KNLTB – A Model for Optimal Rating in Tennis Matches #SWI2020

KNLTB

KNLTB is the Royal Dutch Lawn Tennis Association. The main goal of KNLTB is to support as many people as possible – of all ages and (dis-)abilities – to play tennis and make sure they keep playing. We represent 560.000 tennis members associated with approximately 1650 local clubs. KNLTB coordinates the annual match calendar, including all official Competitions and Tournaments at all Tennis levels.

Player Profile

The skills of a Tennis player for both Singles and Doubles are expressed in a Level and in a Rating. The Tennis Level is assigned annually and ranges from 9 (beginner) to 1 (professional). The actual Rating (e.g. 4,2125) represents the progress of the player after all the tennis matches played during the season.

Player Profile: Singles (4; 4,3594), Doubles (Level 3; Rating 4,2125) and Padel (5; 3,9199)

Problems
Rating is an average of at least 6 rating results during the year, in combination with 6 results based on the previous year. As a consequence, undesirable rating fluctuations occur at the start of the year, whereas at the end of the year matches have little effect on the rating. The current rating method was designed for Singles and does not have the desired effects in Doubles.

1. The calculation has multiple constraints, especially in Doubles where the two partners have diverse levels. Despite winning a match, players often don’t receive a positive Rating result;
2. The system makes use of a combined Rating for both Tennis partners. We believe research is needed to decide if the match results should be calculated combined and/or individually;
3. At the end of the season multiple data corrections in match results are needed, to assign a reliable annual skill Level. In Doubles the compensation for irregular Rating results is too complicated;
4. The final question is how to deal with new players and re-entrants without Rating or prior matches.

Challenges

The Rating has to become a better indicator of the Tennis skills and predictor for results of future matches. Our goal for the SWI2020 team is to find a mathematical model that:

• Calculates a representative positive result for winners and a negative result for lost matches;
• Takes into account the most relevant factors that influence the outcome of a tennis match;
• Gives weight to the (un)certainty in the players Level and Rating based on previous matches;
• Focuses on complex Doubles Rating, suggestions to improve Singles algorithm are welcome;
• Allows for players of diverse levels to play Tennis together;
• Reduces the data corrections in the annual assignment of skill Levels;
• Can be explained to our members, convincing them of the fairness of the algorithm.

An anonymized data set is available at Player, Rating, and Match level for multiple years.

Solution

If the model is able to track the Rating results and progress (or decrease) during the season, Doubles players will be stimulated to keep playing official matches. The improved Rating can support higher quality match schedules in Competitions and Tournaments. Ultimately Players are better rewarded for their Match results and the Rating will be a valuable indicator and predictor for future matches.