pymetrics Faces Game: Complete Practice Guide | Game Assessment Prep

What is the pymetrics Faces game?

Faces is an emotion-recognition task near the end of the core pymetrics battery. You see a photographed face and choose one of ten labels: Anger, Determination, Disgust, Fear, Happiness, Hope, Pain, Sadness, Surprise, or Puzzlement. Some trials show only the photo. Others add a short situation that helps distinguish expressions with overlapping facial cues.

On a context trial, judge the whole item. The face narrows the plausible emotions, while the situation can resolve close alternatives such as anger versus determination, fear versus surprise, or happiness versus hope. Neither source should be ignored or treated as an automatic answer key.

Our tutorial and scored imagery are deliberately separate. The untimed tutorial uses higher-intensity teaching portraits so the rules are clear. Three interchangeable scored banks provide 42 different low-intensity portraits with neutral mouths and restrained changes around the eyes and brows. Each session balances the three banks across its 14 trials, and repeated emotions use different identities, preventing a simple face-to-label memorization shortcut.

What does Faces measure?

The primary skill is facial-emotion recognition, often discussed as one component of emotional intelligence. You must combine cues from the eyebrows, eyes, nose, cheeks, and mouth rather than relying on one isolated feature. Accuracy is observable, so Faces is a skill game and can support a practice score.

Context trials add an integration demand. A restrained expression can plausibly fit several labels by itself, so the written situation supplies information needed to choose the best overall match. Performance on those trials can show whether you combine verbal and visual evidence consistently under time pressure.

This does not mean a short game measures your entire capacity for empathy. Real empathy includes listening, perspective-taking, culture, relationship history, and behavior over time. A ten-label photo task captures a narrow perceptual and contextual skill. We use careful language because pymetrics does not disclose its exact model, and employers may customize the capabilities emphasized in their reports.

Parameters we know—and what remains uncertain

Faces has low confidence at the UI-detail level. Pymetrics patents identify facial-affect or mind-in-the-eyes style tasks, but they do not disclose the image set, option words, trial count, or timers. No verified production recording resolves those details.

The most coherent preparation report describes 14 trials, with seven seconds for photo-only trials and 30 seconds when a story is present. Other sources claim 40 or 90 faces. We use 14 with the 7/30-second split because it forms a plausible short game and is the recommended build default, not because the number is authoritative.

The production stimulus set is unknown. It may use licensed, custom, or established research images. Our portraits are original controlled practice assets, not copied pymetrics content. Every trial keeps all ten documented Harver labels visible. The scored expressions are intentionally subtle, with three possible identities for every trial slot and additional identities for repeated emotions, while the tutorial retains clearer examples for instruction.

The real assessment is generally described as giving no right/wrong feedback. Our practice version deliberately flashes a check or cross after each response. That difference is disclosed on the instruction screen: immediate feedback teaches distinctions that an opaque simulation cannot.

Six practical strategies

1. Read the whole face first

Take one quick global impression before inspecting details. Emotion is expressed as a configuration. A smile-like mouth with tense eyes may not indicate happiness, and wide eyes alone can fit both fear and surprise.

2. Use the eyes and brows to separate close categories

Fear and surprise both often involve widened eyes. Fear tends to add brow tension and a stretched or tense mouth; surprise is more likely to show raised brows and an open, less tense mouth. Anger often lowers and draws the brows together. Sadness may raise the inner brow corners.

3. Check the nose and upper lip for disgust

Disgust is commonly signaled by a wrinkled nose, raised upper lip, or compressed expression around the center of the face. It can be confused with anger when you focus only on narrowed eyes. Look for the nose-and-lip pattern before deciding.

4. Use context to separate plausible labels

On a context trial, identify two or three labels the face could plausibly support, then use the situation to select the best overall fit. Check that the chosen label remains compatible with the expression. This prevents either the caption or one facial feature from dominating the judgment.

5. Eliminate by incompatible cues

When uncertain, remove options that clearly conflict with the expression and context. Visible eye tension may rule out relaxed happiness; a hopeful situation may distinguish a soft almost-smile from simple happiness. Reducing ten options to a close pair makes a difficult distinction more manageable without inventing details.

6. Answer within the window without panic

Seven seconds is enough for a global read, one feature check, and a choice. Do not spend the entire window searching for certainty that a subtle photo cannot provide. Context trials allow longer because you must read the situation, combine it with the expression, and then answer.

How to read your practice result

Overall accuracy is the percentage of all 14 faces labeled correctly. Context-trial accuracy isolates the photo-plus-situation subset. The most useful comparison is the gap between those measures. If photo-only accuracy is solid but context accuracy falls, slow down enough to use the situation to distinguish the closest plausible labels.

Our insight reports how many context trials were identified correctly. The detailed session log records stimulus id, trial type, the situation, all ten options, your choice, the intended label, timeout status, and reaction time. That transparency supports targeted practice without claiming access to pymetrics' private scoring formula.

A site percentile appears only after enough comparable sessions exist. It compares accuracy in this reconstruction, not job fit and not a universal emotional-intelligence rank. Image-bank familiarity can raise repeated scores, so use fresh images and focus on transferable distinctions.

Faces FAQ

Should I answer from the story or the expression?

Use both. First read the facial cues, then use the situation to resolve closely related possibilities. The best answer must fit the complete face-and-context item rather than either source in isolation.

What happens if time runs out?

The trial is recorded as unanswered and incorrect, and practice feedback shows the verified label. Move on immediately; one timeout should not affect the next face.

Are facial emotions universal?

Some broad expression patterns are recognized across cultures, but culture, display rules, context, and individual differences all matter. Ten-label tasks simplify a complex social signal. Treat the categories as the rules of this game, not a complete theory of human emotion.

Can I memorize the images?

Each 14-trial session uses different identities, including different people when an emotion repeats. Three possible portraits exist for every trial slot, so later attempts vary the faces instead of preserving one fixed face-label map. You may eventually recognize items across many attempts, so focus on why the face and context support a label; the real assessment may use a completely different image set.

Is there a pass mark?

No public pymetrics pass mark or production accuracy threshold exists. Employers use role-specific models, and the candidate normally sees no game score. Our result is a transparent practice benchmark intended to improve recognition skill and context control.