The flow state (sometimes simply called “flow”) is a sensation in which one's own skills are adequate to face the challenges encountered (Jackson & Csikszentmihalyi, 2002). This sensation emerges during activities that are goal-directed and governed by rules providing clear feedback on performance (Nakamura & Csikszentmihalyi, 2014). This state of concentration happens to be so intense that it would decrease attention to irrelevant aspects of the environment as well as inducing a temporal distortion during the experience; that is, the feeling that time goes by faster than normal (Jackson & Csikszentmihalyi, 2002; Nakamura & Csikszentmihalyi, 2014).

Nakamura and Csikszentmihalyi (2014) proposed the existence of two conditions in order to enter Flow State. The first refers to the perception of a challenge, or opportunities for action, that require (without exceeding or underusing) existing skills. It is the feeling that one is facing challenges that have a level of difficulty appropriate to one's abilities. Thus, one has the feeling that he can control his own actions; that is, you can deal with the situation because you know how to respond to whatever comes next (Nakamura & Csikszentmihalyi, 2014).

The second condition is to have clear goals and to have immediate feedback on the progress that is being made. The Flow State is about a subjective experience of facing manageable challenges by addressing a series of goals, processing feedback about progress, and adjusting actions based on that feedback (Nakamura & Csikszentmihalyi, 2014).

Under these conditions, the experience unfolds seamlessly from moment to moment, and one gets in a subjective state that presents intense concentration and focus on what is being done, developing a fusion between action and consciousness (Nakamura & Csikszentmihalyi, 2014). This characteristic focuses on performing tasks automatically, allowing performing the activity in a more fluid way, thus avoiding the occurrence of intrusive thoughts (Calvo et al., 2008; Jackson, 1996).

According to Nakamura and Csikszentmihalyi (2014), during the flow state, the experience of the activity becomes intrinsically rewarding. The autotelic experience refers to the intrinsic satisfaction produced by the task. This is related to self-determination theory (Deci & Ryan, 2002), which postulates that a task is easier to perform when one feels satisfaction simply by doing it, without the need for any external reward (Calvo et al., 2008).

In a recent study (Himmelstein & Shapiro, 2017), it was stated that to achieve optimal performance, players should be able to adapt to their opponents, communicate adequately with their teammates (in cases where the game is not individual), and fully trust their abilities (Bányai et al., 2019). In addition, they must be able to develop themselves and their team (i.e., engage in individual skill practice and analyze team and self-performance) and set a varied set of goals (short-term and long-term). On the other hand, barriers or difficulties in execution are related to trust problems, inadequate coping strategies, past achievements and mistakes, lack of personal and team development (Himmelstein & Shapiro, 2017).

López-Torres et al. (2007) discovered that the flow state occurs in a situation of optimal performance in elite athletes. Particularly, they discovered that flow values ​​are higher in better sports performance experiences, while worse sports performance experiences present lower values ​​(López-Torres et al., 2007). However, there is still debate about whether research results in other sports could be generalized to Esports. In addition, this is reinforced by the lack of research related to Esports and the performance of its players.

The purpose of this study is to determine whether CSGO and LoL players are able to reach flow state while playing these games, and whether their performance improves or not depending on it. Finally, to assess if players manage to experience a more satisfying subjective experience when they are in flow state.

Hypothesis 1a [H1a]: The Esport CSGO video game allows its players to enter in flow state.

Hypothesis 1b [H1b]: The Esport LoL video game allows its players to enter in flow state.

Hypothesis 2a [H2a]: The performance of CSGO participants who enter the flow state is higher compared to those who do not enter the flow state.

Hypothesis 2b [H2b]: The performance of LoL participants who enter the flow state is higher compared to those who do not enter the flow state.

Hypothesis 3a [H3a]: CSGO participants who enter the flow state report having more satisfying subjective experiences than participants who cannot access this state.

Hypothesis 3b [H3b]: LoL participants who enter the Flow State report having more satisfying subjective experiences than participants who cannot access that state.

The surveyed players for both Esports were all general public over 18 years of age, with an average age of 20.4 years (SD = 6.4) for CSGO and an average age of 22.5 years (SD = 3.5) for LoL. The majority of both samples were men (94.2 % for CSGO and 77.5 % for LoL) and the nationality of the participants was mostly Argentine (93.9 % for CSGO and 82.5 % for LoL), while the rest of the sample were from other Latin American.

It was observed that participants slept an average of 7.5 hours (SD = 2) for CSGO and 7.35 hours (SD = 0.7) for LoL the night before the survey was carried out. On the other hand, these participants reported playing daily an average of 4.8 hours a day (SD = 2.1) for CSGO and 3.3 hours (SD = 1.4) for LoL.

In order to measure the performance of each participant within the video game, the statistics thrown at the end of each game were used. A screenshot of the results of their last game was requested when completing the survey^[1]. In case the images were unreadable results were consulted by mail. This allowed to record the number of kills (i.e., the number of times the participant killed another player within a game), deaths (i.e., the number of times the participant died within the game) and assists (i.e., the number of times the participant collaborated in a kill by damaging the health of another player during a game). The performance of each player was determined based on the K / D ratio, which implies dividing the number of kills divided by the number of deaths that each participant had.

Additionally, for LoL, if the player used a support character in the team, his performance would be calculated by dividing the number of assists by the number of deaths. (A / D). In any other role, his performance was calculated by dividing the number of kills by the number of deaths (K / D) he had in the game.

In order to measure the Flow State, the Spanish adaptation (Calvo et al., 2008) of the Flow State Scale (Jackson & Eklund, 2002;Jackson y Marsh,(1996), was used. Its construction was based on the nine dimensions of flow: Balance between ability level and challenge, Merging of action and awareness, Clear goals, Clear direct feedback, Concentration on task, Sense of control, Loss of self-consciousness, Distorted sense of time, Autotelic experience (Jackson & Csikszentmihalyi, 2002; Jackson y Marsh, (1996). Each dimension was assessed by four questions. In other words, the scale is made up of 36 items in total. The format for responding consisted of a 10-point Likert-type scale, going from “1” (strongly disagree) to “10” (strongly agree). According to literature consulted, the selected cut-off score was 6.8, Jackson y Marsh, (1996), reported that the internal consistency in the nine subscales was α = .83 (ranging from .80 to .86) after having administered it to 394 athletes. Also, the Spanish adaptation (Calvo et al., 2008) showed an adequate internal consistency with an average Cronbach's alpha coefficient higher than .70 for all factors. Regarding goodness-of-fit indexes, both the Comparative Fit Index (CFI) and the Tucker-Lewis Index (TLI) had values higher than .90, and the root mean square error of approximation (RMSEA) had values lower than .08. These values showed to be acceptable according to literature (Hu & Bentler, 1999).

In addition, with the purpose of evaluating the emotional tone and cognitive activation of the participants, a section of the Optimal Experience Questionnaire (Mesurado, 2008) was used, which allows evaluating the emotion caused by carrying out the task (emotional tone), and the alert sense of competence that one has during the task (cognitive activation). Emotional tone and cognitive activation were evaluated with the 7-point semantic differential method through 14 items.

The data of this study was analyzed using the programming language R. To obtain the value of the flow state in each participant, and in order to test the authors first hypothesis, responses of the 9 subscales that make up the Flow State Scale were averaged. In the first instance, it was necessary to define a cut-off score on said scale, which was set at 6.8, following the criteria used by Kawabata and Evans (2016).

For subsequent analyses of mean comparisons, Student's t-tests were used as long as the assumptions of the t-test were fulfilled. These are: a) the normal distribution of the variable (measurable with the Shapiro-Wilk test); b) the homogeneity of variances (Levene's test).

A one-sample Wilcoxon signed rank test was performed to assess whether the Esports evaluated allowed their players to enter a Flow State (μ = 6.8), considering that this variable was not normally distributed for either CSGO (V = 25215; p < .001) or LoL (V = 581; p = .011) players.

Regarding the analysis of performance based on the flow state for CSGO, no evidence was found that the variable related to the performance of the participants does not satisfy the assumption of distributional normality for those who passed the cut-off score of the Flow State (W = .94; p = .06) nor for those who did not enter a Flow State (W = .92; p = .26). Likewise, no significant differences were found between the variances of the subjective experiences of the participants in both conditions (p = .13). Therefore, a two-sample t-test was carried out to compare performance means according to the condition of the Flow State of the participants to respond to the hypothesis (H2a). On the other hand, in the case of LoL it could not be proved that the variable relative to the performance of the participants satisfies the assumption of distributional normality for those who passed the cut-off score of the Flow State (W = .85; p = .001) or for those who did not enter a Flow State (W = .57; p < .001). As it was not possible to apply a t-test for comparison of means, a Wilcoxon test was performed to answer the research question (H2b).

For the analysis of the results about optimal experience, it was necessary to perform a comparison means test in order to respond to the hypothesis raised (H3a and H3b). First, it was analysed whether the variable related to the subjective experience of the participants fulfilled the assumptions of normality and homoscedasticity of the variances. Thus, after performing a Shapiro-Wilk test for CSGO participants, it was observed that data was not normally distributed, both for the participants who entered the flow state (W = .96, p < .001), and for those who did not enter the Flow State (W = .95, p < .001). Finally, significant differences were found between the variances of the subjective experiences of the participants in both conditions (p < .05). Therefore, it was decided to perform a non-parametric test of medians to answer this research question.

On the other hand, in order to answer the last hypothesis (H3b), it was necessary to perform a means comparison test. Thus, whether the data related to the subjective experience of the participants fulfilled the assumptions of normality and homoscedasticity of the variances was tested. After performing a Shapiro-Wilk test, it was observed that the distribution of the variable was normal, for both the participants who entered a flow state (W = .97, p = .56), and for those who did not (W = .91, p = .19). Since no significant differences were found between the variances of the subjective experiences of the participants in both conditions (p > .20), a t-test was carried out to compare the means of subjective experience according to the flow state condition of the participants.

As Table 1 evidence, the descriptive statistical analysis showed that 63.6 % (n = 180) for CSGO and 70 % (n = 28) for LoL of the players entered into flow state in the CSGO game prior to completing the questionnaire. Participants for CSGO who surpassed the cut-off score in the flow scale had a mean score of 7.05 (SD = 1.5), and those for LoL had a mean score of 7.19 (SD = 1.43). According to the one-sample t-test, there is evidence that players enter in flow state both while playing CSGO [t (310) = 3.44; p < .001] and LoL [t (39) = 1.72; p = .047].

Table 1.
Data of the participants who passed the state of flow and those who did not

Results show that players who surpassed the cut-off flow score in any of the subscales did not necessarily manage to do it in all of them. It is observed that the majority of the players scored high on subscales such as clear goals (75.5 %; 76.2 %)[2], concentration on the task (74.4 %; 73.2 %), autotelic experience (74.1 %; 73.5 %), sense of control (73.9 %; 72.4 %) and balance between ability level and challenge (73.6 %; 74.2 %). However, a low percentage of players managed to reach the cut-off score in subscales such as distorted sense of time (25.2 %; 19.2 %).

A correlation analysis was performed (see Figure 1 and figure 2) to test the correlation between the different subscales.

Although all flow subscales correlated positively with each other.

Figure 1.
Correlogram of the scores of the different flow subscales for CSGO

Notes:

All comparisons between flow subscales were significant after applying a Bonferroni p value adjustment (p < .001), except for comparison between flow 3 and 8 (p = .01).

Flow 1: Balance between ability level and challenge, Flow 2: Merging of action and awareness, Flow 3: Clear goals, Flow 4: Clear direct feedback, Flow 5: Concentration on task, Flow 6: Sense of control, Flow 7: Loss of self-consciousness, Flow 8: Distorted sense of time, Flow 9: Autotelic experience.

Figure 2.
Correlogram of the scores of the different flow subscales for LoL.

Notes:

All comparisons between flow subscales were significant after applying a Bonferroni p value adjustment, except for comparison between flow 3 and 8 (p = .01).

Flow 1: Balance between ability level and challenge, Flow 2: Merging of action and awareness, Flow 3: Clear goals, Flow 4: Clear direct feedback, Flow 5: Concentration on task, Flow 6: Sense of control, Flow 7: Loss of self-consciousness, Flow 8: Distorted sense of time, Flow 9: Autotelic experience.

For CSGO analyses, the strongest correlations were found between the subscales “Concentration on task” and “Sense of control" (r = .79; p < .001), “Balance between ability level and challenge” and “Clear direct feedback” (r = .76; p < .001), “Balance between ability level and challenge” and “Clear goals” (r = .7; p < .001), “Balance between ability level and challenge” and “Sense of control” (r = .71; p < .001), “Clear goals” and “Sense of control” (r = .75; p < .001), “Clear direct feedback” and “Sense of control” (r = .75; p < .001), and, finally, “Clear goals” and “Clear direct feedback” (r = .7; p < .001).

Also, for LoL the strongest correlations were found between the subscales “Balance between ability level and challenge” and “Clear direct feedback” (r = .85; p < .001), “Autotelic experience” and “Concentration on task” (r = .84 ; p < .001), “Clear goals” and “Clear direct feedback” (r = .82; p < .001), “Clear direct feedback” and “Sense of control” (r = .75; p <.001 ), “Autotelic experience” and “Sense of control” (r = .74; p < .001) and, finally, “Concentration on task” and “Sense of control” (r = .72; p < .001). On the other hand, it was observed that the subscale “Distorted sense of time” presented non-significant correlations with the rest of the scales for both games.

Cronbach's alpha coefficient was calculated for the total scale and for each subscale, in order to assess the internal consistency of the instrument for both CSGO and LoL players. The values obtained for CSGO players were: α = .94, 95%IC [.93, .95] for the total Flow State Scale, α = .82, 95%IC [.78, .85] for Balance between ability level and challenge, α = .84, 95%IC [.81, .87] for Merging of action and awareness, α = .80, 95%IC [76, .84] for Clear goals, α = .87, 95%IC [.84, .89] for Clear Direct feedback, α = .88, 95%IC [.86, .90] for Concentration on task, α = .85, 95%IC [.81, .87] for Sense of control, α = .79, 95%IC [.75, .83] for Loss of self-consciousness, α = .78, 95%IC [.73, .82] for Distorted sense of time, and α = .88, 95%IC [.85, .90] for Autotelic experience. On the other hand, the values obtained for LoL players were: α = .95, 95%IC [.92, .97] for the total Flow State Scale, α = .86, 95%IC [.77, .92] for Balance between ability level and challenge, α = .84, 95%IC [.74, .91] for Merging of action and awareness, α = .89, 95%IC [83, .94] for Clear goals, α = .92, 95%IC [.87, .95] for Clear Direct feedback, α = .91, 95%IC [.85, .95] for Concentration on task, α = .76, 95%IC [.61, .86] for Sense of control, α = .77, 95%IC [.63, .87] for Loss of self-consciousness, α = .81, 95%IC [.69, .89] for Distorted sense of time, and α = .88, 95%IC [.81, .93] for Autotelic experience.

A confirmatory factor analysis (CFA) was performed with the 36 items to test whether the factor structure of the Flow State Scale in CSGO and LoL playes is equivalent to the original structure. The nine latent variables corresponding to those suggested by the original authors were introduced into the model. The model for fit was determined by the Comparative Fit Index (CSGO CFI = .89; LoL CFI = .5), the Tucker-Lewis Index (TLI = .88; LoL TLI = .43) and the root mean square error of approximation (CSGO RMSEA = .07; LoL RMSEA = .23).

Significant differences were found, showing a significantly better performance [t (43) = 3.57; p <.001; Cohen’s d = 1.17] for those who entered a flow state (M = 1.69; SD = .77) compared to those who did not in CSGO (M = .87; SD = .51) (Figures 3 and Figure 4). Similarly, in LoL players, the performance of those who entered flow state was higher (Mdn = 3) than in those who did not enter into flow state (Mdn = .82; W = 233, p = .001). The Glass rank biserial correlation coefficient rg = .63 indicated a large effect size.

Figure 3.
Comparative boxplot of performance in participants who entered or did not enter the flow state for CSGO.

Note: *Significant differences were found (p < .001) between performance in players who entered in flow state (M = 1.69; SD = .77) and those who did not (M = .87; SD = .51).

Figure 4.
Comparative boxplot of performance in participants who entered or did not enter the flow state for LoL

Note: *Significant differences were found (p < .001) in performance of players that entered flow state (Mdn = 3) compared to those who did not (Mdn = .82).

As shown in Figure 5, it was observed that the CSGO players’ subjective experience was significantly more satisfying [t (253) = 7.5; p < .001; Cohen’s d = .97] for those who enter into flow state (M = 73.09; SD = 9.67) compared to those who did not enter into flow state (M = 62.72; SD = 12.155). This could be observed in LoL players too (Figure 6), in whom their subjective experience was significantly more satisfying [t (38) = 6.69; p < .001; Cohen's d = 2.31] for those who enter into flow state (M = 75.79; SD = 6.88) compared to those who did not enter into flow state (M = 56.83; SD = 10.79).

Figure 5.
Subjective experience score for participants who had or had not passed the flow state cut off score for CSGO

Note: More satisfying subjective experiences (higher raw scores) in participants who entered a flow state were observed.

*Significant differences were found (p < .001).

Figure 6.
Subjective experience score for participants who had or had not passed the flow state cut-off score for LoL

Note: More satisfying subjective experiences (higher raw scores) were observed in participants who entered in a flow state.

*Significant differences were found (p < .001).

Currently, there are indicators in CSGO such as the ADR (i.e. average damage per round) and the HVTL index, which can be observed in external web pages. Similarly, in LoL, there are indicators of damage to champions, structure damage, gold earned and objectives achieved (the latter essential for players in the role of "jungle"). In future research, it would be relevant to have these performance indices to ensure that players actually performed better within the game.

Also, some of the variables that can affect the performance of players in Esports are the internet connection quality and speed, and the computer or device in which the game is played. In this study, data collection was limited to remote methods, leading to variables that remained uncontrolled. Although robust results were found as predicted, future research should ensure that all players have the same connectivity and device.

The main conclusions that can be drawn from this research are that Esports players experience flow episodes in the same way as in other sports modalities. The dimensions most experienced by players are autotelic experience, clear goals, balance between ability level and challenge, and concentration on the task; while there is a marked contrast with the dimension referred to the distortion of the sense of time. The results obtained allow us to report that the fact of entering in a flow state not only implies performing in the best possible way, but also that the experience becomes subjectively more satisfying accompanied by positive emotions and an optimal level of cognitive activation for the development of the task. In this way, players do not get bored or fatigued in their activity, achieving an ideal performance at every moment of the competition.