Buildup Path Cluster Analysis

Using cluster analysis (K-means or similar) to identify distinct archetypes of buildup paths — the spatial trajectories teams take when building from their own quarter to the offensive zone. Rather than analyzing individual possessions, clustering reveals 4-6 recurring buildup patterns across a league, each with different shot and goal probabilities. The key finding: sideline-to-center cross patterns are the most dangerous (~50% shot rate), while buildups trapped on the sideline are the least effective.

For each buildup possession (starting in own quarter, reaching the final quarter): (1) extract the sequence of ball positions as a spatial trajectory, (2) normalize trajectories to a standard pitch, (3) cluster using K-means or hierarchical clustering on the spatial features (positions along the trajectory, lateral spread, final entry point). The four main clusters typically found:

• Right sideline: stays right, enters zone near sideline — moderate danger
• Right to center: goes right then crosses to center before entering zone — high danger
• Left sideline: stays left, enters near sideline — moderate danger
• Left to center cross: goes left then crosses to center — highest danger (~50% shot rate)

The cross-to-center patterns are most dangerous because the entry point is central. The sideline patterns are less effective because the team gets "trapped" wide. Footedness effects appear: left-to-center-cross may be more effective than right-to-center due to right-footed players cutting inside.

1. Too few clusters: 2-3 clusters may miss the critical cross-to-center pattern. Start with 4-6 and evaluate.
2. Ignoring footedness: Left-to-center and right-to-center patterns may have different effectiveness due to dominant foot cutting inside.
3. Prescribing only the highest-value pattern: A team that always crosses to center becomes predictable. The analysis informs when the pattern is available, not that it should always be attempted.