This thesis grew out of a five-month internship at Scala Sports, a company that runs online amateur tennis leagues, and it tackles two questions that turned out to be closely related: why do players keep coming back, and how should a league rank and pair them so that the competition feels fair? The first part is an empirical study built on the company's own data. After extracting and cleaning match, player and rating records from the production database, I looked at what keeps players in a league from one season to the next. Across roughly 351 USTA leagues and more than 12,000 players, retention turns out to hinge on a handful of things a league organizer can actually control: how long a season lasts, the gap before the next one starts, the format, and how quickly a new player gets their first match. I also designed a simple metric to capture how exciting a match was, and trained churn models whose explanations (SHAP, permutation importance) line up with the descriptive findings and feed a short list of practical, early-warning recommendations. The second part moves from diagnosis to design. Motivated by those fairness concerns, I studied the Kaizer ladder, a ranking system in which a player's worth comes from the opponents they beat rather than a fixed rating. Using simulated tournaments, I tuned its main rules: how much credit to give for skipping a round or joining late, where to place newcomers in an established ladder, and how to pair players so they rarely meet twice while matches stay balanced. The result is a validated and shareable implementation of the system, together with the evidence behind each design choice.

From Retention Analytics to System Design: A Two-Part Study for Scala Sports Flex Leagues

GASTALDELLO, MATTEO
2025/2026

Abstract

This thesis grew out of a five-month internship at Scala Sports, a company that runs online amateur tennis leagues, and it tackles two questions that turned out to be closely related: why do players keep coming back, and how should a league rank and pair them so that the competition feels fair? The first part is an empirical study built on the company's own data. After extracting and cleaning match, player and rating records from the production database, I looked at what keeps players in a league from one season to the next. Across roughly 351 USTA leagues and more than 12,000 players, retention turns out to hinge on a handful of things a league organizer can actually control: how long a season lasts, the gap before the next one starts, the format, and how quickly a new player gets their first match. I also designed a simple metric to capture how exciting a match was, and trained churn models whose explanations (SHAP, permutation importance) line up with the descriptive findings and feed a short list of practical, early-warning recommendations. The second part moves from diagnosis to design. Motivated by those fairness concerns, I studied the Kaizer ladder, a ranking system in which a player's worth comes from the opponents they beat rather than a fixed rating. Using simulated tournaments, I tuned its main rules: how much credit to give for skipping a round or joining late, where to place newcomers in an established ladder, and how to pair players so they rarely meet twice while matches stay balanced. The result is a validated and shareable implementation of the system, together with the evidence behind each design choice.
2025
From Retention Analytics to System Design: A Two-Part Study for Scala Sports Flex Leagues
Player retention
Sports analytics
Ranking systems
Matchmaking
MC simulation
File in questo prodotto:
File Dimensione Formato  
Gastaldello_Matteo.pdf

Accesso riservato

Dimensione 5.05 MB
Formato Adobe PDF
5.05 MB Adobe PDF

The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12608/115837