2 players
Lengths
5 sessions from 3 players
Cards
12 sessions from 3 players
Faces
104 sessions from 68 players

What changes with practice

Not enough to say yet. To show how scores move from a first to a third attempt, a game needs 50 players whose first three sessions were all complete and inside the counting window, and no Pymetrics game here has that many so far.

Try the games behind these numbers

Keypress, Digits Memory and Faces are free to play with a free account.

Play free

How we counted

What is counted
Completed sessions of our Pymetrics practice games. Staff and test accounts are left out, and so are sessions our scorer flagged as incomplete, where the game could not be scored from what was played. Where we changed a game's scoring or timing, only sessions since the change count, as the rows below say; the other games count every session since launch.
Keypress
We changed how the game is scored on 28 August 2026, so earlier sessions are left out.
Repeat players
Nobody counts more than 10 times in a spread: we keep each player's first 10 sessions, so a few people replaying a game many times cannot set its shape. The median first attempt uses every player's first session.
Minimum sample
A spread needs 200 sessions and 100 players, enough to place the median to within about 7 percentile points either way. First to third attempt needs 50 players.
Privacy
The snapshot holds only totals, medians and shares. No name, email, account or single session is in it or on this page.

Sources

Pymetrics facts checked against these sources.Practice figures are CogniPrep's own, from the snapshot dated 8 November 2026.

Read next