Random Walks with Memory Applied to Grand Slam Tennis Matches Modeling

0506074 - ÚTIA 2020 GR eng C - Konferenční příspěvek (zahraniční konf.)
Kouřim, Tomáš
Random Walks with Memory Applied to Grand Slam Tennis Matches Modeling.
MathSport International 2019 Conference Proceedings. Athens: Propobos Publications, 2019 - (Karlis, D.; Ntzoufras, I.; Drikos, S.), s. 220-227. ISBN 978-618-5036-53-9.
[MATHSPORT INTERNATIONAL 2019 CONFERENCE. Athens (GR), 01.07.2019-03.07.2019]
Institucionální podpora: RVO:67985556
Klíčová slova: random walk * statistical analysis * sports statistics
Obor OECD: Statistics and probability
http://library.utia.cas.cz/separaty/2019/SI/kourim-0506074.pdf

The contribution presents a model of a random walk with varying transition probabilities implicitly depending on the entire history of the walk, which is an improvement of a model with varying step sizes. The transition probabilities are altered according to the last step of the walker using a memory parameter to either reward or punish success by increasing or decreasing its probability in the next step. This walk is applied to model Grand Slam tennis matches and fitted on their entire history since 2009. The suitability of the model is thoroughly tested on a number of real datasets.
Trvalý link: http://hdl.handle.net/11104/0297358