Game-Based Assessments 2026: Every Game

A game-based assessment replaces question-and-answer tests with short interactive tasks: pumping balloons, sorting shapes, predicting outcomes, reading faces. The software records how you play, often thousands of data points per task, and turns that into estimates of reasoning, attention, learning and work style. Pymetrics (now Harver), Arctic Shores, Aon, HireVue, Cappfinity and McKinsey Solve all use the format. A typical battery takes 20–60 minutes. This guide compares the main vendors, explains what is and is not "scored", and shows how to prepare.

Key facts

What "game-based" actually means

Researchers separate three things that marketing pages lump together (Landers & Sanchez, 2022):

• Game-based assessment: the game is the measurement. You play a loop, the system infers traits from your behaviour. Pymetrics and Arctic Shores are this.
• Gamified assessment: a classic test with game elements added (points, levels, animation). Aon's smartPredict challenges sit close to this: the content is numeracy, logic and memory, the wrapper is a game.
• Gamefully designed assessment: a test designed around game ideas such as a scenario world. Cappfinity's immersive assessments and McKinsey Solve's ecosystem simulations lean this way.

The distinction matters for preparation. In a gamified ability test, there is a correct answer and practice raises your score. In a behavioural game, there is often no correct play, and "practising" means learning the controls so that your real tendencies, not confusion, are what gets recorded.

How the main vendors compare

Game lists change, and every vendor lets employers pick a subset. Use the table to recognise what you are looking at, then check the vendor guides linked below.

What a game-based assessment looks like: a first-hand example

In an autumn 2026 application round, a German-language Arctic Shores battery had three games in a row, about 32 minutes in total:

1. Order ("Rätsel"): one practice puzzle, then 17 scored puzzles. Three frames are shown, and you build the fourth from four shapes, five colour patterns, a 45° rotate button, a resize button and a 3×3 grid of anchor points. Puzzles grew from one rule to three or four interacting rules. About 15 minutes.
2. Predict ("Nacht oder Tag"): about 90 rounds. Each round shows a card with two symbol patterns and you pick Night or Day, learning from outcomes which symbols lean which way. About 6–7 seconds per round, 10–11 minutes.
3. Expressions ("Gesichtsausdruck"): 7 practice faces then 60 scored faces, each labelled with one of seven emotions. About 4 seconds per face.

Three things from that sitting generalise to most game-based tests:

• Time is recorded but not displayed. A stopwatch icon filled up during each item without ever showing a number.
• No feedback on reasoning items. You never learn whether a puzzle was right.
• The "Next" button is locked until you have answered, so you cannot skip and come back.

How game-based assessments are scored

Against a reference group. Arctic Shores compares your behaviours with a representative norm group and lets the employer choose which traits matter. Pymetrics compares your trait profile with high performers already in the role and applies a role-specific threshold.

From process, not just outcome. Arctic Shores says its reasoning tasks credit partial work. McKinsey Solve is reported to score both the final answer and how you got there.

With no "best" profile on behavioural games. A high-risk balloon player is not better than a cautious one; each fits some roles. What hurts is noise: random play, misunderstanding the rules, distraction.

Practically: on cognitive games (puzzles, memory, maths, emotion labels), be accurate and steady. On behavioural games (risk, trust, effort, strategy), understand the mechanics and then play the way you actually decide.

Original practice examples

All examples are original and mimic common mechanics, not any vendor's items.

1. Balloon (risk and learning)

Each pump earns 2p. A balloon of this colour bursts at a random pump between 1 and 40, every pump equally likely. If it bursts you lose the money on that balloon. How many pumps maximise the expected payout, and what is it?

Answer: 20 pumps, 20p expected. - If you plan k pumps, the chance the balloon survives is (40 − k) / 40, and the payout is 2k. - Expected value = 2k × (40 − k) / 40, which peaks at k = 20: 40 × 20 / 40 = 20p. - Lesson: very low pumping and very high pumping both leave money on the table. In real games you do not know the range in advance; you learn it from the first few bursts of each colour. Your risk style is what gets profiled, so this is about understanding the trade-off, not hitting a number.

2. Cards (learning from losses)

Deck X pays +80 per draw, but every block of 10 draws contains penalties totalling 1,000. Deck Y pays +40 per draw with penalties of 200 per 10 draws. Which deck is better over time?

Answer: Deck Y. - Deck X: 10 × 80 − 1,000 = −200 per 10 draws. - Deck Y: 10 × 40 − 200 = +200 per 10 draws. - The big-reward deck feels better early because penalties arrive in bursts. These tasks test whether you notice the long-run pattern (they derive from the Iowa Gambling Task).

3. Arrows (attention and switching)

Rule: if the arrows are blue, respond to the centre arrow; if orange, respond to the outer arrows.

Answer: right, right, left, left. - Blue `< < > < <`: centre is `>` → right. - Orange `> > < > >`: outer arrows `>` → right. - Orange `< < < < <`: all left → left. - Blue `> > < > >`: centre is `<` → left. - The hard trials are the incongruent ones (centre and flanks disagree), especially right after a colour switch.

4. Order (build the next frame)

A small purple pentagon sits in the bottom-left and grows each frame (small, medium, large). An orange cross appears in the centre in frames 1 and 3 only. A blue drop moves along the border: frame 1 middle-left, frame 2 top-left, frame 3 top-centre.

Answer: purple pentagon in the bottom-left, one size above large; no cross; blue drop in the top-right. - Pentagon: growth only. - Cross: present in odd frames, absent in even frames, so frame 4 has none. - Drop: one border step clockwise per frame (middle-left → top-left → top-centre → top-right). - Build the elements with clear rules first; leaving out the cross is part of the answer, not a gap.

Game-based vs classic psychometric tests

Traps and how to avoid them

• Inventing rules. In prediction and card games the rule is learned from outcomes, not from how many symbols are on the screen.
• Over-reacting to noise. Probabilistic feedback means you will be "wrong" sometimes even when you choose well.
• Playing a persona. Trying to look bold or cautious breaks consistency across games that measure the same trait in different ways.
• Rushing the instruction screens. Many games explain the one control you need (a grid, a key, a colour rule) on the last intro screen.
• Bad hardware. A trackpad, a lagging browser or a phone makes speed-sensitive games harder than they need to be.

How to prepare for a game-based assessment

1. Identify the vendor. The invitation URL or e-mail usually names it (pymetrics, una-arcticshores, Aon, HireVue).
2. Learn the mechanics, not "answers". Read the vendor guide, do the original examples above and the vendor's own practice task.
3. Practise the cognitive skills. 20 minutes a day of inductive puzzles, digit span and simple weight logic for two or three days is plenty.
4. Set up properly. Laptop, mouse, current browser, stable connection, notifications off, 15 minutes of buffer.
5. Rest. Reaction time and attention drop with fatigue faster than reasoning does.
6. Be consistent. Behavioural games cross-check each other. Play as yourself.

Sources

• Landers, R. N., & Sanchez, D. R. (2022). Game-based, gamified, and gamefully designed assessments for employee selection: Definitions, distinctions, design, and validation. International Journal of Selection and Assessment, 30(1), 1–13. https://doi.org/10.1111/ijsa.12376
• Arctic Shores, candidate guidance and science page https://arcticshores.com/science
• Arctic Shores, "ChatGPT vs task-based assessments" (partial credit)
• Game Assessment Prep, Arctic Shores games catalogue
• Psychometric Success, pymetrics games (12 games, 25–35 min, comparison with high performers)
• MyConsultingCoach, pymetrics guide (Harver acquisition)
• JobTestPrep, Aon smartPredict challenges
• GraduatesFirst, HireVue games (Shapedance, Digitspan, Flashback)
• Hunt Scanlon, HireVue acquires MindX (2018)
• Intervyo, Cappfinity immersive assessments
• RoadToOffer and CaseStar, McKinsey Solve 2026 format https://www.casestar.io/guides/mckinsey-solve/what-changed-2026
• Bechara, A., Damasio, A. R., Damasio, H., & Anderson, S. W. (1994). Insensitivity to future consequences following damage to human prefrontal cortex. Cognition, 50, 7–15 (Iowa Gambling Task).
• Lejuez, C. W., et al. (2002). Evaluation of a behavioral measure of risk taking: The Balloon Analogue Risk Task (BART). Journal of Experimental Psychology: Applied, 8, 75–84.
• Eriksen, B. A., & Eriksen, C. W. (1974). Effects of noise letters upon the identification of a target letter in a nonsearch task. Perception & Psychophysics, 16, 143–149 (flanker task).

Written by

Adrian, Specialist

MSc Psychology, 7 years of test taking.

Last updated: Oct 1, 2026