Player value
Each skater begins with an individual value profile built from GAR, Evo-WAR, scoring rates, usage, and workload context rather than a simple points-based ranking.
Methodology
How the optimizer actually builds a lineup.
This model does not just rank players individually. It weighs player value, line chemistry, and role-specific deployment fit to decide which combinations make the most sense for Vegas.
Line chemistry
Forward trios and defense pairs rise or fall based on direct combo results, chance share, differential, and sample strength when those units have real tracked data.
Role logic
The same trio can grade differently depending on whether it is being judged for top-line creation, second-line support, checking-line balance, or hard-minute defensive work.
Deployment realism
Defense pair scoring includes continuity and deployment context so proven Vegas pairs are not broken apart unless the model sees a clear enough gain to justify it.
Model Inputs
The optimizer is built around four core stats inputs.
skaters
21
Vegas playoff skater pool with value and deployment inputs.
goalies
3
Three-goalie set with performance metrics for starter and backup scoring.
forward combos
61
Forward trio efficiency rows used to reward real chemistry instead of guesses.
defense pairings
22
Defense pair deployment and efficiency rows used for pair selection.