TacSIm: A Dataset and Benchmark for Football Tactical Style Imitation

Abstract

Current football imitation research primarily aims to opti mize reward-based objectives, such as goals scored or win rate proxies, paying less attention to accurately replicat ing real-world team tactical behaviors. We introduce Tac SIm, a large-scale dataset and benchmark for Tactical Style Imitation in football. TacSIm imitates the acitons of all 11 players in one team in the given broadcast footage of Pre mier League matches under a single broadcast view. Under a offensive or defensive broadcast footage, TacSIm projects the beginning positions and actions of all 22 players from both sides onto a standard pitch coordinate system. Tac SIm offers an explicit style imitation task and evaluation protocols. Tactics style imitation is measured by using spatial occupancy similarity and movement vector similarity in defined time, supporting the evaluation of spatial and tem poral similarities for one team. We run multiple baseline methods in a unified virtual environment to generate full team behaviors, enabling both quantitative and visual as sessment of tactical coordination. By using unified data and metrics from broadcast to simulation, TacSIm estab lishes a rigorous benchmark for measuring and modeling style-aligned tactical imitation task in football.

Figures (3)

• Figure 1: Overview of TacSIm. (a) Video matches in real world: Frames captured from televised football matches, segmented into offensive and defensive phases to display players' real-time positions on the field. (b) Player coordinate reconstruction: Mapping real-time player positions from broadcast frames to a normalized pitch coordinate system. (c) Tactical replication in a virtual football environment. The current contexts of reconstructed players, such as the actions and positions, are fed into a virtual football simulation platform where each player is treated as an agent. by multi-agent system learning, the following contexts of each player are reproduced and can be compared with the real contexts of each player in the following time of the real football match. The reproduced behaviors in virtual football environment can be visualized.
• Figure 2: Overview of the trajectory completion framework with demonstrator–learner imitation structure.
• Figure 3: Visualization of tactical imitation.This figure shows the comparison between Ground Truth (purple solid line) and Inference (white dashed line) for player trajectory predictions in different tactical contexts. The colorful blocks on the field represent the spatial grid, showing the distribution of the player's movement across different zones during the match. From left to right, the grid resolutions are $15 \times 10$, $20 \times 12$, and $30 \times 20$, illustrating how varying granularity affects spatial coverage and prediction consistency.