The study employed a two-arm, parallel-group, cluster randomized controlled trial (Cluster-RCT) design. Given the educational setting, randomisation was performed at the class level rather than the individual student level to maintain ecological validity and prevent contamination between experimental conditions within the same instructional environment (Li et al., 2017). Intact classes were randomly assigned to either the experimental group (EG), which received the gamified didactic module, or the control group (CG), which followed a traditional direct instruction curriculum. Assessments were conducted at two time points: baseline (T0) and post-intervention (T1), separated by a 12-week intervention period. The randomisation involved exactly 10 intact classes (clusters), assigned equally to the EG (k = 5) and the CG (k = 5), with an average cluster size (m) of 21 students. The study adhered strictly to the CONSORT guidelines extended for cluster trials (Campbell et al., 2012), and a detailed flowchart of participant enrolment, allocation, and analysis is provided in Figure 1.

Participants were recruited from three public high schools located in Northern Italy. The initial pool consisted of 225 students. Inclusion criteria were: (a) enrolment in the first or second year of secondary school (aged 14–17 years); (b) absence of physical or cognitive disabilities that would preclude safe participation in motor testing; and (c) willingness to adhere to the measurement protocol. Students who were professional athletes (training > 10 hours/week) were excluded to avoid confounding effects related to advanced motor competence (Burton et al., 2023). Based on these criteria, 15 students were excluded. The final sample consisted of 210 adolescents (118 males, 92 females; mean age = 15.4 ± 0.9 years).

An a priori power analysis was initially conducted using G*Power 3.1 (Faul et al., 2007) for a repeated measures framework (f = 0.25), power = .80, α = .05), indicating a baseline requirement of 128 individuals. However, considering the nested structure, the Design Effect (DE) was explicitly calculated to determine the required sample size for a cluster trial. Assuming an average cluster size (m) of 21 and a conservative a priori Intraclass Correlation Coefficient (ICC) of .05 based on similar school-based trials (Salway et al., 2024), the DE was estimated as 1 + (m – 1 ) x ICC = 2.0. Consequently, the cluster-adjusted required sample size was 128 × 2.0 = 256 students. While the final recruited sample (N = 210) was slightly below this optimal target, the risk of underpowering was mitigated through the use of robust parameter estimation within the linear mixed models (LMM) and the inclusion of powerful individual-level covariates (e.g., maturity offset, baseline scores) known to substantially reduce residual variance and enhance statistical power in mixed-effects designs.

Ethical approval for this study was granted by the University Territorial Ethics Committee (Reference: UTEC 2024–0029). The research was conducted in accordance with the ethical principles of the Declaration of Helsinki. Written informed consent was obtained from the parents or legal guardians of all participants, and written assent was provided by the students before participation. Personal data were collected and processed in compliance with the General Data Protection Regulation (GDPR).

To assess the multidimensional physical self, the Italian Revised Physical Self-Perception Profile (PSPP-R-IT) was utilized (Nicolosi et al., 2024). This instrument, validated specifically for the Italian adolescent population (15–20 years), consists of 30 items assessing four specific subdomains: sports competence (SC), physical conditioning (PC), body attractiveness (BA), and physical strength (PS). Additionally, a superordinate domain, physical self-worth (PSW), evaluates general feelings of happiness and satisfaction with the physical self. Items are rated on a 4-point structured alternative format to minimise social desirability bias. The Italian version has demonstrated excellent internal consistency (Cronbach’s α = 0.89–0.95) and construct validity (Nicolosi et al., 2024).

The Italian Perceived Importance Profile (PIP-IT) was administered to measure the subjective value students attach to each of the PSPP-R subdomains (e.g., how important it is for the student to be good at sports). This scale allows for the examination of the discrepancy between competence and importance, a key predictor of self-esteem according to James’s hypothesis (Fox & Corbin, 1989; Nicolosi et al., 2024).

Objective physical fitness was assessed using the High-Priority ALPHA Health-Related Fitness Test Battery, an evidence-based gold standard for European adolescents (Ruiz et al., 2011). Importantly, the variables included in the present study (i.e., cardiorespiratory fitness and musculoskeletal strength) were employed as operational proxies of physical performance related to perceived competence, rather than as direct indicators of skill-based motor competence. The selected tests were chosen to align with the domains of the PSPP-R, as detailed below.

Cardiorespiratory fitness (corresponding to physical condition, PC) was assessed using the 20-m shuttle run test. Participants ran repeatedly between two lines set 20 m apart, synchronising their pace with audio signals that increased progressively in frequency at one-minute intervals. The final stage successfully completed was recorded and used to estimate maximal oxygen uptake (V̇O₂max) according to the equation proposed by Léger et al. (1988).

Musculoskeletal fitness (corresponding to physical strength, PS) was evaluated using two complementary measures. Lower-body explosive strength was assessed through the standing broad jump, in which participants jumped horizontally from a stationary position; the longest distance achieved across two attempts was recorded in centimetres. Upper-body strength was assessed via handgrip strength using a Jamar Plus+ digital dynamometer, with grip span individually adjusted according to hand size, following standardised recommendations (Ruiz et al., 2011).

Agility and speed were assessed using a timed 4 × 10 m Shuttle Run Test. While recognizing that this test does not measure contextual or tactical sporting intelligence (as required by broader definitions of sport competence), it is a validated measure of speed of movement, agility, and coordination (Ortega et al., 2008), which serve as fundamental physical proxies for the sport competence domain assessed in the PSPP-R-IT.

The emotional response to the PE lessons was measured using the Physical Activity Enjoyment Scale (PACES), Italian version (Carraro et al., 2008). The scale comprises 16 items rated on a 5-point bipolar Likert scale (e.g., “I enjoy it” vs “I hate it”). Prior to data analysis, the seven items with negative valence were reverse-scored. A total enjoyment score was then computed by summing the responses to all 16 items, in accordance with the protocol established by Carraro et al. (2008). A higher total score indicates greater enjoyment.

To control for the confounding effects of growth and maturation on physical performance (Burton et al., 2023), standing height, sitting height, and body mass were measured. These values were used to calculate the maturity offset (years from peak height velocity) using the sex-specific equations proposed by Mirwald et al. (2002).

To ensure the internal validity of the study, a treatment fidelity checklist was developed based on the framework by Li & Xiang (2023). External researchers observed a random subsample (20%) of the lessons in both groups to verify adherence to the assigned protocol. The checklist assessed key indicators such as “use of gamified language,” “provision of autonomy-supportive feedback,” and “adherence to lesson content.” Only data from classes meeting a fidelity threshold of >80% were included in the final analysis.

Students assigned to the EG participated in MoveQuest, a 12-week gamified instructional module designed to support basic psychological needs. The didactic approach was implemented through a dedicated mobile web application, in which physical activities were framed as progressively structured “missions”.

Game mechanics were based on the completion of motor tasks (e.g., completing five laps of the shuttle run), which rewarded students with experience points (XP). The accumulation of XP enabled progression through hierarchical levels (e.g., from novice to elite), thereby providing structured feedback on engagement and progression.

Assessment followed a criterion-referenced logic (Yang et al., 2024), whereby students’ performance was evaluated against predefined standards rather than through normative peer comparison. For instance, an individual improvement of 5 cm in the standing broad jump resulted in the award of a “personal best” badge, irrespective of the absolute distance achieved.

From a pedagogical perspective, the teacher assumed the role of a “game master”, fostering autonomy by allowing students to select task difficulty levels (“challenge zones”) that were congruent with their perceived competence and current ability.

Students in the CG followed a standard multi-activity physical education curriculum covering the same core physical content (i.e., endurance, strength, and agility) as the experimental group, but delivered through direct instruction. Lessons typically adhered to a conventional structure consisting of a warm-up phase, teacher-led demonstration, structured practice in the form of drills, and a closing phase.

Assessment and feedback were based on a norm-referenced approach (Yang et al., 2024). Students’ performance was evaluated relative to their peers and commonly aligned with standard school grading norms. For example, specific performance thresholds (e.g., six completed laps corresponding to Grade 6, eight laps to Grade 8) were used to rank or grade students, thereby emphasising social comparison.

Data were analysed using IBM SPSS Statistics (version 29.0). Preliminary screening included the Shapiro–Wilk test to assess normality and Levene’s test to evaluate homogeneity of variance.

Given the hierarchical structure of the data, with students nested within classes, the use of standard analysis of variance would have violated the assumption of independence of observations (Li et al., 2017). Consequently, linear mixed models (LMMs) were employed to examine intervention effects. Models were estimated using Restricted Maximum Likelihood (REML) to provide unbiased estimates of variance and covariance parameters, which is particularly appropriate for small-to-moderate numbers of clusters. Degrees of freedom for fixed effects were computed using Satterthwaite’s approximation.

The final models specified fixed effects for group, time, and the group × time interaction. With respect to the random-effects structure, random intercepts were included at both the class level (Class ID), to account for clustering effects, and the individual level (Student ID), to model repeated measurements. Random slopes for Time at the class level were initially tested but subsequently removed, as they did not improve model fit according to Akaike’s information criterion (AIC) and Bayesian information criterion (BIC), or resulted in convergence issues.

Baseline scores, mean-centered Perceived Importance (PIP-IT scores), and Maturity Offset were included as covariates to control for pre-existing differences and inter-individual variability in biological maturation.

Outcomes were hierarchically defined a priori, with physical self-perception, health-related fitness, and enjoyment specified as primary outcomes, and Perceived Importance as a secondary outcome. Given the multiple outcomes assessed, the risk of inflated Type I error was acknowledged. However, rather than applying a highly conservative global adjustment (e.g., Bonferroni) that could obscure theoretically grounded specific effects, exact p-values were reported alongside the corresponding effect sizes (d), with significance interpreted primarily on the basis of effect magnitude and a priori hierarchical hypotheses. Effect sizes were calculated as Cohen’s d for between-group differences at T1, dividing the difference in estimated marginal means by the pooled baseline standard deviation (SDpooled). Magnitude was interpreted as small (0.2), medium (0.5), or large (0.8), in accordance with established conventions (Wang & Chen, 2022). Statistical significance was set at α < .05.

No missing data were observed for the primary outcomes as make-up testing sessions were organized for absent students. The analysis of treatment fidelity, conducted on 24 randomly selected lessons (20% of the total), indicated high adherence to the protocol. The mean fidelity score was 92% for the EG (gamified) and 94% for the CG (traditional), exceeding the 80% threshold required for inclusion (Li & Xiang, 2023). Internal consistency reliability for the psychometric instruments in the current sample was satisfactory. Cronbach’s alpha coefficients for the PSPP-R-IT subscales ranged from .83 to .88 (sports competence = .86; physical conditioning = .84; body attractiveness = .88; physical strength = .83; physical self-worth = .87). The PIP-IT subscales also demonstrated adequate reliability (α range = .79–.85), and the PACES showed excellent internal consistency ( = .92).

Baseline characteristics of the participants are presented in Table 1. Independent samples t-tests revealed no significant differences between the experimental and control groups at baseline for age (p = .64), BMI (p = 0.58), maturity offset (p = .72), or any of the motor competence and psychological variables (p > .05). This confirms that randomisation was successful in creating equivalent groups.

The Intraclass Correlation Coefficients (ICC) for the primary outcomes indicated non-negligible clustering effects, confirming the necessity of using linear mixed models to account for the class-level variance. Specifically, the baseline variance components (between-class variance [τ00] and residual individual variance [σ2]) were as follows: Sports competence (τ00 = 0.024, σ2 = 0.324, ICC = .07), physical self-worth (τ00 = 0.031, σ2 = 0.312, ICC = .09), V̇O2max (τ00 = 1.05, σ2 = 25.1, ICC = .04), and PACES (τ00 = 0.041, σ2 = 0.472, ICC = .08).

Note. Data are presented as Mean (Standard Deviation) unless otherwise indicated. BMI = Body Mass Index; PACES = Physical Activity Enjoyment Scale. p-values derived from independent samples t-tests (for continuous variables) and Chi-square tests (for categorical variables).

The linear mixed model revealed a significant group × time interaction for the sports competence subscale (F(1,208) = 14.52, p < .001) (β = 0.35, SE = 0.09, 95% CI[0.17, 0.53]). Follow-up pairwise comparisons indicated that while both groups improved, the EG showed a significantly larger increase from T0 (M = 2.45, SD = 0.60) to T1 (M = 2.98, SD = 0.55) compared to the CG (T0: M = 2.42, SD = 0.58; T1: M = 2.60, SD = 0.62). The effect size for the between-group difference at post-test was medium-to-large (d = 0.65). Similarly, a significant interaction was found for the superordinate domain of physical self-worth (F(1,208) = 11.34, p = .001) (β = 0.31, SE = 0.09, 95% CI[0.13, 0.49]). The Gamified group demonstrated a substantial improvement (d = 0.58), whereas the CG showed only marginal gains.

In contrast, no significant interaction effect was observed for the body attractiveness subscale (p = .42). Scores for this domain remained relatively stable in both groups over the 12-week period, suggesting that perception of body image was not significantly altered by the short-term didactic intervention.

Results for the ALPHA-FIT battery tests are summarized in Table 2. For cardiorespiratory fitness (20-m Shuttle Run), a significant main effect of Time was observed (F(1,208) = 35.40, p < .001), with both groups improving their V̇O2max estimates. However, the group × time interaction was also significant (p = .023) (β = 2.21, SE = 0.96, 95% CI[0.31, 4.11]), with the EG showing a greater magnitude of improvement (d = 0.42) compared to the CG (d = 0.21).

Regarding musculoskeletal fitness (standing broad jump), both groups showed significant improvements over time (p < .001), but the interaction effect was not significant (p = .18), indicating similar gains regardless of the teaching method.

For the 4 x 10m shuttle run (sports competence/agility), the analysis revealed a significant interaction (p = .004) (β = -0.50, SE = 0.17, 95% CI[-0.84, -0.16]). Students in the gamified condition improved their agility times significantly more than those in the traditional condition (d = 0.51) vs (d = 0.25).

Note. Exp = Experimental Group; Con = Control Group. β represents the fixed effect estimate for the Group × Time interaction derived from Linear Mixed Models (REML estimation, Satterthwaite df). represents the between-class variance (random intercept variance at cluster level); represents the residual variance (within-individual variance). ICC = Intraclass Correlation Coefficient (). †Lower values indicate better performance (negative β indicates greater improvement in Exp); *p < .05, **p ≤ .01, ***p ≤ .001.

A strong group × time interaction was found for PACES scores (F(1,208) = 28.6, p < .001) (β = 0.95, SE = 0.18, 95% CI[0.60, 1.30]). The EG reported a marked increase in enjoyment (d = 0.82, large effect), whereas enjoyment scores in the CG remained static or slightly declined (d = 0.10).

The inclusion of mean-centered PIP-IT scores as covariates in the LMM revealed a significant moderating effect. For physical self-worth, the three-way interaction between Group Time Sports Competence Importance score was significant (p = .015, β = 0.18, SE = 0.07, 95% CI[0.04, 0.32]). As illustrated in the interaction plot (Figure 2), students in the gamified group who attributed high importance to sports competence (+1 SD above the mean PIP-IT score) experienced the greatest gains in self-worth d = 0.72) compared to those with low perceived importance (-1 SD; d = 0.35). This suggests that the intervention was particularly effective for students who already value the domain being trained.

A key finding of this investigation was the superior improvement in sports competence and physical self-worth observed in the gamified group compared to the traditional instruction group. Furthermore, the medium-to-large effect size observed for physical self-worth (d = 0.58) is substantial, especially when benchmarked against recent meta-analyses indicating that physical activity-based interventions typically yield more modest pooled effects on physical self-concept in adolescents (Zamorano-García et al., 2023). This result aligns with the theoretical tenets of Self-Determination Theory embedded in the intervention design. As anticipated, replacing normative peer comparisons with criterion-referenced mastery goals likely fostered a more supportive environment for competence satisfaction (Yang et al., 2024). This shift is supported by recent findings indicating that reducing the sense of public judgment and surveillance in fitness activities is essential for fostering student agency and comfort (Alfrey & Young, 2025). According to Ryan & Deci (2020), comparative grading schemes typically provide little competence-relevant feedback, merely informing students of their rank relative to others. By removing this ‘controlling’ pressure—which inadvertently highlights incompetence in less able students (Li et al., 2017)—the gamified intervention allowed feedback to be perceived as informational rather than evaluative. This distinction is critical: as observed in a previous qualitative analysis of Italian students (Cereda, 2025), technology can create a detrimental “digital panopticon” when used for public comparison. By contrast, the current intervention utilized digital tools to foster private progression, effectively mitigating the anxiety and ‘shame’ associated with public performance tracking.

The significant increase in enjoyment (PACES) within the experimental group further corroborates this interpretation. The magnitude of this effect (d = 0.82) is particularly notable when compared to recent meta-analytical pooled estimates for game-based PE interventions (e.g., 0.53 reported by Mo et al., 2024), confirming the high emotional engagement elicited by the didactic format. This observation is consistent with Ashley & Kawabata (2023), who identified the experience of enjoyment as a pivotal factor in shaping positive perceptions of fitness testing, specifically within a Self-Determination Theory framework. As suggested by Iannaccone et al. (2025), the affective domain is a critical component of physical literacy. This is consistent with recent interdisciplinary reviews highlighting that mastery-oriented learning environments, which prioritize personal improvement over normative comparison, are essential for fostering positive motivational patterns and comprehensive physical literacy (Cereda, 2026). The “fun” element inherent in gamification—achieved through badges, levels, and narrative—appears to act as a catalyst, transforming repetitive motor tasks into engaging challenges. This emotional engagement is crucial during adolescence, a period often marked by a decline in PE interest and participation (Burton et al., 2023).

However, it is noteworthy that Body Attractiveness perceptions did not change significantly in either group. This stability is consistent with the findings of Nicolosi et al. (2024), who suggest that body image is a deeply entrenched construct, heavily influenced by external societal factors and pubertal changes, and thus less permeable to short-term school-based interventions compared to competence-based domains.

Regarding health-related fitness, the results were mixed. While both groups improved in musculoskeletal fitness (standing broad jump), suggesting that the physiological stimulus of the lessons was effective regardless of the pedagogical wrapper, the gamified group showed significantly greater gains in cardiorespiratory fitness and agility. This differential effect might potentially reflect variations in behavioral engagement and “time on task”. In the gamified condition, the desire to unlock the next level or earn a badge may have encouraged students to sustain a higher volume of active practice during the shuttle runs, potentially contributing to greater physiological adaptations. However, as internal load (e.g., heart rate or Ratings of Perceived Exertion) was not monitored, definitive conclusions regarding higher cardiovascular intensity in the gamified group remain unwarranted. Therefore, any causal inferences regarding physiological mechanisms require cautious interpretation and future objective verification. This supports the notion that didactic strategies influence how students move, not just that they move (Wang & Chen, 2022). Indeed, recent evidence indicates that participation in structured physical activity alone does not guarantee motor competence if the pedagogical focus remains on repetitive technical drills rather than diverse movement experiences. The use of the ALPHA-FIT battery (Ruiz et al., 2011) ensured that these findings are robust and comparable with European normative data, adding validity to the observation that gamification can support health-related fitness without compromising skill acquisition. Furthermore, a distinction must be drawn between the fitness components assessed here (fitness and agility) and motor competence (skill mastery). However, recent theoretical reconceptualization suggests that improved product-oriented outcomes (e.g., jump distance, running speed) may reflect a shift from stability to mobility movement strategies, signalling a breakthrough of the proficiency barrier (Brian et al., 2020). Thus, the improvements observed likely reflect both physiological adaptations and enhanced movement efficiency.

A novel contribution of this study, enabled by the use of the PIP-IT (Nicolosi et al., 2024), was the identification of perceived importance as a significant covariate. The finding that gains in self-worth were magnified for students who valued sports competence suggests that educational interventions are not one-size-fits-all. Gamification seems particularly potent for students who already have a vested interest in the domain. However, this also raises a pedagogical challenge: future interventions may need to explicitly target the value students assign to physical activity (perhaps through cognitive engagement or reflection) to broaden the impact to those who are currently disengaged or devalue PE.

The findings of the present study should be interpreted in light of several limitations. First, the intervention duration (12 weeks), while typical for school-based research, may be insufficient to observe changes in more stable constructs such as body image. Furthermore, it is not possible to definitively rule out a novelty response (e.g., the 'Hawthorne effect'), whereby the initial engagement elicited by the technological and gamified elements might wane over time. Longitudinal follow-ups are therefore essential to assess the long-term retention of the observed psychological and physiological gains. Second, although the cluster-randomized design is methodologically appropriate for educational settings, it inherently reduces statistical power compared to individual randomization. While the use of Linear Mixed Models (LMM) mitigated this risk by correctly modeling the nested variance (Li et al., 2017), the present sample size was not adequately powered to conduct complex three-way interactions (e.g., exploring gender-specific or maturity-specific trajectories), which remains an important avenue for future research. Third, despite the rigorous assessment of treatment fidelity, it was not possible to fully isolate a potential 'teacher effect'; the enthusiasm or specific pedagogical style of individual educators may have clustered within classes, thereby influencing the outcomes. Finally, the study was conducted in a specific region of Northern Italy, which may limit generalizability to other cultural or educational contexts.

Conversely, notable strengths of the study include the rigorous adherence to CONSORT guidelines extended for cluster trials, the explicit calculation of the design effect, and the high treatment fidelity verified via a standardized checklist (Li & Xiang, 2023). Furthermore, the concurrent use of objective health-related fitness testing (ALPHA-FIT) and valid, culturally adapted multidimensional psychological instruments (PSPP-R-IT, PIP-IT, PACES) provides a comprehensive, dual-domain evaluation that addresses the fragmentation frequently observed in previous gamification research (Arufe-Giráldez et al., 2022; Carl et al., 2022). This combined approach is particularly relevant, as recent systematic evidence confirms that physical fitness outcomes and motor competence variables continue to be largely absent from gamification research in secondary PE (Sal-de-Rellán et al., 2025).

During the preparation of this manuscript, the author used DeepL Write to improve linguistic clarity, grammar, and style, as English is not the author's native language. All AI-generated outputs were critically reviewed and thoroughly edited by the author to ensure factual accuracy, clarity of expression, and compliance with academic standards. The author takes full responsibility for the integrity and content of this manuscript.