How GamePrep scores

The short version

When you finish a game, your raw performance is sent to our server and scored there against a fixed norm table for that specific game. The table maps raw performance to a percentile between 1 and 99. The same performance always produces the same percentile, for everyone, every time.

Scoring happens on the server, not in your browser

Your browser reports what happened — the sequences shown, what you entered, the timings — and nothing else. The score itself is computed in our backend. Nothing in the page can change your result, which is why a percentile here means the same thing from one session to the next.

Submissions are also validated for structural plausibility. A payload describing a result the game could not actually produce is rejected rather than scored. This is the same discipline real assessments use, and it is the reason we can compare your runs to anyone else's.

What the norm tables are

Each game has its own lookup table: thresholds of raw performance mapped to percentiles. For a speed task the thresholds are response times in milliseconds; for a memory task they are span lengths; for a reasoning task they are counts of correct answers under a timer.

These tables are fixed and static. They do not shift as more people play, they are not recalculated from our own user base, and they are not personalised. That is a deliberate trade-off:

• What you gain: stability. Your 60th percentile in January means the same as your 60th percentile in June, so improvement you see is real improvement rather than a moving comparison group.
• What you give up: live calibration. A percentile from a dynamic population would track the actual pool of players, at the cost of your past scores quietly changing meaning underneath you.

The thresholds are set to the ranges commonly reported for these classic cognitive tasks — digit span, simple reaction time, Stroop interference, flanker interference, numerosity judgement and so on. They are our reference points, published here so you can see what they are rather than being asked to trust a number.

What our percentiles are not

This is the part that matters most, and the part most often misread.

• Not an employer's benchmark. No assessment vendor publishes its cut scores, and no employer publishes the bar it screens at. Our percentile cannot tell you whether you would pass a specific company's assessment, because that threshold is not public information — see what counts as a good game assessment score.
• Not vendor data. We have no access to HireVue, pymetrics/Harver, Arctic Shores, cut-e/Aon or any other vendor's scoring, norms or item banks, and no relationship with them.
• Not a clinical or diagnostic measure. These are practice games, not psychometric instruments. They are useful for familiarity and for tracking your own improvement. They are not an assessment of your cognitive ability, and nothing here should be read that way.
• Not a ranking against other GamePrep users. Because the tables are fixed, your percentile is a position on a reference scale, not a live leaderboard position.

What the number is genuinely good for

• Finding your weakest area. Comparing your own percentiles across games is the reliable use of the scale — it is the same yardstick applied to each, which is exactly what the readiness check uses to tell you what to practise next.
• Tracking improvement. A fixed scale means a rising percentile reflects you getting better, not the comparison group getting worse.
• Removing unfamiliarity. The score is secondary to this. The largest measurable gain available before a real assessment is simply having played the format before.

Which games respond to practice, and which do not

We draw this line explicitly because a lot of preparation advice does not.

• Cognitive tasks improve. Memory span, processing speed, attention control and numerical fluency all respond measurably to short, repeated practice — partly through genuine adaptation, largely through strategy (chunking, endings-first scanning, reading the cue before the item).
• Behavioural and trait tasks do not. Risk-appetite and response-style games read traits rather than skills. There is no high score to reach, and trying to perform an idealised profile tends to produce inconsistency that reads worse than an honest run.

Corrections

If you think a norm table or a claim on this site is wrong, tell us and we will fix it. Contact us — corrections are welcome, and being right matters more to us than looking authoritative.