Accenture Assessment #1 — Behavioural + Cognitive
Practice software for the gamified round. Three playable games, a full 40-minute rehearsal, and the evidence trail behind every number in them.
Start here → games/00-battery.html for a full run, or
games/G1-bubble-arithmetic.html for the game that matters
most. Double-click either; there is no server and no install.
The 30-second version
| The round | 40 minutes. ~54 behavioural statements (~20 min), then three games (~20 min) |
| What eliminates | The games. 3–4 people at one candidate's test centre were cut after this round |
| The only number with evidence behind it | 20 of 25 on the bubble game — two independent candidates |
| The three games | Bubble arithmetic · arrow-tile pathfinding · keys & doors with invisible walls |
| Is there a memory game? | No separate one. The "memory game" a candidate reported is the keys-and-doors board — remembering which doors are locked is the memory test |
| Platform | Mercer Mettl, Secure Exam Browser 1.6.8. It counts navigation away from the window |
Play
| File | What it is | Exam-mode parameters |
|---|---|---|
games/00-battery.html |
The full 40 minutes, once, in order. Behavioural half, then all three games, then one report | 54 statements / 20 min, then the three below |
games/G1-bubble-arithmetic.html |
Three bubbles, click them smallest result to largest. Build your score here first | 25 questions × 15 s, decimals on |
games/G2-shortest-path.html |
One-way arrow tiles. Rotate, swap and place them to route the rocket to Earth | 8×8, 4:00 |
games/G3-key-collection.html |
Collect every key, reach the exit, through walls you cannot see | 10×10, 3 keys, 4:00 |
Every game has Exam mode (locked to the reported real parameters) and Practice mode (difficulty ramp, every rule exposed as a toggle). Each keeps your last 10 sessions and shows the trend. Each opens with a panel saying exactly which of its rules are reported and which are reconstructed.
Read
| File | What it is | When |
|---|---|---|
99-TRAINING-PLAN.md |
14-day spaced schedule, rising daily targets, and the stop rule | Before you start playing |
behavioural/PERSONALITY-STRATEGY.md |
The other half of the round. Two pages, deliberately | Once, the night before |
The evidence trail
Read these if you want to challenge anything above. Every number in the games traces to one of them.
| File | What it holds |
|---|---|
research/00-game-inventory.md |
Every local source, side by side, and what none of them says |
research/01-mechanics-spec.md |
The build contract. Every rule tagged [REPORTED] or [RECONSTRUCTED], with its source |
research/02-online-research-log.md |
All 54 network operations, including the ones that found nothing, plus the claims examined and discarded |
research/03-game-roster.md |
How the roster conflict was resolved, the conflict table, and the three disagreements still standing |
research/04-qa-report.md |
What QA found and fixed, including my own mistakes |
data/reviews-online.csv |
18 online sources, REV-101–REV-118, each dated and credibility-rated |
data/build-plan.md |
The ranked build list and the evidence-coverage table |
PROGRESS.md |
Phase-by-phase record of how this was built |
_backlog/PARKED.md |
Findings about other rounds that surfaced here |
Code
| Path | What |
|---|---|
engine/ |
Shared modules — seeded RNG, timer, scoring, storage, results screen, battery state, stylesheet. One implementation, so a fix lands in every game |
behavioural/items.json · items.js |
The 54 behavioural statements (27 traits × forward/reverse). Original wording — nothing copied from any real inventory |
verify_games.py |
The test harness. Drives every game in real Chrome |
Everything is a plain file. No build step, no npm, no server. Games load engine/*.js as
classic scripts, which works from file:// — so keep the folder structure as it is.
To re-run the tests
python3 verify_games.py
It drives all four pages in real Google Chrome: full sessions, deliberate wrong answers, deliberate timeouts, perfect runs, mid-session refreshes. Every claim has a negative control, and the generated arithmetic and both path optima are re-derived independently in Python and compared against what the games score you on.
Three things to be honest about
- No photograph of any of these game screens exists — not locally, not online. Every screen here is designed from candidate descriptions and coaching reconstructions. The mechanics are well sourced; the look is mine.
- Nobody has ever stated what a wrong answer costs. Across 54 network operations, no candidate, coach or vendor says. Every scoring rule in these games is reconstructed, labelled as such in-page, and toggleable.
- The newest first-hand account describes the batch-2026 sitting. The target is batch 2027. The roster is well evidenced for Oct 2025 – Sep 2026 and nothing guarantees it is unchanged. The games train transferable skills for exactly that reason.