Games and game-style tests
Game-based assessment practice
Short games that measure traits, attention and decision style. Know the format; do not chase a score.
Choose an exercise
About this test
Game-based assessments replace or sit alongside questionnaires. You play short tasks, such as pumping a balloon, sharing money with a partner, pressing a key on a signal or recalling a sequence, and the platform turns your behaviour into traits like risk preference, attention or planning. Three publishers account for most of what UK finance candidates report: a per-employer matching platform, a task-based personality and ability publisher, and an ability-game suite from a large assessment group.
We build practice for each from public sources: vendor science pages, published audits, academic task protocols and candidate guidance. We cannot copy any employer's scoring, because none is public, and several publishers say plainly that practice should not change your result. So the report for each game explains what was measured in plain English and never calls a result better or worse.
What we know about this format
Each point carries a badge showing where it comes from. Where only candidates have reported something, the badge says so.
What do the confidence badges mean?
- Vendor-published Stated in the test publisher's own document or practice site.
- Academic From peer-reviewed research.
- Official employer Stated on the employer's own page.
- Near-official Near-official: an employer-run practice portal, or a vendor or employer-quoted press release naming the employer.
- Candidate-reported Reported by candidates. Treat as likely but unconfirmed.
- Third-party From prep sites, press or forums. Not confirmed by the employer or publisher.
- Vendor-published
Per-employer matching platform: it builds a separate model for each employer from that employer's own high performers. A 2021 audit co-written by the publisher's staff describes this.
Source · Peer-reviewed conference paper, co-authored by publisher staff.
- Vendor-published
The same publisher's candidate FAQ says only your first results are used and that scores should not change much.
Source · 2018 FAQ reproduced in an employer guide.
- Vendor-published
Task-based personality and ability publisher: each trait is scored on a 1 to 10 sten scale and compared with a target range the employer sets, with 'danger zones' at the extremes.
- Vendor-published
Ability-game suite from a large assessment group: four games of roughly 6 to 15 minutes each, with worked examples first and tests generated per participant.
Source · 2018 press release.
- Academic
Retest reliability of the balloon risk task is about r = .77 across days with no average change, so practice barely moves it.
Source · White, Lejuez and de Wit 2008.
- Academic
The traditional gambling-task score has a retest reliability of only about r = .37, and players learn the deck rules, so layouts must be randomised.
Source · Sullivan-Toole et al.
Aon-style set: all four games
- Vendor-published
Aon-style game tests are timed and level-based: timed levels must be solved to unlock the next, worked examples come first, and each test is generated for the individual candidate
Source · Press release from the provider, 2018. The provider's own product pages give no game detail.
- Vendor-published
A set of four short ability games, each 6 to 15 minutes including instructions
Source · Press release, 2018.
Missing digits (Aon-style)
- Candidate-reported
The result is shown and you fill in the missing numbers or operators; about 5 minutes; scored on correct answers per minute
Source · Candidate-reported.
- Third-party
Related numeracy format: fill the gaps in an equation with digits 1 to 9
Source · Prep site description of a related test.
Grid memory (Aon-style)
- Candidate-reported
Memorise a dot position, do a symmetry, rotation or changed-grid check, repeat for 3 to 5 dots, then recall the sequence; 9 rounds, about 9 minutes
Source · Candidate-reported. The checks used here are our own versions of those three types.
- Academic
The standard symmetry-span design: items to remember alternate with symmetry judgements, then recall in order
Source · Kane et al. 2004, Journal of Experimental Psychology: General 133, 189-217.
- Academic
Partial-credit scoring (items placed correctly, even in a round that is not perfect) and accuracy on the in-between task
Source · Conway et al. 2005, Psychonomic Bulletin & Review 12, 769-786. Researchers usually set results aside below 85% accuracy on the checks.
Slide to exit (Aon-style)
- Candidate-reported
Rush Hour-style block sliding where a ball must reach the exit; a counter shows the minimum moves; 6 minutes; skipped items cannot be revisited
Source · Candidate-reported.
- Vendor-published
Boards are generated per candidate, so there is no fixed answer to memorise
Source · Press release, 2018.
Switch code (Aon-style)
- Candidate-reported
Four symbols pass through a funnel and you pick the 4-digit code that reorders them; a third row appears at harder levels; keep going for as many levels as you can
Source · Also described by GraduatesFirst and Intervyo. Candidate-reported, not confirmed by the provider.
- Candidate-reported
About 6 minutes (some versions about 3)
Source · Versions differ between employers.
- Third-party
Top performers at one bank reportedly reach level 40 or higher
Source · Anecdotal press report. Not used to set any target here.
- Third-party
Morgan Stanley is reported to source its tests from this provider (since about 2018); UBS, Nomura and BNP Paribas are reported by candidates
Source · Press and candidate reports only. Check your own invitation email for the real provider.
Balance (Arctic Shores-style)
- Vendor-published
Up to 9 tasks; the first one or two may be timed; breaks are allowed between tasks; a practice task comes first
Source · From the vendor’s candidate guidance. The vendor’s Help Centre was not readable, so task names and counts come from the sources below.
- Vendor-published
Each behaviour is shown on a 1 to 10 spectrum against a norm group, and there is no better or worse
- Vendor-published
Each trait gets a sten (1 to 10). The employer’s target profile gives a target range per trait plus danger zones, and a sten of 10 is not always the best
Source · Vendor sample report hosted by Newcastle University careers.
- Vendor-published
Fit combines how close each sten is to its target range, weighted by how important the trait is for the role, then becomes a percentile against a norm group
Source · Kvanta shows the same method with a role it made up and percentiles only against Kvanta players.
- Vendor-published
The vendor does not publish which task feeds which trait, or its scoring models and norm tables
Source · So every game-to-trait link in Kvanta’s report is Kvanta’s own inference, and the reference values are provisional.
- Vendor-published
Aptitude tasks give partial credit, and tasks use no language
- Vendor-published
Aptitude percentiles are reported in three bands: below the 30th, 30th to 50th, above the 50th
Source · Kvanta shows these bands only against Kvanta players, once 30 or more people have played.
- Academic
Retaking reasoning tests gains about a quarter of a standard deviation, mostly over the first few sittings
Source · Hausknecht et al. 2007. The report says practice raises reasoning scores.
- Candidate-reported
A scales-and-shapes puzzle appears in a pilot battery
Candidate report in Kvanta’s research brief. No public link.
- Candidate-reported
10 puzzles in 10 minutes with adaptive difficulty
Counts are from the research brief. The five-level staircase, the shapes and the puzzle generator are Kvanta’s own; each puzzle has exactly one balancing option, checked by brute force.
Balloon (Arctic Shores-style)
- Candidate-reported
Balloon: pump and bank; about 45 balloons
- Third-party
Balloon, Leaflets and Power are named in press coverage of the vendor’s older tasks
Source · Press coverage, quoting candidates.
- Academic
Balloon Analogue Risk Task: bursts uniformly between 1 and the ceiling; adjusted average pumps on balloons that did not burst
Source · Lejuez et al. 2002. Kvanta’s three balloon types have ceilings of 128, 64 and 32 pumps.
- Academic
Retest reliability of the balloon task is about .77 with no average change, so practice barely moves it
Source · White, Lejuez and de Wit 2008.
- Candidate-reported
Balloons of different types burst after different numbers of pumps
Prep sites say colours may differ in risk. Kvanta marks types by pattern and label, never by colour alone.
Arctic Shores-style battery
- Vendor-published
A battery has up to 9 tasks and takes about 15 to 25 minutes, with a practice task first and breaks between tasks
Source · Task count and the practice task are in the vendor’s guidance. Kvanta’s nine tasks and their order are its own pick from the reported games.
- Third-party
Results from the real assessment last about 9 to 12 months; the vendor advises against retaking within a year
Source · Press quoting the vendor. The sample reports state the 9 to 12 months.
- Vendor-published
Sten scores per trait against an example role: a role Kvanta made up, with target ranges, danger zones and importance weights
Source · Method from the vendor’s sample report. The role itself is not any employer’s.
Direction (Arctic Shores-style)
- Candidate-reported
Arrows / Direction: report the centre arrow and ignore the others; 80 to 85 trials, about 2 minutes, and it speeds up
- Vendor-published
A task is done under varying pressure
- Academic
Flanker task: congruent versus incongruent side arrows
Source · Eriksen and Eriksen 1974.
- Academic
Cued rule switching: switch cost in time and errors
Source · Rogers and Monsell 1995.
- Candidate-reported
85 rounds, 30 on the middle rule then mixed rules, a window shrinking from 2.0 to 1.1 seconds, 6 practice rounds
The trial count and speeding up are reported. The rule switches, window values and practice rounds are Kvanta’s own design.
Energy (Arctic Shores-style)
- Candidate-reported
Power Generator: press buttons; some generators add power and some drain it; learn the pattern
- Vendor-published
Learning Agility task: speed of learning, use of feedback and transfer to a new problem (mechanics not disclosed)
Source · Kvanta’s second round with new symbols is a stand-in for transfer.
- Academic
Multi-option reward learning: choose among options with different average payoffs
Source · The standard bandit paradigm (for example Daw et al. 2006).
- Academic
Learning games re-randomise which choices are good every session, because a fixed layout teaches the answer and changes what is measured
Source · Sullivan-Toole et al. on practice effects in the Iowa Gambling Task. Kvanta applies the same rule to Energy and Predict.
- Candidate-reported
Two rounds of 30 choices, three panels that add power and three that drain it
The structure is Kvanta’s own design. Which panels are good, and where they sit, is drawn at random every session.
Expression (Arctic Shores-style)
- Candidate-reported
Emotions: name the facial emotion; 50 rounds, about 3 minutes
- Academic
Forced-choice emotion recognition scored with the unbiased hit rate
Source · Wagner 1993.
- Candidate-reported
Faces are synthetic and the 7 emotions are anger, disgust, fear, happiness, sadness, surprise and neutral
Seven emotions is from the research brief. Faces are Kvanta’s own synthetic images.
Lock (Arctic Shores-style)
- Candidate-reported
Security door: stop a spinning dial on a target; up to 15 rounds; candidates report a final level they could not complete
- Candidate-reported
What is measured is what a person does after failing, not whether they succeed
This is Kvanta’s inference from the reported unwinnable level. How the vendor scores it is not public.
- Candidate-reported
15 doors, a speed that changes every door, a 4-minute limit, a locked door at door 8, 9 or 10
Up to 15 rounds is reported. The speeds, zone sizes, time limit and locked-door position are Kvanta’s own design.
Questions people ask
What is a game-based assessment?
It is a set of short interactive tasks, often drawn from psychology research, that record how you behave rather than what you say. The platform turns that behaviour into trait or ability measures.
Is there a best score?
For trait games, no. Employers compare your pattern with a profile set for the role, and the target ranges are private. Faking a pattern can backfire because the extremes are sometimes flagged too.
Can I practise for them?
You can learn the format, controls and pace, which removes the surprise. Publishers say practice should not change trait results much, and some measures such as reaction time barely move. Ability games, such as memory spans and logic levels, do show some practice effect.
Which finance employers use these?
Reports vary and few are confirmed by the employers themselves. The firm pages list each firm's reported vendor with a confidence badge, so you can see what is official and what is only reported.
Can I retake a game assessment?
Publisher rules differ and are set with the employer. One publisher's quoted guidance is not to retake within 12 months, and a bank's page was reported to apply a 330-day lock. Both are reported secondhand, so check your own invitation.
I am neurodivergent. Are there alternatives?
Ask the employer before you start. Alternative formats and extra time are standard reasonable adjustments. Our adjustments page has email templates, and each of our games has an opt-out note.
Are these the real games?
No. We build original tasks from public protocols and published descriptions. They do not use any publisher's artwork, names or items, and the scoring is academic, not an employer's model.
Where this appears in firm processes
- BDO UK Official employer
- Santander UK Official employer
- Schroders Official employer
See all firm pages. Check your own invitation, because processes change each cycle.
Independent practice. Not affiliated with or endorsed by Aon, Arctic Shores and Pymetrics.
Practice builds familiarity. Trait games have no ideal score: there is no better or worse result to chase, and employers compare you with a profile they set for the role. Reaction-time measures barely change with practice.
Use Kvanta to practise and learn. In a live assessment, do not use AI tools, notes or another person unless the employer says you may. Read our approach to integrity.