at Google ScholarTitle data
Rambau, Jörg ; Richter, Rónán R.C.:
Finding optimal strategies for substitutions in soccer using a two-scale Markov Decision Process approach.
2019
Event: MathSport International 2019 Conference
, 1.-3. Juli 2019
, Athen, Griechenland.
(Conference item: Conference
,
Speech
)
Abstract in another language
While a coach in soccer has much influence on his team’s chances of winning before the match, e.g. by designing the training, giving tactical plans or choosing the starting line-up, his abilities during the match are very limited. One of his main ways of influence are strategic substitutions. A substitution will be called strategic, if it aims to change the formation of the team. Possible questions in that context are: In which cases should a coach send an additional defensive player onto the field in order to save a lead or when to introduce a striker for more offensive possibilities? Questions of this type can be tackled by modeling a soccer match as a two-scale Markov Decision Process (2MDP), that consists of a detailed MDP modeling the gameplay and a coarse MDP modeling the strategic decisions. This concept has recently been introduced by Hoffmeister and Rambau, who used beach volleyball as a first application. One main feature of this model is, that it does not relay on historic data on the success of the different decision rules against a specific opponent, that is rarely available, but only on data of the players’ abilities, that can be observed independent of other influences. In this talk, a 2MDP model for strategic substitutions in soccer is presented. The differences to net sports are outlined. Additionally, first results for optimal substitution strategies on artificial data are given and the importance of decisions on substitutions in various scenarios is discussed.