Factors that affect handball execution in Australian Football

ABSTRACT A handball is one of two technical skills used to dispose of the ball in Australian Football. Previous research has only considered handball effectiveness in the analysis of team performance and there is a need to understand whether there are other more important characteristics of handball execution that explain effectiveness (i.e., performance). 1342 handballs from Australian Football League matches were analysed. Ten characteristic variables were created that represent the context and execution of each handball included in the analysis. A mixed effects generalised linear model was used to evaluate the effect of the (ten) characteristics on handball outcome. Three out of ten characteristics were associated with handball outcome, and these related to the posture of the handballer and the type of defensive pressure applied to the handballer and the receiver of the handball. These findings explain both how to increase handball effectiveness and how to reduce the handball effectiveness of an opposition team. Given the important role of handballing in passing sequences and maintaining ball possession, the practical application of these findings could enhance overall team performance.


Introduction
Australian Football (AF) is a field-based, contact invasion team sport played between two teams of 18 players with the intent to score more points than the opposition (Gray & Jenkins, 2010).Players share possession by passing the ball, which in AF is referred to as a "disposal".The two legal disposals in AF are a handball (whereby the ball is held in an open palm and the other hand is clenched and used to punch the ball), or a kick.These two technical skills allow teammates to move the ball close to their scoring end to enable shots at goal (Kempton et al., 2015;Sullivan et al., 2014).Six points (goal) are achieved by kicking the ball through the goal posts (without it being touched by any other player), or a single point can be scored by kicking the ball through the point posts to the left and right of the goal posts, or if the ball is touched by any player before crossing the goal line.Improving the execution of the technical skills associated with disposals will increase the likelihood of a team maintaining possession and therefore improve their potential to score.
Handballs comprised 41% of total disposals in the 2019 AFL season (Australian Football League, 2019).The percent of handballs from total disposals for seasons 2015 to 2018 ranged from 43% to 45%, whilst seasons subsequent to 2019 have ranged from 40% to 41% (2020-2022;AFL Tables, 2023).Handball efficiency in 2019 was shown to be 82% (i.e., percentage of handballs that were received by a teammate) compared to 64% for kicking (Australian Football League, 2019).Furthermore, handballs (relative to opposition) have been found to have a significant association with match outcome (Robertson, Back et al., 2016).Historically, within the professional league (Australian Football League (AFL)), the official match statistics collector (Champion Data) and the media, typically report a variety of statistics that relate to kicking performance, such as kick type, location, distance, direction, type of defensive pressure and outcome (Stats Glossary: Every Stat Explained, 2022;Vella et al., 2022).However, for handballs, reports generally include only the outcome (i.e., efficiency) or errors (a "clanger handball").Compared to kicking performance, less is known about handball performance and there is no understanding of the contextual factors that affect handball effectiveness.If teams had a better understanding of the factors that affect handball performance, they may be able to increase handball efficiency and possibly their overall performance in a match.Vella et al. (2022) conducted a systematic review assessing the physical, technical, and tactical analyses in the AFL.Within this review, the technical requirements of the AFL were stated, highlighting that many aspects of technical performance have been thoroughly investigated in previous research.Whilst there are a small number of papers to include handball count and/or efficiency in their research, there is a lack of research to analyse the execution of this skill, despite handballs representing 40-50% of all disposals (Australian Football League Statistics, 2019).There is limited research with a focus on handball performance in AF.Parrington, Ball and MacMahon (2013) sought to "profile handball performance" by analysing several factors including outcome, technical, decision-making, and gameenvironment factors.They reported that short handballs to the front, rather than across the body were more efficient and similarly, when players were presented with low levels of pressure and had passing options available, they were more successful (Parrington et al., 2013).Handball efficiency was higher when players were either in motion or if stationary using a "knees-bent stance".Handballs were most often executed in the midfield and received from a ball travelling in the air but were most effective in the defensive 50 and when the ball was received in-flight or as an "easy-receive".While this study is useful, it does have some limitations.No statistical testing was used to support their comparisons, instead the authors made qualitative comparisons of descriptive statistics (Parrington et al., 2013).Furthermore, although the data represented multiples teams and matches, it was drawn only from the first quarter of matches, which introduces the potential for a small bias in the results.Consequently, the conclusions of this study may not be reliable, however they provide a valuable starting point for understanding handballs in AF.Ireland et al. (2019) examined the representativeness of training in an Australian Football League team and in doing so, provided some extra context around the execution of handballs.They contextualised time in possession, the distance handballs travelled, whether the handballer was static or dynamic, whether the receiver was stationary or on the move, pressure on the handballer and receiver, and how the handballer received the ball.This increased the understanding of how handballs occur in training and matches, however there was no measure of effectiveness provided so it is not understood what factors may contribute to more successful handball performance.Teune et al. (2021) reported an effectiveness of 83.2% (for total disposals), which is higher than previously reported during competition (Australian Football League, 2019), however, this may have been a result of no pressure being present in 55% of the skill involvements.They also noted that higher player density (team in possession and opposition players) was weakly related to more effective disposals.This is thought to be due to potentially lower player densities at the target player (Teune et al., 2021), however, this was not measured.The Parrington et al., (2013) and Ireland et al. (2019) studies provide useful frameworks for the analysis of handball performance, but there is also an opportunity to develop these frameworks further.Therefore, the aim of the current study was to determine whether the context and the execution of handballs, affects the outcome of handballs.The primary hypothesis is that neither the context nor execution characteristics of handballs will affect their outcome.

Material and methods
Ethical approval was granted from the relevant institution Human Research Ethics Committee.Handballs (n = 1342) from a series of AFL matches in the 2019 season were analysed.The sampling procedure was designed to represent a wide range of handball contexts and execution characteristics, to optimise the generalisability of the findings (Hughes & Bartlett, 2002).Each match contained handballs from a specific club that was a partner in this study, and their opponent.Thus, handballs were sampled from four matches for the club and their opposition.The four matches chosen were from the second half of the season and were selected to represent a variety of end-of-season ladder positions for the opposition team analysed (top four, top eight, bottom eight and bottom four).In addition, matches were selected that had varying score margins and outcomes (win/loss) of both quarters from within the match and the overall match outcome.The club involved won two of the matches and lost two of the matches and quarter scores ranged from winning or losing a quarter by less than 10 points to winning or losing a quarter by greater than 15 points.
Pre-coded information from the official statistics provider of the AFL (Champion Data, Melbourne, Australia) was used to identify when each handball occurred in matches.This data file was then linked to a corresponding video file showing a "broadcast" camera angle of the match.The subsequent data collection was conducted using performance analysis software (Sportscode.Version 11.3.0,Hudl, Australia).Using the Champion Data pre-coded information to identify each handball, handballs were then analysed using a framework related to the only other study of handballs (Parrington et al., 2013).Handball performance was considered in three phases: context, execution, and outcome.Handball context characteristics were used to describe the factors that could influence a handball (e.g., how the ball was gained).Handball execution characteristics were used to describe the factors that could influence disposing of the ball (e.g., pressure when disposing of the ball) and handball outcome characteristics were used to quantify the success of the handball (e.g., possession maintained or lost).See Supplementary Table S1 for a summary of the characteristics, their categories, and definitions.
After completing data collection, some levels of several characteristics (seven of nine) were redundant and were collapsed into the next most similar level.This decision was made to ensure the quality of analysis was maintained and to assist with the interpretation of results.For example, the levels of pressure such as closing, corralling, and chasing were collapsed to the category "Implied", consistent with Champion Data's definition.For our dependent characteristics "Handball -outcome", all levels were collapsed into either "complete" or "incomplete".A "complete" handball refers to a handball that is directly passed to its intended target and possession is maintained within the player's team, and this aligns with Champion Data's "handball efficiency" measure.
A total of 1,419 handballs were identified in the four matches.Seventy seven (77) handballs were excluded from the analysis due to an inadequate view in the video images or an umpire deeming that an attempted handball was an illegal throw.Therefore, data from 1342 handballs was used in the analysis.During data processing, additional metadata was added to each handball, including match, team, player, and quarter identifiers.
All data analysis was conducted using SPSS software (Version 29.Armonk, New York).A mixed effects Generalised Linear Model (GLM) was used to determine whether any handball characteristics were associated with handball outcome.A GLM was chosen for this analysis because of its ability to manage variables with different types of distributions.In addition, this type of model can evaluate factors that are presumed to have a consistent (i.e., fixed) effect on handball outcome (e.g., body posture) and those that might have an inconsistent (i.e., random) effect (e.g., field zone).Therefore, the ten handball characteristics listed in Table 3 were included as fixed effects, whereas, match (n = 4), team (n = 5), quarter (n = 4), field zone (n = 2) and player ID (n = 114) were included as random effects in the GLM.The link function was a binomial logit for handball outcome (in/complete).The general formula for the model was: Where y is the handball outcome, β and χ are the coefficient and observed value of ten fixed effects.There are five random effects γ and one residual error term ε.The model was trained on 100% of the data and its accuracy was assessed by simply comparing observed versus predicted handball outcomes.In a separate procedure, intra-rater reliability was assessed using two quarters of play, which equated to 8% of all handballs coded.Cohen's kappa and intra-class correlation coefficient results showed substantial (0.61-0.80) to almost perfect (0.81-1.00) agreement in line with Cohen (1960).

Results
Descriptive statistics are reported for the metadata characteristics (i.e., match, team, quarter, and field zone) in Supplementary Table S2.A total of 1,002 handballs were classified as complete and 340 as incomplete (see Table 1).The table can be used to identify the handball characteristics that occurred most and least frequently.Handball efficiency ranged from 16.0 to 95.8%, across different levels of the contextual and execution characteristics.
While the sampling procedure was designed to be representative of all handball contexts and execution characteristics, there was a need to determine whether the sample characteristics (e.g., team or field zone) influenced the outcome of handballs.These sample characteristics were included in the GLM as random effects and none of them had a significant effect (see Table 2.) The GLM was able to classify handball outcome with an overall accuracy of 87.7% (pseudo r 2 = 0.634).The only characteristics that have a significant effect on handball outcome are the type of pressure experienced by the handball receiver and the handballer, and the posture of the handballer (see Table 3).

Discussion
This study aimed to gain a greater understanding of the context and execution characteristics of handballs in AF and to determine whether these characteristics are related to the outcome of handballs.There are many combinations of contextual and execution characteristics that lead to a wide range of outcomes (i.e., handball efficiency ranged from 16.0 to 95.8% efficiency).Several characteristics were associated with handball outcome that related to the nature of defensive pressure applied in the context and on the execution of the handball.We reject the hypothesis that neither the context nor execution characteristics of handballs affect their outcome.These findings demonstrate that there are opportunities to improve offensive and defensive performance, with respect to handballs, which may affect match outcome.Pressure in the context of this manuscript relates to the nearest defender of the ball carrier and the nearest defender to the handball receiver.Three of the ten context and execution characteristics had a significant association with handball outcome, and two of these related to pressure.Pressure was assessed with respect to the handballer receiving the ball, then trying to execute the handball and finally, the pressure exerted on the player trying to receive the handball.Handball effectiveness was lowest when the type of pressure was "physical", which represents direct contact between the handballer and a defender.The highest level of handball effectiveness was achieved when the type of pressure was open (i.e., no pressure) or implied (i.e., when a defender was chasing, corralling, or closing).The present findings were similar to Parrington et al., (2013) with both studies confirming what is likely to be a widely held belief, that handball effectiveness is highest when pressure is lowest.Nevertheless, the present study analysed pressure in more detail, and it provides novel insights about when or where pressure has the greatest effect on handball outcome.Further work should be done to explore both the pressure and player density on the handballer and the receiver to understand this relationship further.
Pressure on the player that is trying to receive the handball, is the most important "pressure point".Pressure on the player executing the handball (the handballer) is important, but less so than the pressure on the receiver.These findings support those of Teune et al., (2021) who observed that more effective disposals (noting this includes both handballs and kicks) occurred when player density around the ball disposing player was higher, however, this was a weak association.The definitions of our pressure characteristics provide an opportunity to evaluate the relative effectiveness of physical versus implied pressure.This comparison helps to elucidate an important, broader tactical issue -is the cost of physical pressure worth the benefit?Physical pressure means a player is focused on the handball receiver or the handballer, which means they are unable to simultaneously apply defensive pressure to any other players.Implied pressure may be less effective but is more likely to allow the player to exert defensive pressure to multiple players simultaneously.The GLM indicates that when pressure is applied to the handball receiver, physical pressure is 25.3 times more effective than implied pressure (i.e., odds ratio of 0.009 versus 0.228).Although, when pressure is applied to the handballer, physical pressure is only 2.6 times more effective than implied pressure (i.e., odds ratio of 0.850 versus 2.199).This indicates that the most effective defence is to apply physical pressure (not implied) to the handballer and the handball receiver/s.In some situations, there are not enough defenders available to apply physical pressure to the handballer and all the potential receiver/s of the handball.Furthermore, for tactical reasons, a coach might want to allocate defenders to a specific role such as "tagging" an opposition player.Therefore, there is a need to understand the relative effectiveness of defensive strategies that do not rely entirely on physical pressure.The results indicate that the next-best approach is to maintain physical pressure on the handball receiver, but use "open" or "implied" pressure on the handballer.These options might make the best use of the relative benefits of different types of pressure and may allow a defensive team to meet a broader range of simultaneous defensive needs.More generally, the effect of pressure on performance in AF has been explored with respect to contests to gain possession of the ball (Robertson, Back et al., 2016;Robertson, Gupta et al., 2016;Spencer et al., 2019;Young et al., 2019), and it has been shown that contested possessions can have an influence on match outcome.This reinforces the possibility that defensive pressure on handballs is associated with and may affect possession and match outcome.
Other than defensive pressure, the only other characteristic that was associated with handball outcome was the body posture of the handballer.When the handballer was standing, a successful handball outcome was 1.87 times more likely than if they were on the ground (i.e., odds ratio of 1.87 versus 1.0).These results align with the findings of Parrington et al., (2013) who reported that a ball received, by a standing player, in a predictable manner, is more likely to result in a successful outcome.Much like our findings about the effect of pressure on handball outcome, these findings are likely to confirm existing beliefs.Wherever possible, players should Table 3.The effect of the context and execution characteristics on the outcome of handballs.The characteristics were evaluated as fixed effects within the model and statistical significance was accepted when p < 0.05.For significant effects, the odds ratios represent the probability of achieving a handball outcome of "complete", compared to the reference category that has an odds ratio of 1.000.It is important to briefly consider the handball characteristics that are not associated with outcome, because they may have practical implications.The duration of possession by the handballer, the hand they use to handball, the relative direction of the handball and even the direction in which pressure is applied to the handballer, did not affect the outcome of the handball in the current study.Notwithstanding contextual characteristics, field position, and player outnumbering, players can feel free to use any versions of these characteristics that they like (e.g., any handball direction), knowing that there are no sub-optimal choices.
Completed handballs accounted for ~75% of the handballs in the current study which is marginally under the 82% efficiency presented previously (Australian Football League, 2019).This may result from approximately 50% of the data being representative of one club rather than all team from the league.Understanding the optimal contextual and executional aspects of handballing may improve players' decision-making during a match.Understanding both, the most common contexts in which handballs occur, including when handballs are less effective, allows coaches to manipulate training design to ensure competition scenarios are replicated, but also to try and improve handball effectiveness in scenarios where it might be less effective but occur regularly (e.g., when the player does not deliberately gain possession of the ball or when pressure is high).Ireland et al., (2019) identified that most handball task-constraint frequencies were greater at training and current training design did not replicate the technical demands of competition, an apparent overload of handballing.The design of drills would better prepare athletes for competition if executing (i.e., handballs) under the same conditions, coaches may see an improvement in skill performance, and thus, improve handball efficiency during competition.
The design of this study involved several limitations.The source of data may not be perfectly representative of all AFL clubs and levels of AF, therefore, caution needs to be applied when generalising to broader contexts.A more extensive analysis may consider analysing a whole season of handballs providing a more representative data sample.There are other handball characteristics that could have been included in this study, including information about team formation around the handballer.Future research could explore the relationships between handball characteristics and broader measures of team performance, such as scoring and match outcome.
Australian Football is dynamic in nature and the measures from the current study reflect static points in time.The majority of pressure metrics used in AF, traditionally represent the nearest defender to the ball carrier.Consequently, they may not be effective at representing the dynamic nature of AF and density of players that is often situated around the ball carrier.Development of continuous quantitative measures have the potential to reflect play more accurately and generate further understanding of how the performance context (e.g., pressure) may affect disposal efficiency.

Conclusion
In conclusion, handball performance is an important aspect of team performance in AF and the present analysis identifies the factors associated with their outcome.The defensive pressure applied to the receiver of the handball and to the handballer, and the posture of the handballer are the most important determinants of a team's handball effectiveness.Of the types of pressure that defenders can apply, physical pressure should be applied to the handballer and the handball receiver, or at least implied pressure to the handball receiver.Where possible, defenders should prevent the handballer from executing the handball whilst standing.These conclusions are valuable for coaches in the elite environment and wider community of AF as they may use this information to design training to improve the team's use and efficiency of handballs or consequently, reduce the opposition team's effectiveness.

Table 1 .
Descriptive statistics of context and execution characteristic, with outcome and efficiency.The results represent all (n = 1342) handballs from four Australian

Table 2 .
Evaluation of the effect of the Match, Team, Quarter and Field Zone on handball outcome.These factors were evaluated within the model as random effects.Field zone was redundant in the model and its covariance was not calculated.