Pymetrics Games: Guide to All 12 Games and How to Prepare

Pymetrics games are 12 short behavioural games that measure over 90 traits. Learn what each game tests, how scoring works, and how to prepare properly.

Pymetrics games are a set of 12 short online games that employers use early in graduate and experienced-hire applications. Each game lasts around two minutes. Together they take roughly 25 to 35 minutes. The games look playful, but they are behavioural measurement tools: they record how fast you react, how much risk you take, how you respond to reward, and how you read faces. The Oxford University Careers Service describes the suite as measuring “over 90 cognitive, social, and behavioural traits”. There is no pass mark in the usual sense. The employer first has current high performers play the same games, builds an average profile from their results, then compares your profile to it. That means preparation is not about learning answers. It is about knowing the rules cold, playing consistently, and not sabotaging yourself with nerves or a bad setup.

TL;DR

The suite is 12 games, about two minutes each, roughly 25 to 35 minutes total, played in one sitting with no pause.
The games measure traits such as risk tolerance, attention, memory, planning, effort, generosity, fairness and emotional recognition.
Employers benchmark you against a profile built from their own high performers, so the “right” profile differs by firm and by role.
You cannot revise for the content, but you can learn every game’s rules in advance so you never waste the first 20 seconds working out the controls.
Consistency matters more than raw speed. Erratic behaviour inside one game reads as noise and weakens your profile.
Set up properly: quiet room, stable connection, laptop rather than phone, and a time of day when you are alert.

What are the pymetrics games and who uses them?

Pymetrics is a game-based assessment now part of the Harver talent platform. Employers use it as a first-round screen, usually straight after the online application form and before any video interview. Banks, law firms, technology companies and large graduate schemes all use it, and the same 12 games appear regardless of the employer. What changes is the benchmark profile behind the scenes.

The games sit in the same family as other task-based assessments, such as the Arctic Shores engine behind some Big Four online assessments and the scenario tests described in our Cappfinity assessment guide. All of them try to score behaviour rather than knowledge. If your process also includes a traditional timed maths test, warm up separately with five free timed numerical questions so you are not switching between two very different skills cold.

What does each of the 12 games measure?

The Oxford University Careers Service lists all 12 games and the trait each one targets. The table below maps them.

Game	What you do	Main traits measured
Balloon	Pump a balloon for reward, cash out before it pops	Risk taking, decision making
Keypress	Press the spacebar only on a given signal	Impulsivity, following instructions
Money Exchange 1	Send money to a partner who may return more	Trust, risk taking
Money Exchange 2	Share or take a further sum, then rate fairness	Altruism, generosity, fairness
Hard or Easy Task	Choose a low-paying easy task or a high-paying hard one	Effort, motivation, decision making
Digits Memory	Recall a growing sequence of digits	Working memory, attention
Stop	Respond only to one target colour among flashing circles	Attention, focus
Arrows	Respond to the middle or outer arrows depending on colour	Learning, adaptivity
Lengths	Compare two very similar images	Attention to detail, effort
Cards	Draw from four decks with hidden win and loss patterns	Learning, decision making
Tower	Rebuild a disc tower in the fewest moves	Planning, decision making
Faces	Identify the emotion in a face, sometimes with a short scenario	Emotional recognition

Two details are easy to miss. In the Tower game you are also scored on how long you take before your first move, so pausing to plan is measured, not penalised by default. In the Cards game the decks follow a hidden pattern, so the game rewards noticing which decks are quietly costing you money.

How is the pymetrics assessment scored?

There is no single score out of 100. The employer asks people already doing the job well to play the games first. Their results form an average profile for that role. Your 12-game profile is then compared to it. This is why the same profile can match one employer and miss at another, and why “play it safe” is bad advice: if the role rewards calculated risk, a very cautious balloon profile moves you away from the benchmark rather than towards it.

It also means faking is a poor strategy. The system records thousands of micro-decisions across 12 games. A profile assembled by guessing what a recruiter wants tends to be internally inconsistent, and inconsistency is visible.

Sample question 1: the balloon game and expected value

You cannot practise the exact balloons, but you can practise the thinking. Here is an original worked example in the same shape.

A balloon pops at a randomly chosen pump number between 1 and 16, with every number equally likely. Each pump adds 5 points. If you cash out after n pumps you keep 5n points. If it pops first, you keep nothing from that balloon. What is the best number of pumps?

Worked solution. The chance the balloon survives n pumps is the chance the pop point is above n, which is (16 minus n) divided by 16.

Expected points = 5n times (16 minus n) divided by 16.

Cash out at 4 pumps: 5 times 4 times 12 divided by 16 = 15 points.
Cash out at 8 pumps: 5 times 8 times 8 divided by 16 = 20 points.
Cash out at 12 pumps: 5 times 12 times 4 divided by 16 = 15 points.

The middle of the range wins. Very cautious and very greedy strategies both underperform. In the real game the pop range differs by balloon colour, so the practical lesson is to treat the first few balloons of each colour as information gathering, then settle into a steady cash-out point for that colour.

Sample question 2: the cards game and hidden loss patterns

Four decks pay out on every draw, but some decks also carry occasional heavy penalties.

Deck A: pays 100 points per card. One card in five costs 600 points.
Deck B: pays 50 points per card. One card in ten costs 250 points.

Which deck should you keep drawing from?

Worked solution. Compare the average value of a single draw.

Deck A: 100 minus (600 divided by 5) = 100 minus 120 = minus 20 points per draw.

Deck B: 50 minus (250 divided by 10) = 50 minus 25 = plus 25 points per draw.

Deck A pays more per card and still loses money. Deck B is the correct deck. The game is built so the flashy deck is the losing one. The behaviour being measured is whether you update after a run of losses or keep chasing the larger headline reward.

How should you prepare for the pymetrics games?

Preparation is logistics plus rule familiarity, not revision.

Read the rules of all 12 games before you start, so no game surprises you. Every game has different controls.
Play on a laptop with a real keyboard where possible. The Keypress, Stop and Arrows games are reaction-time games and a laggy trackpad or phone touchscreen costs you milliseconds on every trial.
Close everything else. Notifications during a two-minute attention game are expensive.
Pick a time of day when you are sharp. The suite runs straight through without pause.
Decide your general approach before you start rather than mid-game. Erratic switching inside a single game reads as noise.
Do not try to fake a personality. Aim for your genuine behaviour, played competently.

If your application also includes scenario questions, the judgement skills overlap with our situational judgement test guide, and the general logistics advice in how to pass online assessment tests applies here too.

What mistakes cost candidates the most?

The three expensive errors are all avoidable. First, learning the controls during the scored game instead of before it: in a two-minute game, 20 lost seconds is a sixth of the data. Second, treating every game as a speed test. The Tower and Cards games reward thinking; the Stop and Arrows games reward speed with accuracy. Third, quitting a game mentally after a bad start. Each game contributes to a profile, and a strong finish still counts.

If your process pairs the games with a timed maths section, the drills in our numerical reasoning test tips are the better use of your remaining preparation time, because that part of the process genuinely rewards practice.

Frequently Asked Questions

Can you fail the pymetrics games?

Not in the sense of getting answers wrong. You can be screened out because your behavioural profile is far from the employer’s benchmark for that role. The same profile may progress at a different firm.

How long do the pymetrics games take?

Around 25 to 35 minutes in total, with each of the 12 games taking roughly two minutes. The games run in sequence and you cannot pause partway.

Can you practise the pymetrics games?

You can practise the mechanics and the underlying thinking, such as expected value in the balloon and cards games, and reaction accuracy in the keypress games. You cannot revise content, because there is no syllabus.

Should you take risks in the balloon game?

Take measured risks. Cashing out immediately every time produces a very low risk score, which is only a match if the employer’s benchmark is very cautious. Popping most balloons is equally unhelpful. A steady mid-range strategy is the safer default.

Do the pymetrics games work on a phone?

They are playable on a mobile device, but a laptop with a keyboard is better for the reaction-time games. Use the setup that gives you the fewest input delays.

Does an employer see your individual game results?

Employers see a profile and a match against their benchmark rather than a leaderboard of raw scores. The individual traits are combined into that profile.