The Role of Age and Emotional Bond in Games

1. Introduction

The study of demographic and relational factors in cooperative gaming remains relatively under-explored, despite growing evidence that both age and pre-existing social ties can shape performance, motivation, and learning outcomes. From a motivational perspective, cooperative goal structures and friendship bonds have been shown to enhance commitment and effort in motor-centered games: Peng and Hsieh (2012)’s balloon-popping experiment demonstrated that cooperation induces greater effort than competition, and that playing with friends further amplifies goal commitment—but only in cooperative contexts. Related to this, a meta-analysis (Chung et al., 2018) involving 1,016 groups revealed a small but reliable friendship advantage in group task performance, particularly for larger teams and output-focused tasks. Taken together, these findings suggest that familiarity can strengthen cooperative engagement and performance under the right task conditions. 

Age introduces a complex interaction between declining cognitive abilities and the potential benefits of social interaction. Research conducted by Horn et al. (2021) showed that in resource-allocation games, younger adults outperformed older adults in strategic decisions and overall rates, with fluid cognitive abilities influencing these results. Another study carried out by Crompton et al. (2022) focused on collaborative learning performance in different groups in terms of relationship level and age. Over multiple trials, a learning effect was observed in both younger and older teams, although more mature participants did not perform with similar efficiency to the adolescent group. 

Building on these conclusions, our study examines performance in ORTHO (Potęga vel Żabik et al., 2024), a two-player game in which one participant controls the x-axis and the other the y-axis of a virtual ball as it moves along a track. ORTHO focuses on sensorimotor coordination and joint decision-making. Grounded in Peng and Hsieh (2012)’s insights on cooperative motivation and Chung et al. (2018)’s friendship-performance link, we predict that the best performance is achieved by those who maintain a friend–friend relationship. Moreover, following on from Horn et al. (2021)’s evidence for youthful strategic superiority and having in mind the results of Crompton et al. (2022)’s research, we hypothesize that teenagers will outperform older adults. To evaluate these hypotheses, we analyse ORTHO’s detailed log data using different performance metrics and nonparametric tests (Kruskal–Wallis, Wilcoxon). 

The full codebase and experiments are available in our repository. 

2. Methodology 

The ORTHO study employs a custom-built interactive game environment to collect movement data from participant pairs. The data collection system generates structured JSON files, each representing a single day of gameplay sessions. Within these files, data is hierarchically organized with game sessions containing players’ personal information (age, relationship type) and details about tracks, including their difficulty. The data also incorporates game session details such as start/end times, completion status, points recording pre cise coordinates, timestamps, and status codes (0-4) for ball position throughout gameplay. Each session is uniquely identified with timestamps recorded for all significant events, which potentially might be useful in further explorations. 

For analysis purposes, we developed a Python-based data transformation pipeline to convert a hierarchical JSON structure into a tabular format. The pipeline traverses all JSON files in the data directory, extracts session metadata and track information. Then the script generates two types of CSV outputs: a consolidated file (games.csv) containing all session and track metadata, and individual files for each game including detailed trajectory data. This transformation also standardizes identifiers across sessions and games. 

Key variables extracted for our analysis include participants’ ages (in years), relationship types (coded as: 0=family member, 1=friend, 2=stranger, 3=significant other), completion times (in milliseconds), and success/failure indicators. We developed three metrics to evaluate performance: completion time, completion rate and error rate. The completion time measure indicates the average time needed to pass the track in the considered group. The completion rate describes how frequently a given part of players successfully completed a track, represented as a proportion from 0 to 1. On the other hand the error rate quantifies how often participants steered the ball off the designated track, also scaled from 0 to 1. 

For categorical analysis, we grouped participant ages into five brackets ("under 18", "18-29", "30-44", "45-59", "60+") and aggregated relationship types into five categories: "Family-Family", "Significant Other-Significant Other", "Stranger-Stranger", "Friend-Friend", and "Complicated" (for declared mixed relationships on both sides). During the age group classifying, each game was attached to a certain class twice, separately for both participants. 

3. Experiments 

Our experimental approach involved a comprehensive analysis of gaming performance across all seven available tracks, recognizing that each track presents unique challenges and difficulty levels. We systematically evaluated three key performance metrics—completion time, completion rate, and error rate—while controlling for track-specific variations that could confound cross-demographic comparisons. 

The age analysis revealed striking and consistent patterns across all tracks. Figure 1b demonstrates that participants aged 18-29 consistently outperformed all other age groups, achieving the fastest average completion time of 32,787 ms and maintaining the best overall ranking (1.0) across all tracks. This superior performance supports our initial hypothesis regarding the advantages of youthful cognitive abilities in cooperative gaming scenarios. Contrary to expectations, participants under 18 did not achieve optimal performance despite their presumed cognitive flexibility. Instead, they ranked fifth overall (4.4) with an average completion time of 41,004 ms, which may be due to the fact that this group is heav ily represented, but could just as well be due to their lack of understanding of Cartesian coordinates.

Figure 1: Impact of age and relationships on Ortho game performance. (a) Error rates by

age groups across different tracks. (b) Average completion times by age groups across all

tracks. (c) Average completion times across tracks by relation groups. (d) Average times of

each age group on track 1. The horizontal lines above the bar graph show which pairs of

groups are statistically significantly different. *** denote differences at the 0.001 confidence

level, ** at the 0.01 confidence level, * at the 0.05 confidence level.

The 45-59 age group presented an interesting paradox: while achieving the lowest error rate (0.479) among all groups (which can be seen in the table 1), they demonstrated slower completion times (39,911 ms), which may indicate a cautious but deliberate approach to gameplay. This could be caused by age-related strategic adaptations that prioritize accuracy over speed. Most surprisingly, participants aged 60+ achieved a relatively strong overall ranking (2.4) despite having the highest error rate (0.544) among age groups. 

The relationship analysis provided the most interesting performance differences observed in our study. Figure 1c illustrates that significant other pairs achieved the best results, with the fastest average completion time (32,465ms) and perfect ranking (1.0) across all tracks, as seen in Table 2. This performance advantage over other relationship types might suggest that intimate relationships create better conditions for cooperative gaming through enhanced communication, trust, and coordination. Friend pairs, while hypothesized to per form best due to competitive dynamics, actually ranked third overall (2.9) with completion times of 35,760ms. However, they showed higher error rates (0.527) compared to family pairs, indicating a reliance on speed rather than accuracy, potentially due to a motivation to compete rather than cooperate. The "complicated relationship" category emerged as an unexpected high performer, achieving second-best ranking (2.3) with completion times of 33,950ms. This type of relationship may be due to the non-serious approach of game par ticipants to the survey, making it impossible to draw meaningful conclusions. Family pairs, despite having the lowest error rate (0.472) across all relationship types, showed the slowest completion times (41,166ms) and worst overall ranking (4.9). This pattern could indicate that familial relationships prefer a strategy that prioritizes caution and consensus-building over rapid task completion. 

We also conducted a comprehensive statistical analysis that employed multiple approaches to verify differences between groups. We used t-tests for independent samples to compare means between all pairs of groups, with results visualized through significance bars on our charts. To address the multiple comparisons problem, we applied Bonferroni correction to control the family-wise error rate. Additionally, we employed Kruskal-Wallis tests to verify differences between groups and Wilcoxon tests for pairwise comparisons with balanced 

sampling. Kruskal-Wallis tests revealed statistically significant differences (p < 0.001) across all categories. Detailed Wilcoxon tests revealed that 60-80% of pairwise group comparisons were statistically significant at α = 0.05 level. Figure 1d presents the mean completion times for different age groups on Track 1, with horizontal lines above the bars indicating which pairs of groups show statistically significant differences. 

Our approach also contained distribution analysis, in which Kolmogorov-Smirnov tests were used to examine whether completion times follow log-normal or Weibull distributions across different groups and tracks. For log-normal distribution, most groups failed to satisfy the distributional assumptions (p < 0.05), with only a few exceptions, including strangers on track 1, several groups on tracks 2 and 4, and the 60+ age group on track 1. Similarly, Weibull distribution testing showed limited fit, with significant deviations observed across most demographic categories and tracks. The results indicate that completion times in the majority of considered groups do not consistently follow standard parametric distributions. 

4. Conclusions 

Consistent with Horn et al. (2021), young adults (18–29yrs) perform the fastest comple tion times, suggesting that this age group combines above-average fluid cognitive abilities with mature motor coordination and high technological familiarity. The results remain consistent with Crompton et al. (2022) who also reported that younger (18–28yrs) participants maintained an advantage over older adults in embodied tasks. This contradicts our hypothesis, which predicted that teenagers would perform better. 

Surprisingly, the strongest relationship effect was observed not among friend–friend pairs—as in Peng and Hsieh (2012) and Chung et al. (2018) —but among intimate (significant other) pairs. While Peng and Hsieh (2012) demonstrated that friends commit more effort under cooperation and Chung et al. (2018) found moderate friendship gains in group tasks, our research suggest that deeper mutual trust and nonverbal connection in intimate rela tionships may further enhance synchronized decision-making in sensorimotor games. Similar observations were noted by Pollmann and Krahmer (2018), while studying the impact of re lationships and non-verbal communication in the game Taboo. 

Our contributions are summarized as follows: 

• We investigate the impact of age and relationship between players on performance in the cooperative games. 

• We are analysing more than 100,000 ORTHO game sessions, making this one of the largest studies on cooperation in sensorimotor games. 

• We indicate that intimate relationships provide a better foundation for cooperative gaming. 

• We suggest that stronger interpersonal relationships may be more important than pure competitive motivation, which has important implications for the design of collabora tive environments. 

• We imply that young adults combine above-average fluent cognitive abilities with mature motor coordination and high technological literacy, which makes them the best performers in the game. 

---

Writiten by Mikołaj Rowicki, Jakub Półtorak, and Filip Langiewicz

References

• Sangcheol Chung, Robert B Lount, Hyo Min Park, and Eun Seon Park. Friends with performance benefits: A meta-analysis on the relationship between friendship and group performance. Personality and Social Psychology Bulletin, 44(1):63–79, January 2018. doi: 10.1177/0146167217733069. URL https://doi.org/10.1177/0146167217733069. Epub 2017 Oct 10. 

• Catherine J Crompton, Maria K Wolters, and Sarah E MacPherson. Learning with friends and strangers: partner familiarity does not improve collaborative learning performance in younger and older adults. Memory, 30(5):636–649, May 2022. doi: 10.1080/09658211. 2022.2041038. URL https://doi.org/10.1080/09658211.2022.2041038. Epub 2022 Feb 22. 

• Sebastian S Horn, Judith Avrahami, Ya’akov Kareev, et al. Age-related differences in strate gic competition. Scientific Reports, 11:15318, July 2021. doi: 10.1038/s41598-021-94626-2. URL https://doi.org/10.1038/s41598-021-94626-2. Received 15 March 2021; Ac cepted 13 July 2021; Published 28 July 2021. 

• Wei Peng and Gary Hsieh. The influence of competition, cooperation, and player relationship in a motor performance centered computer game. Computers in Human Behavior, 28(6): 2100–2106, 2012. ISSN 0747-5632. doi: https://doi.org/10.1016/j.chb.2012.06.014. 

• Marloes M H Pollmann and Emiel J Krahmer. How do friends and strangers play the game Taboo? a study of accuracy, efficiency, motivation, and the use of shared knowledge. Journal of Language and Social Psychology, 37(4):497–517, September 2018. doi: 10. 1177/0261927X17736084. URL https://doi.org/10.1177/0261927X17736084. Epub 2017 Oct 12. 

• Katarzyna Potęga vel Żabik, Dor Abrahamson, and Ilona Iłowiecka-Tańska. It takes two to ortho: A tabletop action-based embodied design for the cartesian system. Digital Experiences in Mathematics Education, 10(2):189–201, 8 2024. ISSN 2199-3254. doi: 10.1007/s40751-024-00139-8. URL https://doi.org/10.1007/s40751-024-00139-8.