Round Outcome Prediction in VALORANT Using Tactical Features from Video Analysis

TL;DR

The study tackles round outcome prediction in VALORANT from broadcast footage by leveraging TimeSformer to analyze minimap visuals. It compares a minimap-only dataset with an augmented dataset that includes labeled tactical events, finding that tactical features boost overall accuracy to around 80% and provide clearer gains in the mid-to-late rounds. The work demonstrates the value of extracting on-screen tactical cues from video for real-time predictions, with implications for enhanced viewer engagement and coaching. It also outlines future directions to incorporate movement and player-history data to further improve predictive performance.

Abstract

Recently, research on predicting match outcomes in esports has been actively conducted, but much of it is based on match log data and statistical information. This research targets the FPS game VALORANT, which requires complex strategies, and aims to build a round outcome prediction model by analyzing minimap information in match footage. Specifically, based on the video recognition model TimeSformer, we attempt to improve prediction accuracy by incorporating detailed tactical features extracted from minimap information, such as character position information and other in-game events. This paper reports preliminary results showing that a model trained on a dataset augmented with such tactical event labels achieved approximately 81% prediction accuracy, especially from the middle phases of a round onward, significantly outperforming a model trained on a dataset with the minimap information itself. This suggests that leveraging tactical features from match footage is highly effective for predicting round outcomes in VALORANT.

Figures (1)

• Figure 1: Prediction Accuracy