The effect of expertise, training and neurostimulation on sensory-motor skill in esports

Recently, increased attention has been directed to the brain to better understand how motor skill expertise develops. One promising technique purported to accelerate motor skill improvement is transcranial direct current stimulation (tDCS). While simple fine motor tasks involving the hands and fingers are most frequently used to investigate the role of tDCS on motor skill learning, less work has examined the role of tDCS on complex sensori-motor tasks applicable to occupational, sport, and daily living activities. Esports require a high degree of sensori-motor control and have become one of the most popular forms of digital entertainment worldwide. Currently, no research has quantified the development of motor skill expertise in esports or whether tDCS can enhance skill improvement. The current study aimed to first differentiate the sensorimotor performance of a key gameplay skill among esports players of different skill levels. Secondly, we quantified the training effect on performance. Finally, we investigated the effect of tDCS on performance improvements. We hypothesized that esport players would perform superiorly compared to novice gamers, that all groups would be able to improve their performance through training, and that tDCS would enhance training induced performance improvements. We found that performance on a single fundamental esport skill can differentiate expertise among novice and skilled players, that training can significantly improve performance among all expertise levels and that tDCS preferentially accelerates the performance improvements of novice players. The implications of this work, specifically regarding the temporal application of tDCS during complex motor skill learning and rehabilitation, are discussed.


Introduction
Sensory-motor control is one of the largest topics in the field of neurophysiology and beyond, and its diverse application to healthy aging, neuro-motor rehabilitation, and sport performance justifies the breadth and depth of the attention it receives. Motor skill acquisition is typically achieved with prolonged training, and the resulting performance gains typically comprise an improved speed-accuracy relationship and/or a reduction in performance variability (Shmuelof et al., 2012). Initially, experience with a motor skill is accrued over one or more practice/training periods (Dayan & Cohen, 2011). The performance improvements that accrue over these shorter time periods, such as within a single training session or day, are typically referred to as online learning. Over time, such as over several hours, days or training sessions, motor memories may transition to a consolidation phase (Stickgold, 2005). Predominantly, research has quantified the difference in performance between novice and experts on a given motor task as well as the acquisition of motor skill over time within individuals. However, more work is now looking to the underpinning neural correlates and mechanisms to further our understanding of motor skill acquisition (Espenhahn et al., 2020;Koch et al., 2020;Krakauer & Mazzoni, 2011;Park et al., 2015;Yarrow et al., 2009).
Our knowledge of the mechanisms underlying motor skill acquisition originally stems from the work of Von Holst & Mittelstaedt (von Holst & Mittelstaedt, 1950), who hypothesized that during motor learning, a copy of the efferent signal produced in the central nervous system is compared with, and updated by, the reafference produced during subsequent movement. Over time, this 'efferent copy' is refined in such a way that motor output becomes less variable and more accurate with respect to the task goal Dayan et al., 2016). This efference copy has been incorporated into more current Bayesian models of sensory-motor control as 'prior' experiential information, which is integrated with sensory reafference 'likelihoods' to evaluate the success of motor performance (Fetsch et al., 2013;Wilke et al., 2013). Recently, research has begun to investigate the potential of accelerating the process in which the efference copy is refined, leading to faster motor skill acquisition.
One technique used to facilitate motor skill acquisition is transcranial Direct Current Stimulation (tDCS) (Davis, 2013). tDCS involves passing a small current (below 2-3 mA) over an area of the cortex to modulate the excitability of the underlying neurons (Angius et al., 2017a;Brunoni et al., 2012;Nitsche & Paulus, 2011). Typically, tDCS is delivered using a bipolar electrode montage (Brunoni et al., 2012). When providing anodal stimulation, the excitability of the underlying cortical neurons is increased. Alternatively, cathodal stimulation has shown to decrease the excitability of the underlying cortical neurons (Brunoni et al., 2012). The neuromodulatory effects from a brief application of tDCS (10-20min) can persist for over 1 h, and repeated stimulation can prolong and stabilise changes that last for weeks (Nitsche & Paulus, 2011). Specifically, when applying anodal tDCS over the motor cortex, researchers have found evidence that motor skill performance and acquisition improve more quickly than during practice alone for various tasks (Yamaguchi et al., 2020).
The primary motor cortex (M1) is a complex network of somatotopically organised neurons, responsible for initiating and controlling movements (Schieber, 2001). The M1 shows a high degree of plasticity and adaptation in response to motor learning and practice (Jensen et al., 2005;Lee et al., 2010). Specifically, motor skill training is known to induce persistent encoded neural activations within the cortex that cascade through M1 to facilitate the precise execution of difficult motor tasks (Adkins et al., 2006;Nielsen & Cohen, 2008). As such, M1 appears to be an ideal target for tDCS insofar as the stimulation can easily facilitate the neural cascade that travels through motor cortical neurons during movement tasks, leading to accelerated plasticity of the associated neural circuitry.
Due to the fact that the M1 cortical neurons controlling the hands and fingers are located bilaterally along the motor cortex, altering the excitability of these neurons with tDCS is much easier than, for example, leg M1 cortical neurons, which are located deeper within the central sulcus. Therefore, simple fine motor control tasks involving the use of the hands and fingers are often used to investigate the role of tDCS on motor skill learning (Buch et al., 2017;Jackson et al., 2019;Reis et al., 2009;Spampinato et al., 2019). However, less work has examined the role of tDCS on the motor learning of more complex sensori-motor tasks involving precise endpoint accuracy requirements of the arms, hands and fingers, which could be more applicable to occupational tasks, sports, and daily living activities. One unique area that is now attracting significant attention where complex motor skill is displayed using the arms and hands that has never been studied through a 'motor learning' lens is that of competitive computer gaming, otherwise known as esports.
Esports are video games played at a competitive, and often professional, level, and their popularity has exploded over the past decade. The growth of esports has also led to the emergence of new scientific research investigating the health and performance of esports competitors. Specifically, research into the motor skills displayed by esports players has gathered momentum (Campbell et al., 2018). Esports predominantly require precise motor control of the hands and arms to operate a controller (console-based games), or a mouse and keyboard (PC-based games), consequently making esports-related tasks ideally suited for studying motor learning and the effects of tDCS. One of the original and most prominent esports over the past 20 years has been the first person shooter (FPS) game, Counter-Strike: Global Offensive (CS: GO) (Rizani & Iida, 2018;Wagner, 2006, pp. 437-442). Within this game, a fundamental skill required for success is 'flicking', a motor skill that involves a precise hand and arm movement and timed click the mouse to quickly and accurately target and destroy an enemy in the field of view. To date, the performance of this skill between players of different expertise has never been quantified, nor has its rate of improvement within and between skilled players. Uncovering the performance differences between expertise levels and quantifying the performance changes associated with the motor learning of esports skills will inform and potentially alter the way esports players train and practice to more rapidly improve. Flicking is also an exemplar skill for studying the effects of tDCS on the motor learning of a complex movement task using the arms and hands and can be used to demonstrate the efficacy of neuro-modulatory techniques for enhancing esport performance.
Taken together, the purpose of the present study was three-fold. First, we aimed to differentiate the performance of a key gameplay skill, flicking, between gamers and non gamers as well as among esports gamers of different in-game skill levels. We hypothesized that gamers would show superior flick performance compared to non-gamers, and that gamers with a higher level of expertise, based on their overall ingame competitive ranking, would demonstrate better flick performance. Secondly, we sought to quantify the effect of training on flick skill learning and performance. Here, we hypothesized that all expertise groups would be able to improve their flick performance on subsequent days when compared to their baseline performance. Finally, we investigated the effect of tDCS on flick skill learning and performance. We hypothesized groups exposed to tDCS would enhance their flick performance across training days more than those who trained the flick skill without stimulation.

Testing & training environment
All testing and training sessions were conducted at the Esports Science Research Lab (ESRL) in Lero at the University of Limerick. 8 identical PC stations were setup to ensure consistency among the data collected from all participants. Each station consisted of a high-powered gaming PC (Intel i7-8700 CPU @3.20 GHz), Logitech Pro Wireless gaming mouse and G613 Carbon Gaming keyboard. Each station was also equipped with a GT Omega™ Gaming chair for participants to sit in. ASUS (NVIDIA G-Sync) monitors were used and each monitor was identically calibrated for color, brightness, contrast, and frame rate (144 Hz).

Participants
One hundred and forty-nine young healthy adult participants were recruited from the University of Limerick and Limerick Institute of technology student populations provided informed written consent prior to participating in an esports training study that required attendance at a similar time from Monday to Friday consecutively and again the following Monday. The experiment was approved by the University research ethics board in accordance with the Declaration of Helsinki. Baseline, Post and Retention tests were conducted on the first Monday, Friday and second Monday respectively. Training sessions were conducted on five consecutive days with the first training session occurring immediately after the Baseline test on the first Monday and the last training session occurring immediately prior to the Post test on the Friday. Participants were excluded if they dropped out of the study prior to completing the Post test or missed at least two of the training sessions on Tuesday, Wednesday and Thursday (15 participants were excluded in this way). For a schematic of the experimental protocol, please see Fig. 1.

Questionnaires
Upon entering the Esports Science Research Lab on Day 1, participants were presented with a questionnaire that captured information regarding their age, sex, the number of hours they reported gaming on average per week, the game they predominantly played and, for those that played the game Counter-Strike: Global Offensive (CS:GO), their current and highest competitive rank. Participants were excluded if they reported any diagnosed neurological or neuromuscular disorder (N = 1), if they were both adextrous and used their mouse with their left hand, or if they suffered from color-blindness.
After gathering this initial information, participants were categorized into one of three groups. Those who dedicated less than 10hrs per week to action video games (game genres including first-person shooter (FPS), third-person shooter (TPS) and massive online battle arena (MOBA)) and did not hold any CS:GO competitive ranking were assigned to the Non Gamer Expertise group. Seventeen classified Non Gamers (NGs) were excluded upon learning they had prior experience playing alternate FPS or MOBA games despite failing to report playing them currently. Those who reported spending more than 5 h per week specifically playing CS:GO and maintained a current competitive rank between Silver 1 and Gold Nova 3 were assigned to the Low Skill Gamer (LSG) Expertise group. Finally, those who reported spending more than 5 h per week specifically playing CS:GO and maintained a current competitive rank between Gold Nova Master and Global Elite were assigned to the High Skill Gamer (HSG) Expertise group. Those who maintained a rank at or above Gold Nova Master but reported playing CS:GO less than 5 h per week were excluded. Fig. 2 shows the distribution of participants across the 18 competitive CS:GO rankings in the Low Skill and High Skill Gamer Expertise groups.
Participants were asked to refrain from caffeine within the 4-6 h, the half-life of caffeine in the average human (Blanchard & Sawers, 1983), prior to attending each daily session and an experimenter recorded whether or not they had consumed any caffeine upon arriving to the lab each day.

Mood, sleep and mouse sensitivity metrics
Each day when participants arrived at the ESRL, they also completed a 32-item Brunel Mood State (BRUMs) questionnaire (Lane & Jarrett, 2005). The questionnaire comprises 32 different mood descriptors and participants were asked to indicate the extent to which each descriptor matched their current mood on a Likert scale from 1 to 4. The Pittsburgh Sleep Quality Index (PSQI) (Buysse et al., 1989) is a questionnaire that poses questions regarding a participant's sleep over the previous month.  All participants completed this questionnaire on Day 1 (First Monday) and then completed a similar questionnaire on Day 8 (Monday the following week) with the only difference being that questions were framed to inquire about sleep patterns over the course of the previous week (time during which participants were in the study). At the start of each training and testing day, participants were also as instructed to select their usual gameplay mouse DPI using the Logitech G gaming software (https://support.logi.com/hc/en-001/articles/3600252980 53-Logitech-Gaming-Software). While Low and High Skill Gamers were able to select a familiar DPI, those in the Non Gamer Expertise group, who had never calibrated their mouse DPI for gaming in a FPS game, were instructed to 'try out' different mouse cursor speeds by altering the DPI using a slider within the software and moving the mouse around to control the on-screen cursor. They were instructed to select a DPI that they felt gave them the most control over the movement of their cursor.

Baseline, Post and Retention testing
All participants performed the Baseline, Post and Retention tests using a bespoke CS:GO Flick Test Software. The software was designed and developed using the CS:GO architecture and game mechanics. During the test, participants were instructed to use their mouse to control their avatar, positioned in the middle of a shooting arena and always armed with the same common weapon used in CSGO, to shoot and eliminate each of 45 enemy targets as quickly and accurately as possible. During each trial (target presentation) participants began by centering their crosshair over a ball presented at one end of the shooting arena. Centering their crosshair over the ball caused it to change color and initiated a random timer lasting between 1 and 2 s, after which a target appeared in the participant's field of view. If participants moved their crosshair aim off of the ball before the timer ended, they were instructed on screen to move their cursor back over the ball and a new random timer began. Participants shot and destroyed targets using the left-click button on their mouse. Targets were destroyed after 2 shots to the chest or one to the head. Participants could fire multiple shots in succession by holding down the left click button (referred to as 'spraying'), however, the trajectory of each successive shot deviated from the participant's aim according to the in-game weapon mechanics. In total, 45 targets were presented along 5 different directions from initial aim (42 • left, 18 • left, 0 • , 18 • right and 42 • right) and at three different distances from the participant's avatar (600, 900, 1200 game units). Three targets were presented at each unique direction by distance combination and targets were presented randomly for each participant.

Training groups
As each participant was categorized into an Expertise group, they were also randomly allocated into one of four Training groups. Participants in each Training group differed by their exposure to a bespoke CS: GO Flick Training Software and transcranial Direct Current Stimulation (tDCS). Table 1 shows the total number of participants in each unique group as well as their average (±SD) age and the number of hours they reported gaming on average per week.
Participants trained using the CS:GO Flick Training Software for 5 consecutive days (Fig. 1). The training software was designed and developed using the CS:GO architecture and game mechanics. During training, participants were instructed to control their avatar, positioned in the middle of a shooting arena and armed with the same weapon used in the test software, to quickly and accurately destroy as many targets in 10 min as they could. All targets could only be destroyed by a shot to the head and every time a target was destroyed, a new target immediately appeared in a random location within the participant's 'on screen' field of view. Constraining the appearance of targets within the participant's field of view prevented differences in recorded metrics that would likely occur if some targets appeared off screen and were not immediately visible at the time of destroying the previous target.

Training STIM group
Participants in the Training STIM (STIM) Group wore a custom headset (HALO Neuroscience™) designed to deliver transcranial Direct Current Stimulation (tDCS) to underlying cortical neurons. The headset was carefully positioned on the head overlying the motor cortex and participants received 2.1 mA 34 of anodal stimulation for a duration of 20 min. The location of each participant's motor cortex was determined to be the frontal band located immediately anterior to Cz, which was measured by an experimenter to be equidistant from the Nasion and Inion skull landmarks (Plowman-Prine et al., 2008;Rossini et al., 1996). Participant's received neurostimulation immediately before each training session. See Fig. 3 for the position of the electrode montage.

Training SHAM group
Participants in the Training SHAM (SHAM) Group wore the same headset, which was positioned over their motor cortex, and received a sham tDCS for a duration of 20 min. The sham stimulation consisted of an anodal current that ramped from 0mA to 1mA and then back to 0 mA over two 30s intervals before remaining at 0 mA for the following 19 min. This stimulation protocol has been demonstrated previously to not affect the long-term excitability of the underlying cortical neurons (Gandiga et al., 2006). The location of each participant's motor cortex was determined to be the frontal band located immediately anterior to Cz, which was measured to be equidistant from the Nasion and Inion skull landmarks. Participant's received the sham neurostimulation immediately before each training session.

Training NOSTIM group
Participants in the Training NoSTIM (NoSTIM) Group wore the same headset, which was positioned over their motor cortex. However, they received no tDCS at all for a duration of 20 min immediately prior to each training session. The location of each participant's motor cortex was determined in the same way as for participants in the SHAM and STIM Training groups for positioning the headset.

Control group
Instead of practicing with the training software for 10 min during each training session, participants who were allocated to the Control group played tetris for 10 min. In order to mitigate any transfer effects from using the mouse, participants used the keyboard arrow keys to play and played with their left hand. Control group participants did not wear the tDCS headset prior to playing tetris but did complete a crossword puzzle for the 20 min preceding tetris training. It is important to note that crossword puzzle completion was replicated for all training groups as well to control the cognitive engagement of participants and mitigate differences in underlying cortical network activity between training groups during the 20-min tDCS STIM, SHAM, and NoSTIM protocols prior to training (Giacobbe et al., 2013;Wokke et al., 2015).

Testing software metrics
A custom built LabVIEW program (National Instruments: LabVIEW 2013) was designed to calculate the Time on Target (ToT), Time to Shoot (TTS), Time to Destroy (TTD), and Ammo to Destroy (ATD) metrics for each target presented during the Baseline, Post and Retention tests. ToT, TTS and TTD were calculated as the time between the presentation of a target and the first overlap of the participant's crosshair on the target, the participant's first shot attempt and the destruction of the target by the participant respectively. ATD represents the number of shots (total ammunition) required to destroy a given target. All metrics were averaged across all targets for each of a participant's Baseline, Post and Retention tests.

Training software metrics
Another custom built LabVIEW program was built to calculate the total Number of Targets Destroyed (NTD), the Max Kill Streak (MKS), the Time to Destroy all targets (TTDall), centred targets appearing between 15 • left and 15 • right of initial aim (TTDcentre), targets appearing between 15 and 90 (left boundary of on screen field of view) degrees left (TTDleft), and targets appearing between 15 and 90 (right boundary of on screen field of view) degrees right (TTDright) of initial aim, as well as Ammo-to-Destroy all (ATDall), centre (ATDcentre), left (ATDleft) and right (ATDright) targets presented during completion of each 10 min training. NTD and MKS metrics were determined for each training session for each participant. TTDall, TTDcentre, TTDleft and TTDright averages were calculated for each training session for each participant and then individual values were excluded if they fell beyond ±3 SD of the mean. The same process was conducted for all ATD variables. After individual data were excluded (less than 2% of total values for any metric), averages were recalculated for statistical analyses.

Demographic, mood and sleep data
All statistical analyses were conducted using SPSS v25 statistical software. Data normality was verified by conducting Shapiro-Wilk tests and investigating Q-Q plots and histograms. Any metrics where data residuals were not normally distributed, data lying beyond 1.5 times the interquartile range were removed and data normality was again verified prior to conducting parametric analyses. Chi-Squared Tests (χ2) were used to test for differences in caffeine consumption between training groups across study days. Two-way ANOVAs were conducted to test for differences in the number of reported gaming hours per week, the combined DPI and in-game mouse sensitivity (D-Sens) used by participants during testing and training sessions, and overall week averages of BRUMs mood and PSQI sleep scores among Expertise and Training groups. Sidak adjustments were used to correct alpha levels for the multiple ANOVAs conducted on BRUMs and PSQI variables.

Performance data
To test whether Baseline performance differed among Gamer groups (Hypothesis 1), one-way ANOVAs were conducted on Baseline test data variables ToT, TTS, TTD, and ATD with post hoc Sidak corrections for multiple comparisons. Two-way ANOVAs [Expertise (NGs -LSGs -HSGs) x Training (NoSTIM -SHAM)] were also conducted on Day 1 training data variables NTD, MKS, TTDall, and ATDall, with post hoc Sidak corrections for multiple comparisons. Where a main effect of Expertise indicated differences among Expertise groups for TTDall or ATDall variables, further two-way ANOVAs were conducted on TTDcentre, TTDleft, TTDright, and ATDcentre, ATDleft, and ATDright variables respectively.
To test whether training improved performance across days for each of Non Gamers, Low Skill Gamers, and High Skill Gamers (Hypothesis 2), we conducted two-way ANOVAs specifically on TTD and ATD performance variables [Test data: Testing Session (Baseline -Post -Retention) x Training Group (Training NOSTIM and Training SHAM only)] [Training data: Training Session (Day1 -Day5) x Training Group (Training NOSTIM and Training SHAM only)] with Baseline-Post and Baseline-Retention a priori comparisons for test data and T1-T2, T1-T3, T1-T4 and T1-T5 a priori comparisons for training data.
Finally 2-way ANCOVAs were conducted on TTDall and ATDall performance metrics for each of Non Gamers, Low Skill Gamers, and High Skill Gamers (Test data: Training Group x Testing Session) (Training data: Training Group x Training Day) to test whether participants who received tDCS improved in their performance more than those in the Control group and those who received sham tDCS or no tDCS (Hypothesis 3). Baseline and T1 scores were used as covariates in the 2-way ANCOVAs conducted on Test and Training metrics respectively. Where a main effect of Training group was observed for TTDall or ATDall variables for any Gamer group, further two-way ANCOVAs were conducted on TTDcentre, TTDleft, TTDright, and ATDcentre, ATDleft, and ATDright variables respectively.
Analyses of the average mood scores captured by the Brunel Mood State (BRUMs) questionnaire revealed a significant main effect of Training group only for Confusion (F (3,116) = 3.016, p = 0.033, η 2 = 0.079). However, post hoc analysis showed no significant differences among the different training groups after adjusting alpha levels for multiple comparisons. No interaction or main effects were observed for any of the other moods captured by the BRUMs questionnaire (Anger, Tension, Depression, Vigour, Fatigue, Happiness, Calmness).
Upon analyzing data from the Pittsburgh Sleep Quality Index (PSQI) questionnaire, we found a significant Training Group by Day interaction effect for Daytime Dysfunction (F (3,105) = 3.525, p = 0.018, η 2 = 0.091).
Post hoc comparisons revealed that the Daytime dysfunction score was 0.320 higher on Day 8 compared to Day 1 only for Control participants pooled across all three Expertise levels (p = 0.032). Main effects of Day for Sleep Disturbance (F (1,105) = 6.269, p = 0.014, η 2 = 0.056), Subjective Sleep Quality (F (1,105) = 188.756, p < 0.001, η 2 = 0.643) and Global Sleep Score (F (1,105) = 22.883, p < 0.001, η 2 = 0.179) indicated average scores were significantly higher on day 1 compared to day 8 of the study. No significant interaction or main effects were found for Sleep medication, Sleep Latency, Sleep Duration or Sleep Efficiency.
χ2 analysis revealed that only on day 4 did participants in the control group consume significantly more caffeine than those in the training groups (Fig. 4). However, Cramer's V (0.247) showed that this effect was weak. No significant associations were found between Expertise and Day.

Non gamers
When examining test data, a main effect of test day was found for comparisons showed T2, T3, and T4 scores were all significantly less than T1 scores for left and right targets (all p < 0.050).

Low Skill gamers
Test data analyses revealed a significant interaction effect for TTD

High Skill Gamers
Test data analyses showed a significant main effect of Test Day for

Non gamers
Analyses of Test TTD data revealed a significant main effect of showed that participants in the STIM group had greater NTDs compared to those in either the SHAM (p = 0.010) or NoSTIM (p = 0.110) groups. No significant post hoc effects were found between groups for ATDall. Participants in the STIM group also displayed faster TTDs compared to SHAM (p = 0.020) and NoSTIM (p = 0.111) groups (Fig. 8D). Further comparisons across Training Days revealed that those in the STIM group significantly improved their performance compared to SHAM and NoSTIM groups by Training Day 3 (T3). Analyses of TTDs for left, centre and right targets showed a significant main effect of Training group for TTDleft (F (2,18) = 3.607, p = 0.048, η 2 = 0.286) and TTDright (F (2,18) = 5.422, p = 0.014, η 2 = 0.376), but not TTDcentre (F (2,18) = 2.757, p = 0.090, η 2 = 0.234). Post hoc comparisons revealed that STIM TTDs for left and right targets were faster than SHAM (left; p = 0.050, right; p = 0.039) and NoSTIM (left; p = 0.287, right; p = 0.038) TTDs across training days (Fig. 10). Training group comparisons across training days showed that the additional TTD improvement by the STIM group for left and right targets was evident by Training Day 3 (T3).

Low Skill gamers
Analyses of Test data TTDs revealed a significant Training group by Test Day interaction effect (F (3,554) = 3.462, p = 0.016, η 2 = 0.018) with post hoc comparisons showing a significant difference between Control and NoSTIM (p = 0.004), SHAM and NoSTIM (p < 0.001), and SHAM and STIM (p = 0.046) TTDs during the Retention test when controlling for Baseline TTD scores. No significant interaction or main effects were observed for ATD. Analyses of Training data revealed no main or interaction effects for any of NTD, TTKall (Fig. 9E) or ATDall.

High Skill Gamers
Analyses of Test data TTD and ATD revealed no significant main or interaction effects. Analyses of Training data revealed no main effect or interaction effects for any of NTD, TTKall (Fig. 9F) or ATDall.

Discussion
The current study aimed firstly, to quantify the flicking performance, a common first-person shooter game skill, involving precise finger, hand and arm motor control, among Non video game players and CSGO Gamers of Low, and High expertise. Overall, HSGs were found to perform superiorly compared to LSGs, who were also found to perform better than NGs. This finding addresses aim 1 and indicates that the motor skill of flicking can differentiate FPS expertise. Secondly, we set out to evaluate whether NGs, LSGs and HSGs could improve their flicking performance over the course of 5 days of training for 10 min per day. We found that significant improvements could be attained as early as the third day of training for all groups. Finally, to address our third aim, we examined the effect of transcranial Direct Current Stimulation (tDCS) on training improvements in flicking performance. We found that  and right (C) targets. *, ** and *** indicate significant differences at p < 0.05, p < 0.01 and p < 0.001 respectively.
training with tDCS was significantly more beneficial for improving performance compared to training alone specifically for our Non Gamer sample only. We discuss the implications of these findings within the context of sensory-motor skill based training in esports and the potential for expertise to moderate the effect of tDCS on motor learning.

Differentiating expertise in esports
In the game CS:GO, an individual's competitive ranking is typically the sole metric used to describe their expertise level. Currently, both amateur and elite players improve their rank and skills by simply dedicating more time to playing matches in the game (Campbell et al., 2018;Toth et al., 2019). However, research in psychology (Boot et al., 2010;Towne et al., 2016), sport (Landin et al., 1993;Yao et al., 2009), andcognition (McKendrick &Parasuraman, 2012) have all demonstrated that the deliberate training and application of individual skills to more complex tasks is a superior method for improving complex task performance compared to simply repeating execution of the complex task. This strategy of testing and training component skills also has the benefit of identifying specific areas of weakness, which can then facilitate faster overall improvement through the deliberate practice of those areas deemed to be inferior. The importance of motor skill training has only recently become recognized in competitive video gaming and companies have only started to develop similar software tools to test individual gaming skills. In this study, we provide the first objective evidence that an individual gaming skill, as evaluated using a customized software with the same mechanics experienced in the game itself, can be used to differentiate gamers of different overall expertise levels (Figs. 5 and 6).

Skill training among individuals of varying expertise
Meaningful performance improvements were found in all Expertise groups when participants trained for only 10 min per day. Moreover, in line with previous literature, we saw that the magnitude of performance change over the course of the 5-day training protocol was much greater for Non Gamers than for LSGs and HSGs (McNeill et al., 2019). Previous work has highlighted that as skill is acquired, increasingly more time and/or effort is required to cause a comparable increase in performance (Côté et al., 2007;Newell & Rosenbloom, 1981). In the current Study, we observed an 8.21% improvement in performance in NGs, an 8.90% improvement in performance in LSGs, and a 6.09% improvement in performance in HSGs TTKs. Moreover, we also found that as expertise increased, the level of variability between participants within a given day's performance decreased, as the average TTK performance variance between participants on a given day in the NG group was 3.32 and 7.68 times the variance observed between LSG and HSG participants respectively. This pattern of variance among expertise levels corroborates existing skill acquisition literature demonstrating that performance

Fig. 8. Baseline, Post and Retention test TTD and ATD metrics (mean ± SE) pooled across SHAM and NoSTIM Non Gamer (A and D), Low Skill Gamer (B and E) and
High Skill Gamer (C and F) participants. *, ** and *** indicate significant differences from baseline at p < 0.05, p < 0.01 and p < 0.001 respectively. among novices is much more variable than among more skilled individuals (McNeill et al., 2019;Phillips et al., 2012). Previous work has highlighted the importance of uncovering the many physical and cognitive skills required for successful gaming performance  and this study is the first to identify and quantify the improvements to be gained by individuals with varying expertise on an important complex sensory-motor skill required for FPS esports. Overall, we found that when both gamer and non gamer participants trained using our software, they were able to rapidly improve their performance on a skill that, as we have highlighted above, differentiates performance among gamers of different expertise levels.

Transcranial direct current stimulation effects on performance
Our study found that NGs who received tDCS over their motor cortex prior to training on the flick task improved their performance on the task over the course of 5 days significantly more than NGs who trained following no such stimulus (SHAM & NoSTIM groups). Previous work has highlighted that tDCS can improve complex motor skill performance across a number of tasks, including music (Rosen et al., 2016), visuo-motor tasks (Reis et al., 2009;Jackson et al., 2019;Waters-Metenier et al., 2014), balance tasks (Kaminski et al., 2016), golf putting (Zhu et al., 2015), and darts throwing (Mizuguchi et al., 2018). However, the efficacy of tDCS for simple reaction time tasks or exercise/strength based tasks remains debated (Angius et al., 2017b;Machado et al., 2019). When we examined the effect of tDCS on training compared to SHAM and NoSTIM groups, we observed a significant effect of tDCS on training for left and right targets, but not centre targets (see Fig. 9). The fact that tDCS exerted an influence on training performance specifically for targets requiring a larger controlled movement (left and right targets) corroborates the assertion that tDCS may be better able to accelerate performance improvements for complex motor movements rather than simple reactions (Seidel & Ragert, 2019).
As research investigating the application of transcranial electric stimulation (which includes tDCS) for cognitive and motor learning has flourished over the past decade, so has the debate over the efficacy of  these techniques for improving cognitive and motor performance. Recently, a paper by Filmer, Mattingley & Dux (Filmer et al., 2019) has identified a number of criteria that may explain the variance in findings among the existing literature and provides recommendations concerning the varied methods of tDCS implementation, including the level of current used, electrode montage, use of SHAM and Control groups and how potential confounding variables such as caffeine intake are controlled for (McIntire et al., 2017). In line with the recommendations provided by Filmer and colleagues, we utilized an anodal tDCS electrode montage (Fig. 3) that delivered 2.1 mA of current to the scalp. Based on the work of Huang and colleagues (Huang et al., 2017), we can be confident that a stimulus of 2.1 mA is likely creating greater than a 0.8V/m voltage gradient over the cortex of our participants. Given that 0.42 and 0.68V/m voltage gradients can induce changes in neuronal excitability (Krause et al., 2017;Vöröslakos et al., 2018), we are confident that our stimulation protocol is altering the motor cortical excitability of participants who received the stimulus.
As also recommended by Filmer and colleagues (Filmer et al., 2019), we compared our tDCS STIM group to both a group who received a SHAM stimulation and a group who trained while receiving no stimulation (NoSTIM). When we asked a subset of 45 participants who were in either the STIM or SHAM groups whether they thought they received a real or fake neurostimulation, we found that 68.75% and 71.43% of participants respectively thought they had received the real/actual neurostimulation. This suggests firstly, that we were able to successfully blind participants to the stimulation and secondly, that despite equally thinking they had the real stimulation, participants that actually received tDCS (STIM) improved their performance more than both the SHAM and NoSTIM training groups. Finally, we also controlled for and recorded the sleep, mood, and the level of caffeine intake of participants during the study. In finding no differences between our Expertise or Training groups, we are confident that performance differences between our groups were not due to differences in sleep, mood or caffeine intake between Training groups during the course of the study. Overall, we demonstrate with this work that when sufficiently addressing the confounding influence of criteria outlined by Filmer and colleagues (Filmer et al., 2019), tDCS can be shown to be efficacious for improving performance on complex sensory-motor tasks.
In our study, we uniquely found that tDCS enhances sensory-motor performance of the investigated flick task only among participants in the Non Gamer Expertise group. To date, no work has explicitly investigated the differential effect of tDCS among individuals with different experience or skill level. Although, there is some evidence indirectly supporting the claim that tDCS preferentially accelerates performance improvements for novel motor skills, of which the flick skill is for Non Gamer participants in this study (Mizuguchi et al., 2018;Reis et al., 2009;Rosen et al., 2016). One explanation for why novices may more greatly benefit from tDCS is that they have a greater capacity for establishing the neural circuitry associated with movement automaticity. Motor learning literature shows that as skill is acquired, plastic changes occur within the neural circuitry (Chang, 2014;Gellner et al., 2020) that may also be accompanied by synaptic conductivity and inhibitory mechanisms at the spinal level (Ruffino et al., 2017). The more novel a skill, the greater capacity for neurostimulation to facilitate the plastic changes that occur during motor learning (Chang, 2014). Based on our results, we hypothesize that to be able to observe an effect of neurostimulation on training in Low and High Skill Gamer groups, longer training durations may be required. Moreover, we encourage future work to examine the dose response to tDCS among performers of different expertise across other motor skills.
An alternative explanation for the observed facilitatory effect of tDCS on training improvements specifically in Non Gamers may be their use of a significantly higher mouse D-Sens. At a higher D-Sens, a smaller physical hand movement is required to move a more sensitive on-screen cursor through a given pixel distance. As such, one might argue that a higher D-Sens places a greater demand on fine motor control during the flick task. This increased demand for fine motor control among NGs may lead to the greater observed effect of tDCS on training improvements. When studying the potential for tDCS as a tool to enhance motor learning, many of the tasks chosen require fine motor hand movements (Reis et al., 2009) , (Spampinato et al., 2019) , (Jackson et al., 2019) (Mizuguchi et al., 2018) , (Pixa & Pollok, 2018). Due to the fact that esports' overtly requires fine motor skills of the hands with either a controller, or mouse and keyboard, esports-related skills may not only provide a unique test bed to examine the effects of tDCS on sensory-motor skill performance, but tDCS may become a useful tool for those seeking to accelerate the training improvements of a number of motor tasks required for successful esports performance.

Implications
We believe the findings from this research may be applied to a number of research areas and real-world situations. For example, when learning a new task, we faithfully replicate that performance can vary considerably in novice performers (Mizuguchi et al., 2018). Awareness of this knowledge may provide encouragement and facilitate resilience in novice performers who may be prone to frustration and decreased motivation due to their initially high performance variability when learning a new skill. This knowledge also can be applied into existing self-regulated learning models of deliberate practice (Tedesqui & Young, 2015). Secondly, due to the fact that more time and effort are required to manifest appreciable increases in performance as expertise increases, cognitive strategies such as mental practice (MP)  and action observation (AO)  can be implemented as they have been shown to be effective for further enhancing performance. The use of motor simulation strategies like MP and AO may be especially relevant in esports, where the effect of these cognitive strategies on performance has yet to be investigated with any rigour. Thirdly, our work has corroborated previous research in showing that tDCS is highly efficacious for skill acquisition among novice performers (Mizuguchi et al., 2018). When considering the learning of sensory-motor skill among different age groups, it is important to consider how children, who are novices for most tasks, acquire skill and how neurostimulation may help or hinder learning in the developing brain. In a review on the effects of tDCS in children, Palm and colleagues (Palm et al., 2016) conclude that although tDCS seems to be safe in pediatric populations, more work is required to confirm these encouraging preliminary findings. Finally, given the evidence that tDCS may be especially beneficial during the learning of novel tasks, patients who have had sensory or cognitive lesions may most greatly benefit from tDCS at the start of their rehabilitation process, when initially re-learning complex movements that were once automatic. For example, recent research examining motor recovery in stroke patients has suggested that there is an early optimal window of time soon after a stroke when the brain is in a very plastic state (Wahl & Schwab, 2014). Given that tDCS affects motor learning through neuroplastic mechanisms (Chang, 2014), it may be that tDCS supplementation during early rehabilitation could accelerate patient mobility and/or recovery. Overall, tDCS is a promising technique in both performance and clinical contexts and we encourage future research to continue to explore its merit.

Limitations
To allow for a fully informed interpretation of our findings and further place their relevance within the research areas of motor learning and neurostimulation, we do highlight certain limitations. Firstly, although our test data metrics were able to differentiate expertise, the effect of training was smaller than when the same metrics were assessed using the Training data. This may be due to the increased variability associated with sampling a lower number of targets/trials, the increased spatial predictability of targets due to the cursor re-centreing in the test software environment and/or the consistency with which targets appeared along a given angle at different distances. The test and training software also differed in their constraints for destroying targets; i.e., two shots to the larger thorax area or a single shot to the head could destroy a target in the Flick Test Software, whereas in the Flick Training Software, targets could only be destroyed by shots to the head. This may have produced strategy differences between Baseline and Post tests, where a participant who employed a strategy to 'spray' targets at Baseline (due to an increased probability of destroying the target by shots to the chest or head) may have altered their strategy over the course of their 5 days of interacting with the Training software (where a spray strategy was likely not optimal due to the constraint on only destroying targets with a single hit to the head). Our data suggest this may be the case. When examining the training data, participants did not significantly alter the amount of ammo used to destroy targets but did improve in the speed at which they destroyed targets; a clear performance improvement. However, from Baseline to Post tests, participants used significantly less ammo to destroy targets in addition to destroying them more quickly (Fig. 8). Therefore, despite our intent to constrain training in the hope of facilitating improvements in Post and Retention tests compared to Baseline, the differences between the test and training software may have altered strategies between Baseline, and the subsequent Post and Retention tests, making more difficult the observation of performance improvements and tDCS effects among test software data.

Conclusion
Overall, this study demonstrates firstly that performance on a single FPS gameplay skill, flicking, can differentiate expertise between gamers and non gamers, and among gamers of different overall in-game expertise. Secondly, performance improvements among participants of all skill levels were observed after only 3 days of training 10 min per day. Finally, and most interestingly, we showed that transcranial Direct Current Stimulation (tDCS) can accelerate motor performance improvements specifically in novice participants, and that the effect of tDCS was confined to stimuli requiring more complex sensory-motor actions. This work significantly contributes to a growing body of research investigating the effects of neurostimulation on sensory-motor performance and demonstrates esports to be an exemplar medium within which to study motor learning and the effects of neurostimulation on motor skill development.

Credit author statement
AT, MC, CC and AM-Conceived and designed the study, NR-Collected data, processed data, recruited participants, AT and MC-ran the statistical analyses, AM and CC-software and server support, AT and MC-drafted the manuscript, AT and MC-interpreted results, final draft of the manuscript.

Declaration of competing interests
Two of the authors (CC & AM) have financial competing interests: Funding: Research support (including salaries, equipment, supplies, and other expenses) by Logitech Switzerland that may gain or lose financially through this publication. Additionally the same two authors (CC & AM) are Logitech employees and have some stocks or shares in companies that may gain or lose financially through publication; consultation fees or other forms of remuneration (including reimbursements for attending symposia) from organizations that may gain or lose financially.