pymetrics Money Exchange 1: Trust Game Guide | Game Assessment Prep

There is no correct answer—and we will never score this

There is no correct answer and we will never score this preference game. Money Exchange 1 asks how much of a $10 endowment you choose to place with an algorithmic partner. Sending more is not better; sending less is not worse. The transfer is a descriptive trust signal, and different roles can value different behavioral profiles.

That distinction matters because conventional test-prep language can be actively misleading here. There is no accuracy percentage, pass mark, practice percentile, green “good” zone, or employer-approved amount on our completion screen. We show a neutral Trust reflection so you can understand what the behavior may signal without pretending to know a secret answer.

How the exchange works

You start with $10 and the partner starts with $0. You choose a transfer from $0 to $10 in one-dollar steps. The amount visibly triples in transit. If you send $4, for example, the partner receives $12 while you retain $6.

After a short “partner deciding” pause, the algorithm returns part of the tripled amount. Our default partner returns a seeded-random 30% to 70%, rounded to cents. You then rate how fair the exchange felt on a neutral 0–10 scale. The round ends with a trait screen, and the game does not report completion to the shared session shell until you select Continue.

Your final money is visible because it is part of the exchange, not because it is a score. A person can send a large amount and receive little, or send a small amount and receive a generous fraction. The partner response is simulated; there is no live candidate on the other side.

What does Money Exchange 1 signal?

The underlying paradigm is the Trust or Investment Game introduced by Berg, Dickhaut, and McCabe in 1995. The amount sent is commonly interpreted as comfort relying on another person under interpersonal uncertainty. A larger transfer exposes more of your endowment to the partner's reciprocal choice; a smaller transfer retains more certainty and can signal self-reliance.

The fairness rating adds context. It describes your reaction to what the partner returned and can reflect expectations about reciprocity. It does not turn the exchange into a test with a correct norm. Two candidates can receive the same return, rate it differently, and both provide coherent information about their own expectations.

We use “signal” deliberately. One simulated decision cannot define your character, and pymetrics does not disclose the exact formulas or role models applied in production. The exchange may contribute to a broader profile alongside eleven other games. No guide can tell you which transfer a particular employer wants.

What is known—and what remains uncertain?

Core mechanics have comparatively high confidence. The $10 endowment, tripling multiplier, and one-shot structure are corroborated across sources and closely reproduce the academic Trust Game. A 0–10 fairness scale is the recommended configurable default, although a 1–10 version remains possible.

The largest unknown is the partner-return algorithm. Pymetrics does not publish how it selects the returned amount, whether return behavior changes across candidates, or whether employer configuration affects it. We use a seeded 30–70% fraction of the tripled transfer because the build specification recommends that transparent reconstruction range. It is not a recovered production schedule.

We call the partner algorithmic rather than implying a hidden human. That framing is the most plausible for a repeatable assessment and avoids deceiving the player. The real interface wording still needs verified production footage.

How to approach the game honestly

Understand the financial state

Before confirming, check three facts: how much you retain, how much is sent, and how much the partner receives after tripling. Understanding the mechanics ensures the choice reflects trust preference rather than a multiplier mistake.

Choose your natural level of interpersonal risk

Ask what you would genuinely be comfortable placing with an unknown partner under the stated rules. Do not automatically maximize the social-looking option or minimize exposure because a preparation site suggested it.

Do not optimize against our return seed

The return is generated only after your transfer and is not predictable from prior sessions. Replaying until a favorable partner appears changes the observed outcome, not the meaning of your initial choice.

Rate the experienced exchange

Use the fairness scale to describe the return you actually saw. You do not need to make the rating mathematically match the transfer. Fairness judgments include expectations, reciprocity, and personal norms.

Keep preference and performance separate

Money retained is not a test score. A lucky return does not make a transfer superior, and a low return does not prove the decision was wrong. The task observes behavior under uncertainty rather than rewarding an optimal strategy.

Avoid role-playing an imagined employer

Trying to appear maximally trusting can make your response less authentic and may conflict with behaviors elsewhere in the battery. Pymetrics describes role-specific matching, not a universal ideal personality. Honest responses are the only defensible approach.

How to read the trait reflection

The reflection names Trust and places the selected transfer on a neutral continuum from self-reliance to comfort relying on others. It repeats the actual amount because that is the behavior you chose, not a number-as-score. No section turns it into a percent, rank, or evaluation.

The shared completion page likewise states only that preference signals were saved and that there is no score, pass mark, or practice percentile. Internally, the session log stores the seed, amount sent, tripled amount, partner return, both final balances, fairness rating, reaction times, and timestamped events. A schema-required zero placeholder may be saved in the database score column, but it is excluded from percentile processing and never presented as a player result.

Money Exchange 1 FAQ

Is sending all $10 the strongest answer?

It is the highest-exposure trust signal, not the best answer. Sending less describes greater retained certainty or self-reliance. Neither end is universally preferred.

Is the partner a real person?

No. This reconstruction explicitly uses a seeded algorithmic partner. Production pymetrics is also likely algorithmic, but its exact framing is not publicly confirmed.

Does the partner always return more than I sent?

Not necessarily. The partner returns a fraction of the tripled amount. Depending on the seeded fraction, that return can be above or below the original transfer.

Should my fairness rating match everyone else's?

No. It is a preference judgment. Use the endpoints consistently enough to express how the exchange felt to you.

Will I receive a percentile?

Never for this game. Money Exchange 1 is explicitly excluded from practice percentiles because ranking a trust preference as higher-is-better would violate the construct.