Right Move, Right Time: Multi-Sport Space Evaluation Platform for Ultimate Frisbee, Basketball, and Soccer

TL;DR

An open, sport-agnostic platform that turns tracking into comparable spatial measures across professional Ultimate, basketball, and soccer is presented and a practical path toward consistent, comparable evaluation across various invasion sports is shown.

Abstract

We present an open, sport-agnostic platform that turns tracking into comparable spatial measures across professional Ultimate, basketball, and soccer. Coaches in all three sports ask the same question: where is the usable space, and when should an off-ball run start? Our workflow standardizes inputs, provides timing-aware spatial evaluations, and makes it possible to reuse the same analysis across sports. We illustrate the approach with Ultimate as a focused testbed and then examine transfer between basketball and soccer. Together, these results show a practical path toward consistent, comparable evaluation across various invasion sports.

Figures (6)

• Figure 1: Overview of our multi-sport pitch control platform. Left: The inputs are Ultimate Frisbee, basketball, and soccer tracking data. Middle: sport-agnostic preprocessing and space evaluation via Preprocessing and SpaceEval packages in the open-source Python library called OpenSTARLab. Right: Obtained Ultimate, basketball, and soccer space value maps. Preprocessing Package: https://github.com/open-starlab/PreProcessing and SpaceEval Package: https://github.com/open-starlab/spaceEval.
• Figure 2: UFATrack dataset. This dataset is derived from a part of the official UFA match footage between the Oakland Spiders and the Salt Lake Shred held on June 27, 2025. The data contain detailed spatio-temporal tracking annotations of player xy coordinates. This dataset anchors the Ultimate analyses and is released for reproducible research. The dataset is shared at: https://zenodo.org/records/17840132
• Figure 3: (a) Histogram of the maximum $V_{frame}$, showing that SLC generated a larger number of medium-quality plays, whereas OAK produced fewer but higher-quality plays. (b) Histogram of $V_{timing}$, which did not differ substantially between the teams, indicating that the five-point margin may not reflect a true difference in overall outcome.
• Figure 4: (a) The line plot in the lower panel shows, for all possessions in the second quarter, the proportion of grid cells with wUPPCF values exceeding 0.7 (High-wUPPCF area ratio) for each team at 5-second intervals, visualizing how safe passing areas for the offense are created over time. Larger values indicate that a wider region of the field is under favorable pitch control for the offense. (b) Example of OAK’s first turnover. In a situation where the offense has limited advantageous space, a pass is thrown into an area with strong defensive pressure, resulting in the disc being blocked. (c) Example of OAK’s fourth turnover. Although a sufficient amount of advantageous space is available for the offense, the turnover is caused by an offensive error. As shown in (a), all six possessions in which the High-wUPPCF area ratio exceeded 0.4 at least once resulted in goals, whereas among the eight possessions in which it never exceeded 0.3, only five resulted in goals.
• Figure 5: Off-ball initiation timing in UFA. (a) actual run; (b) counterfactual with the receiver’s start delayed by +10 frames. Blue markers indicate offensive players, red markers indicate defensive players, and the black marker denotes the disc. The darker-colored markers represent the offensedefense pair corresponding to the detected movement initiation. $V_{frame}$ is defined as the average wUPPCF value within the blue circle (with darker blue in the background indicating a more advantageous situation for the offense). The graph at the bottom shows the temporal evolution of $V_{frame}$, and the maximum of its 10-frame moving average is defined as $V_{scenario}$.