VolleyAI uses an equivariant graph neural network to read your film and surface the highest-probability plays — serve receive, setter options, and formation tweaks.
Live preview
Tap through the tabs — this is the real VolleyAI workflow on a sample rally.
Serve Receive Analyzer · predicted pass-quality heatmap
Predicted receiver L1 · pass rating 2.4/3
How it works
Drop a match clip — phone or broadcast, any camera angle.
Tracking → court stabilization → equivariant graph inference.
Win probability, setter options, and formation tweaks in seconds.
The science
VolleyAI is built on VolleyAI: A Graph Neural Network Assistant for Volleyball Tactical Decision Support — modeling the rally as a VolleyGraphwith players, ball contacts, and court zones as nodes, and passes, blocks, and responsibilities as edges. Inference is equivariant under the D₂ × C₆ symmetry group, so reads hold for any court rotation or reflection.
The approach is inspired by Google DeepMind’s TacticAI (football corner kicks, Liverpool FC, 2024), adapted to volleyball’s faster rotations and shorter rallies — and trained on the VREN dataset (7,000+ NCAA/professional rallies).
81.2%
rally accuracy
59.1%
set location
86.4%
hit type
68%
coach-preferred
Preprint submitted to Nature Communications / MIT Sloan Sports Analytics Conference 2026. Code: github.com/volleyai/volleyai (MIT). Data: VREN (CC-BY-4.0).
Inspired by
Google DeepMind
TacticAI proved geometric graph neural networks with D₂-equivariant convolutions can predict receivers and generate coach-preferred adjustments for football set-pieces. VolleyAI applies the same family of methods to volleyball.
Model
voll3y-gnn-v1
Trained on VREN rally tracking · serves, passes, sets, attacks, blocks, digs.
Features
Predicted pass quality & receiver heatmap from tracking data.
“Middle vs outside?” with kill probabilities per set option.
Minimal player adjustments that maximize rally win probability.
D₂ × C₆ symmetry-aware GNN — court-rotation invariant.
Benchmarks tuned to high-tempo college volleyball tempo.
Built on published graph-neural-net volleyball research.
Pricing