Pymetrics risk games: balloon, cards, and the trait Pymetrics is actually measuring

Two of the twelve Pymetrics games measure risk tolerance: the balloon-pump game and the card-deck task. Candidates often assume "safer is better" — that's not how the trait is scored.

The balloon game

You pump up balloons to earn money. Each pump adds value; balloons can pop at random and you lose what's in the current balloon. You decide when to bank.

What Pymetrics actually measures:

• Average pumps per balloon — pure risk appetite
• Consistency — do you adjust after a pop?
• Learning rate — how quickly do you find the optimal pump count?

The trait is not "low" or "high" — it's calibrated against the employer's benchmark. BCG's profile tends to favour moderate risk with high consistency (steady learner), not maximum reward-seeking.

The card-deck task

Four decks. Two are "good" (small wins, rare small losses). Two are "bad" (big wins, occasional huge losses). You draw cards and try to maximise money.

What it measures:

• How fast you detect the bad decks (learning under uncertainty)
• Whether you abandon a bad deck after a single bad draw (overreaction)
• Stickiness — do you keep going back to a high-reward but losing deck?

Candidates who alternate decks randomly look like noise. Candidates who lock onto a "good" deck too early miss the better one. The signal is adaptive learning, not pattern matching from prep guides.

The wrong way to "prep"

Reading playthrough guides and trying to mimic a specific pattern backfires. The scoring is comparative — playing a textbook pattern when your natural tendency is different produces an inconsistent trait profile, which actually flags you.

The right way

Play each game three times in a relaxed mood. Notice your natural tendencies. Then play your real assessment in the same mood. Pymetrics looks for stability across games — the trait estimate is more reliable when your behaviour is consistent.

What firms see

Pymetrics returns a trait vector, not a score. Employers map their high-performer benchmark and look for fit. There's no universal "good" risk profile — BCG's benchmark differs from JPM's, which differs from McKinsey's (where applicable).

Practice plan

• Day 1: Play each risk game once, untimed, just to understand mechanics.
• Day 2: Play timed, naturally. Note your pump average and deck-switch pattern.
• Day 3: Play timed in your real assessment mood (early morning, after coffee). Same conditions as the real test.

Three sessions is enough. More creates fatigue patterns that distort the trait estimate.