These probabilities can then be added up per team and yield a “result-agnostic” description of the teams' performance. An expected goals model (hereafter xG model) tries to estimate the probability of any given shot being converted to a goal based on various different factors describing the shot. However, assessing shots just by being successful or not is a too rough abstraction that warps reality. As another consequence of this low-scoring nature, the role of shots as a success proxy within several studies in football is fortified ( Spearman et al., 2017). Consequently, the rareness and therefore importance of goals makes such a metric even more relevant when assessing teams and players. The fact that football is the lowest scoring game of the above-mentioned sports, makes it harder to develop such models, because of the scarcity of data. Similar shot prediction models were also developed for ice hockey ( Macdonald, 2012) as well as for return plays in tennis ( Wei et al., 2016) and table tennis ( Draschkowitz et al., 2015). The high scoring nature of basketball enables clubs to go even further and to apply individual shooting efficiency models ( Beshai, 2014).
Another famous example is basketball: By calculating scoring probabilities of different shot locations ( Reich et al., 2006 Chang et al., 2014 Harmon et al., 2016 Jagacinski et al., 2019), the NBA's shooting behavior changed significantly 3. In baseball, scouts and experts focused their attention on homeruns or hits for decades until more complex evaluation metrics changed the assessment procedure of hitters' performance significantly ( James, 1985). For example, the performance from an outstanding creative player could be made void by strikers missing all their chances.įor this reason, in football as well as in other sports, it has become typical to consider more granular process-based metrics. Nevertheless, judging performances solely based on this binary metric ( goal or no goal) loses a lot of information and places results over process. For example, both the best goal scorers 1 and the players with the most assists 2 receive a lot of attention from experts and the media. However, goals alone decide the outcome of a game and are the most common metric to judge both a team's and individual player's performance.
#Shot online weather calculator professional
In professional football (soccer), only 1% of all attacking plays and only around 10% of all shots taken end up in a goal ( Pollard and Reep, 1997 Tenga et al., 2010 Lucey et al., 2014).
This approach allows us to assess team and player performances far more accurately than is possible with traditional metrics by focusing on process rather than results. With a ranked probability score (RPS) of 0.197, it is more accurate than any previously published expected goals model. The best performing model uses an extreme gradient boosting algorithm and is based on hand-crafted features from synchronized positional and event data of 105, 627 shots in the German Bundesliga. This model is validated statistically and with professional match analysts. The aim of this study is to objectively quantify the quality of any given shot by introducing a so-called expected goals (xG) model. However, not all shots are created equally and their quality differs significantly depending on the situation.