Team Dynamics in DotA2 through Attention Mechanism - GREYC image
Communication Dans Un Congrès Année : 2024

Team Dynamics in DotA2 through Attention Mechanism

Résumé

We analyse team dynamics in the popular e-sports game DotA2 using an approach that combines convolutional and LSTM networks with a featuretemporal attention mechanism. Our goal is to identify strategic behaviours that lead to successful goals, such as scoring kills, during World Championship matches. Each team's formation is represented by a polygon, which feeds an RNN that learns kill events from this polygon under its area, diameter, and moments around the centroid. By exploiting the attention mechanism, our network highlights the most relevant features at each time step, providing insights into strategic team movements and formations. Our results demonstrate the effectiveness of our approach in capturing critical dynamics that influence the outcome of engagements in DotA2.
Fichier principal
Vignette du fichier
MLSA2024_Team_Dynamics_in_dota_through_Attention_Mechanism.pdf (6.35 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04695698 , version 1 (12-09-2024)

Identifiants

  • HAL Id : hal-04695698 , version 1

Citer

Alexis Mortelier, Sébastien Bougleux, François Rioult. Team Dynamics in DotA2 through Attention Mechanism. Machine Learning for Sports Analytics, Sep 2024, Vilnius, Lithuania. ⟨hal-04695698⟩
0 Consultations
0 Téléchargements

Partager

More