The story of KNLTB
The challenge
The Royal Dutch Lawn Tennis Association (KNLTB) was looking for an improved ranking methodology for Double players in national and regional official competitions, which has to become a better indicator of the tennis skills and better predictor for results of future matches. Sometime we observe strange things in the ranking of Double players, explains Helga Wiersma of KNLTB. “Like minus points when you win a match”. In that case a couple wins the tennis match, however the ranking of one of them is deteriorating to a lower level.
The mathematics
The Study Group Mathematics and Industry performed preliminary research and analysis:
- Logistic regression analysis to evaluate the current rating system DSS and to compare to alternative methodologies: Universal Tennis Rating (UTR), Elo, Glicko and Microsoft TrueSkill
- Development of a new statistical model based on logistic regression to combine ratings for doubles matches satisfying the practical constraints
- Bayesian update rules based on Kullback-Leibler divergence to distribute the rating change over the players of a doubles team based on how much they contributed
- Nonlinear optimization to compute likelihood needed for Bayesian updates.
Tau Omega further developed the model and implemented it into tooling for direct use by KNLTB.
The result
Accurate player levels that motivate members to play official matches leading to increase of positive publicity for KNLTB.
Helga Wiersma – KNLTB
“Ik heb bewondering voor wat de groep in een week heeft bereikt. Het was een project waarvan we al een tijdje wisten dat we er wat mee wilden. Door SWI2020 heeft het vleugels gekregen.”
Additional information
Popular proceedings of SWI2022 (in Dutch) can be found here.