NHL

NHL

What Predicts an NHL Goalie’s Next Season?

What Predicts an NHL Goalie’s Next Season?

Forecasting the sport’s most volatile position means separating shot environment, repeatable skill, workload, and regression.

Forecasting the sport’s most volatile position means separating shot environment, repeatable skill, workload, and regression.

BY STAT PULSE MEDIA

Save Percentage Is Contextual

Traditional save percentage treats every shot alike even though danger varies enormously. Team defense influences pre-shot movement, screens, rebounds, rush chances, and shot locations. Forecasts should begin with shot-quality-adjusted measures and still include team context, because no public expected-goals model observes every detail a goaltender faces.

Goals Saved Above Expected

GSAx compares actual goals allowed with an expected total based on shot characteristics. It better isolates performance than raw goals against average, but model inputs and tracking quality matter. Use multiple public models when available, normalize for workload, and avoid treating small differences as definitive rankings. The metric estimates value within a model, not pure talent.

Regression Is the Default

Year-to-year goalie results fluctuate because finishing, rebounds, traffic, and limited shot samples contain noise. Extreme seasons should be pulled toward a career and league baseline. The amount of shrinkage depends on workload and history: a veteran with thousands of shots earns more individual weight than a rookie whose spectacular month contains little evidence.

Age and Workload

Goalie aging is not a smooth universal curve. Heavy workloads may reveal durability while also increasing fatigue and injury exposure. Include age, recent starts, cumulative shots, rest patterns, and injury history, with flexible nonlinear terms. Forecast availability separately from per-shot performance because a strong rate estimate has limited team value if expected starts decline.

Rebounds and Hidden Skills

Public rebound-control measures are imperfect because teams influence recoveries and clears. Still, repeat chances, frozen pucks, lateral movement, and performance after east-west passes can provide clues beyond first-shot stopping. Combine tracking proxies with scouting rather than forcing incomplete observations into one confident number.

Changing Team Environments

A goalie moving teams may face different rush defense, penalty killing, screens, and puck retrieval. Translate performance by modeling shot types and defensive structure, not by carrying last season’s goals-against average forward. Coaching and equipment changes may matter too, but they require cautious priors because public samples are thin.

Build a Distribution, Not a Point

Blend multi-year GSAx, save performance by shot type, age, workload, injuries, and projected team defense in a hierarchical model. Simulate a range for both rate and starts. Fans and teams need the chance of elite, average, and poor outcomes—not a forecast of exactly .914 that implies unrealistic precision.

Conclusion

Goalie forecasting improves when current results are treated as noisy evidence about an underlying skill level. Shot quality, multi-year history, age, workload, and defensive environment all matter, but uncertainty remains larger than at most positions. A wide honest interval is a feature of the analysis.

THE FINAL PULSE

THE FINAL PULSE

The best goalie forecast starts with shot-quality-adjusted performance, then applies regression and context. Talent can be estimated; next season’s exact save percentage cannot—and responsible projections make that uncertainty visible.

The best goalie forecast starts with shot-quality-adjusted performance, then applies regression and context. Talent can be estimated; next season’s exact save percentage cannot—and responsible projections make that uncertainty visible.

BY STAT PULSE MEDIA