Despite many studies on burnout among educators and coaches, significant gaps remain-particularly regarding burnout among soccer coaches working with amateur and youth teams. For instance, burnout in elite soccer coaches has been depicted, with Swedish data showing that 71% of women’s league coaches and 23% of men’s league coaches experienced moderate to high emotional exhaustion (Hjälm et al., 2007). Similarly, qualitative work with elite coaches in Sweden has linked burnout to work-home conflict and inadequate recovery (Olusoga et al., 2019). However, little is known about burnout in coaches leading amateur or youth soccer, who often face limited support, varied workloads, and diverse coaching qualifications. It is also unclear how coaching experience, professional development, and workload differences contribute to risk—or potentially buffer against burnout. Addressing these gaps is essential for informing tailored strategies to support coach wellbeing across different settings. The authors in the following article attempted to explore these relationships. To this end, an attempt was made to analyze the model, which is shown in Figure 1.

Taking into account the model above the three hypotheses were created as below:

Although the present study did not directly measure basic psychological needs, variables such as age, coaching experience, professional development, and salary satisfaction can be understood as contextual factors that may shape the degree to which these needs are met.

The study involved licensed soccer coaches who were participants in a training conference organized by the Polish Football Association. All participants held UEFA coaching licenses at the time of study. In both the conference and the survey, participation was voluntary and anonymous. The participants were informed about the purpose of the study and the procedure for its conduct. The respondents completed the survey questionnaire after the conference under identical conditions and with no time limit. Data were collected using a dedicated Microsoft Forms survey, created specificity for the purpose of the study. The questionnaire was accessible only during the data collection period involving the study participants. The study received ethical approval from Ethics Board for Research Projects at the University of Physical Education and Sport in Gdansk.

The study sample consisted of 412 participants. Regarding gender, the majority identified as male (n = 386, 93.69%), while a small proportion identified as female (n = 26, 6.31%). In terms of geographical distribution, most participants were from the Pomeranian Voivodeship (n = 395, 95.87%), with a minority from other regions (n = 17, 4.13%). Participants reported various places of residence: cities (n = 142, 34.47%), towns (n = 113, 27.43%), villages (n = 105, 25.49%), and small cities (n = 52, 12.62%). With respect to education, 169 participants (41.02%) held a higher education degree in sports, 141 (34.22%) had primary or secondary education, and 102 (24.76%) reported higher education in fields other than sports. Concerning coaching qualifications, 143 participants (34.71%) held a UEFA B license, 139 (33.74%) had a UEFA C or GRASSROOTS C license, and 130 (31.55%) were qualified at the UEFA A level. Table 1 shows descriptive statistics for numerical variables.

Coaches were given instructions on how to complete the survey. They were then given access to an online survey, which they could complete on any device with internet access. The survey was completed without interference from the authors and with no time limit. All participants provided informed consent prior to the study, and their data were anonymized to ensure confidentiality. The study received approval from the Ethics Board for Research Projects at the University of Physical Education and Sport, Gdansk, Poland (Resolution number 1/15.07.2024).

Soccer Coach Questionnaire—the questionnaire is prepared for the purpose of this study and is divided into four parts which describe and investigate different aspects. The sections were named accordingly:

The questionnaire varies in terms of possible answers and does not have regular Likert scale, therefore to get more information please contact the first author. In the current study, the questionnaire is undergoing its first pilot test.

This is a self-report questionnaire which provides new burnout indicators for adults who work in different professions including education, health service, sport and many others. The questionnaire was created by Santinello (2007) and adapted to polish conditions by Jaworowska (2014). The LBQ consists of 24 items with a six-point Likert-type response scale for study of four dimensions, each with three positive and three negative elements: the psychophysical dimension (energy-exhaustion), relationships (involvement-deterioration), professional competence (efficacy-inefficacy), and existential expectations (satisfaction-disillusion). Within the scope of this study, the Cronbach’s alpha for the complete questionnaire was α = 0.85. For the purpose of this study, we examined the general score for the burnout of the investigated coaches.

This analysis was performed in the Automated Statistical Description System - SZTOS (Hryniewicz and Milewska, 2023). The results were visualized using the graphics package “ggplot2” (Wickham, 2016). The SZTOS application is software developed by the Polish company SZTOS. The software is written in R language and its distinctive feature is the generation of comprehensive descriptions of results and drawings based on logical decision processes. Effect sizes were reported accordingly to Cohen guidelines (Cohen, 1988). Preliminary data exploration showed that missing values accounted for 1.63% of the entire dataset. They were filled using the “missForests” method (Stekhoven and Bühlmann, 2012), implemented in the SZTOS program. This is a nonparametric method that uses random forests to maximize the prediction of values at missing data locations (Breiman, 2001). To analyze the results in terms of LBQ burnout scores, an analysis of variance (Fisher, 1921) was conducted with two main effects and its interaction. The analysis of variance used type III sum of squares, which means that each variable (main or interaction effect) is evaluated taking into account all others. Furthermore due to the non-normal distribution of the LBQ score, this variable was subjected to a normalization procedure. To select the best method for normalizing the LBQ score variable, a series of transformations of this variable were performed using a range of normalization methods. The R package bestNormalize (Peterson, 2021) was used for this purpose implemented in the SZTOS program. The analysis showed that the orderNorm technique yielded the best normalization properties. Compared to other popular transformations, this method returned the lowest score of fit statistics value, P/df = 0.01. Visual diagnostics of the tested models in terms of posterior predictive values, variance homogeneity, collinearity outliers, and normality of residuals performed in the performance package (Lüdecke et al., 2021) showed that models with normalized LBQ scores were closer to meeting the assumptions of the analysis of variance model than raw LBQ scores. However, the diagnostics revealed that the model’s linearity assumption was not met and that the multicollinearity statistics for the main effects were slightly elevated. However, this is not a cause for concern when testing models containing interaction components resulting from simultaneous testing of main effects. Additionally, formal comparative analyses of the models showed that models with normalized values of the LBQ variable were significantly and significantly better fitted to the data than models tested on non-normalized data. Results section includes tables comparing model performance, and diagnostic plots are shown both before and after LBQ normalization. All results reported below are expressed in results after normalization and standardized.

To analyze the results, an analysis of variance (Fisher, 1921) was conducted, which showed that the variable Age did not significantly influence LBQ score [F(1, 408) = 0.59, p = 0.444, η² = 0.00, η p 2  = 0.00]. In contrast, the variable Motivation for Education: Earning Points had a significant effect [F(1, 408) = 10.62, p = 0.001, η² = 0.02, η p 2  = 0.03]. Furthermore, the analysis of interaction effects in the tested model revealed a significant interaction between Age and Motivation for Education: Earning Points in predicting LBQ score [F(1, 408) = 8.47, p = 0.004, η² = 0.02, η p 2  = 0.02]. The analysis of the model’s R² coefficient and its adjusted value indicated that the model explained approximately 5.09% (adjusted: 4.39%) of the variance in LBQ score. Figure 2 shows that the relationship between Age and LBQ score differed between the “Points-Motivated” and “Non-Points-Motivated” groups. The magnitude of this difference was small, with Cohen’s d = 0.03, 95% CI [0.01, 0.05]. In the Non-Points-Motivated group, Age was not significantly related to LBQ score, b = 0.00, t(408) = −0.77, p = 0.444. In contrast, in the Points-Motivated group, higher Age was significantly associated with lower LBQ score, b = −0.03, t(408) = −4.10, p < 0.001 (Tables 2, 3).

To analyze the results, an analysis of variance (Fisher, 1921) was conducted, which showed that Age had a significant effect on LBQ score [F(1, 406) = 15.22, p < 0.001, η² = 0.04, η p 2  = 0.04], as did Coach qualifications [F(2, 406) = 3.12, p = 0.045, η² = 0.01, η p 2  = 0.02]. Furthermore, the analysis of interaction effects in the tested model revealed a significant interaction between Age and Coach qualifications in predicting LBQ score [F(2, 406) = 3.60, p = 0.028, η² = 0.02, η p 2  = 0.02]. The model explained approximately 5.08% of the variance in LBQ score (adjusted: 3.91%). Figure 3 shows that the effect of Age on LBQ score differed between qualification groups. The difference between UEFA A and UEFA B was small, b = −0.02, t(406) = −2.03, p = 0.043, Cohen’s d = −0.03, 95% CI [−0.05, 0.00]. In the UEFA A group, higher Age was significantly associated with lower LBQ score (b = −0.04, t = −3.90, p < 0.001), whereas in the UEFA B group, this relationship was not statistically significant (b = −0.01, t = −1.39, p = 0.166). The difference between UEFA A and UEFA C GRASSROOTS C was also small, b = −0.03, t(406) = −2.59, p = 0.010, Cohen’s d = −0.03, 95% CI [−0.06, −0.01]. In the UEFA C GRASSROOTS C group, the relationship between Age and LBQ score was not significant (b = 0.00, t = −0.48, p = 0.635). No significant difference was found between UEFA B and UEFA C GRASSROOTS C, b = −0.01, t(406) = −0.63, p = 0.532, Cohen’s d = −0.01, 95% CI [−0.03, 0.02] (Tables 4, 5 and Figure 3).

To analyze the results, an analysis of variance (Fisher, 1921) was conducted, which showed that Coaching experience did not have a significant effect on LBQ score [F(1, 408) = 1.07, p = 0.301, η² = 0.00, η p 2  = 0.00]. In contrast, Motivation for Education: Earning Points had a significant effect [F(1, 408) = 11.58, p < 0.001, η² = 0.03, η p 2  = 0.03]. Furthermore, the analysis of interaction effects in the tested model revealed a significant interaction between Coaching experience and Motivation for Education: Earning Points in predicting LBQ score [F(1, 408) = 8.22, p = 0.004, η² = 0.02, η p 2  = 0.02] (Figure 4). The model explained approximately 5.06% of the variance in LBQ score (adjusted: 4.36%). Figure 1 shows that the effect of Coaching experience on LBQ score differed between the “Points-Motivated” and “Non-Points-Motivated” groups. The magnitude of this difference was small, with Cohen’s d = 0.04, 95% CI [0.01, 0.06]. In the Non-Points-Motivated group, Coaching experience was not significantly related to LBQ score [b = −0.01, t(408) = −1.03, p = 0.301]. In contrast, in the Points-Motivated group, higher Coaching experience was significantly associated with lower LBQ score [b = −0.04, t(408) = −4.02, p < 0.001] (Tables 6, 7).

To analyze the results, an analysis of variance (Fisher, 1921) was conducted, which revealed that Number of teams managed had a significant effect on LBQ score [F(1, 406) = 6.25, p = 0.013, η² = 0.01, η p 2  = 0.02]. Similarly, Education showed a significant effect [F(2, 406) = 5.72, p = 0.004, η² = 0.03, η p 2  = 0.03]. Furthermore, the analysis of interaction effects indicated that Number of teams managed and Education interacted significantly in predicting LBQ score [F(2, 406) = 3.05, p = 0.048, η² = 0.01, η p 2  = 0.01]. The tested model explained approximately 3.40% of the variance in LBQ score (adjusted: 2.21%). Post-hoc comparisons of the moderating effect of Education showed that in the Higher education group, the effect of Number of teams managed on LBQ score was significantly different from that in the Higher sport group [b = −0.29, t(406) = −2.02, p = 0.044], with a small effect size (d = −0.29, 95% CI [−0.57, −0.01]). In the Higher group, more teams managed was significantly associated with lower LBQ score (b = −0.27, t = −2.50, p = 0.013), whereas in the Higher sport group, this relationship was not statistically significant (b = 0.02, t = 0.19, p = 0.852). Similarly, the Higher group differed significantly from the Primary and Secondary group [b = −0.34, t(406) = −2.30, p = 0.022], also with a small effect size (d = −0.34, 95% CI [−0.63, −0.05]). Again, the relationship was significant in the Higher group indicated that more teams managed was significantly associated with lower LBQ score (b = −0.27, t = −2.50, p = 0.013) but not significant in the Primary and Secondary group (b = 0.07, t = 0.70, p = 0.486). By contrast, the Higher sport and Primary and Secondary groups did not differ significantly (b = −0.05, t(406) = −0.39, p = 0.700, d = −0.05, 95% CI [−0.32, 0.22]). In both of these groups, the association between Number of teams managed and LBQ score was non-significant (Higher sport: b = 0.02, t = 0.19, p = 0.852; Primary and Secondary: b = 0.07, t = 0.70, p = 0.486). Figure 5 illustrates the moderating role of Education in the relationship between Number of teams managed and LBQ score, with shaded areas indicating 95% confidence intervals for the regression estimates (Tables 8, 9).

The final analysis of variance showed that the Number of Teams Managed variable did not significantly influence the explanation of the LBQ score variable [F(1, 406) = 1.26, p = 0.263, η² = 0.00, η p 2  = 0.00], while the Coach Qualifications variable showed a significant influence [F(2, 406) = 4.35, p = 0.014, η² = 0.02, η p 2  = 0.02]. Analysis of the interaction effect in the tested model showed that the variables Number of Teams Managed and Coach Qualifications did not significantly interact with each other [F(2, 406) = 2.42, p = 0.090, η² = 0.01, η p 2  = 0.01]; however, it is worth noting that this effect was on the borderline of a statistical trend. Analysis of the R² coefficient and its adjusted value showed that the tested model explained approximately 2.41% (1.20% after adjustment) of the variance in the results of the dependent variable LBQ score. The strength of the influence of the tested variables was assessed in terms of the value of the Cohen’s d statistic (Cohen, 1988). Figure 5 and the analysis of comparisons of the intensity of the effect of the Number of teams managed variable on the LBQ score variable showed that in UEFA group A it was statistically different than in UEFA group B b = 0.32; t(406) = 2.20; p = 0.029. The effect of the strength of differences between these groups was weak, Cohen’s d coefficient was d = 0.32; 95% CI [0.03; 0.61]. In UEFA group A, the effect of the Number of teams managed variable was not significantly associated with the change in the results of the LBQ score variable, however in UEFA group B an increase in the results of the Number of teams managed was associated with a significant decrease in the intensity of the LBQ score results, the results were, respectively, UEFA A b = 0.13; t = 1.16; p = 0.248 Vs UEFA b = −0.19; t = −2.00; p = 0.046. Further analysis of the effect of the number of managed teams on the LBQ result showed no significant differences between groups: UEFA B and UEFA C GRASSROOTS C [b = −0.11; t(406) = −0.78; p = 0.438] and UEFA A and UEFA C GRASSROOTS C [b = 0.21; t(406) = 1.49; p = 0.137] (Tables 10, 11).