New methodological approaches to better understanding the attacking process during open play moments in football

Attacking team possessions in football has been classified in terms of the number of passes, duration or using descriptive labels such as build up and direct play. However, little consideration has been given to the fact that teams probably utilise all possession types at various times and can be a consequence of teams playing with a high press or low block. In addition, playing styles are normally described only when football teams attack during open play moments i.e. not when a shot occurs directly from a set piece as these are very different situations involving specialist formations for each team. Furthermore, not all possessions will necessarily aid the description of playing style e.g. low duration ones, whereas some may be more complex than to be categorised into one style. The three studies in this thesis aimed to describe the attacking process using novel methodologies, during open play moments which started when the ball was regained or from set pieces that developed into open plan situations. All studies assessed all matches involving the first and second placed teams (Manchester City and Liverpool) in the 2018-19 English Premier League season. Study 1 defined indeterminate team possessions as possessions that did not allow the attacking intent to be determined. These accounted for 25% of the data and were excluded from the two step cluster analyses. This analysis detailed the attacking process in terms of where possession started, how the ball progressed forward and where the possession ended, including whether a resultant shot or goal occurred. Manchester City and Liverpool differed on 6 of the 16 derived clusters with results showing that Manchester City tended to build up play from the back and had more possessions that started in the final third of the pitch compared to Liverpool. The second study used an additional variable to classify how a team tried to break the opponent’s defensive line. If this was successful, but the team subsequently passed the ball back behind the opponent’s defensive line, this was coded as a new phase of the same team possession. Data mining was used to produce association rules for the sequential possession data and decision trees then used to determine which sequential patterns produced the most shots at goal. Association rules allowed the identification of repeated possession sequences for different specific situations. For example, Manchester City had 34 different attacking patterns (Liverpool 33) after patterns with rule support of less than 1% were removed. Manchester City’s most used team possession was build-up play culminating in a pass to the wide area. Liverpool utilised both fast transitions from defence to attack and build-up play with the significantly greater use of long passes from behind than Manchester City. The third study assessed the attacking strategies of the two teams under the combined influence of all levels of four contextual variables (match status, match venue, opposition quality and time period). Decision tree models for each team, confirmed that match status was the most important contextual variable, followed by match venue. Post hoc Node diagrams illustrated how the possessional phases changed in these decision tree models. Both Manchester City and Liverpool typically maintained similar possessional phases throughout matches, suggested to reflect their ability to control matches. Manchester City’s did increase their use of direct play at the expense of fast breaks during losing situations in the second half of matches, albeit a relatively unusual occurrence. Liverpool tended to have more fast breaks in winning situations and during the last 30 minutes of matches there was an increased use of direct play. The fine-grained analysis presented in this thesis has aided the closing of the so-called theory-practice gap. However, future studies should consider the most appropriate methods to examine multiple individual teams within an analysis that maintains practical significance.