Goals Above Replacement (GAR)
A comprehensive metric that estimates how many goals a player contributes above what a replacement-level player would provide in the same playing time.
Goals Above Replacement (GAR) is the flagship all-in-one metric in modern hockey analytics. It answers a deceptively simple question: how many goals is this player worth compared to a freely available minor-league call-up? GAR rolls offensive production, defensive impact, special teams play, and penalty behaviour into a single number, making it possible to compare a shutdown defenseman to a goal-scoring winger on the same scale.
Hockey Alchemy's GAR model combines counting-stats components (EVO, EVD, PPO, SHD, Penalties) with a Regularized Adjusted Plus-Minus (RAPM) overlay that isolates individual contributions from teammate and opponent effects. Each component is calibrated so the league-wide sum of GAR correlates strongly with actual goal differentials and standings points.
A replacement-level player is defined as the caliber of skater a team could acquire for minimal cost -- typically a waiver pickup or AHL call-up. Because that baseline is set from what freely available talent actually produces, and not from league average, a substantial minority of NHL regulars finish below it: across the RAPM era roughly a quarter to a third of qualified skater seasons end with negative GAR.
Read GAR as a rank within its season rather than against fixed thresholds. The median qualified regular sits only a few goals above replacement; the top quarter is comfortably positive; and 'elite' is best understood as the top decile, with Hart-caliber seasons in the top one percent. Deliberately no ceiling is quoted here -- the RAPM model is re-fit as new games land, so the single best season in the database moves between runs, and any hard number printed in a glossary goes stale by construction.
Formula
GAR = EVO + EVD + PPO + SHD + Finishing + Penalties (drawn + taken) WAR = GAR / 5.57 goals per win
Components
| Component | Description |
|---|---|
| EVO | Even-strength offensive value: goals and expected goals generated at 5v5. |
| EVD | Even-strength defensive value: goals and expected goals suppressed at 5v5. |
| PPO | Power-play offensive value: contribution on the man advantage. |
| SHD | Shorthanded defensive value: penalty-kill impact. |
| Finishing | Shooting talent: goals scored above what the shot locations alone expected. |
| Penalties | Penalties drawn (positive) and penalties taken (negative), carried as two separate terms. |
Frequently Asked Questions
What is a good GAR in the NHL?
Think in percentiles rather than fixed cutoffs. A replacement-level player contributes 0 GAR by definition. Among qualified regulars, the median season lands only a few goals above that, the top quarter is clearly positive, the top decile is the usual bar for 'elite', and Hart-caliber years live in the top one percent. Roughly a quarter to a third of qualified regulars finish below zero in any given season -- that is the baseline working as intended, not a bug.
How does GAR differ from WAR?
WAR (Wins Above Replacement) is simply GAR converted to wins. Hockey Alchemy uses an empirical 5.57 goals per win, measured from its own stored GAR and WAR totals. The two metrics rank players identically.
Does GAR account for goaltending?
Skater GAR does not include goaltending. Goalies have a separate GAR based on Goals Saved Above Expected (GSAx), which measures how many goals a goalie saves relative to the expected goal quality of shots faced.
Why might two GAR models disagree on a player?
Different models (Evolving Hockey, Hockey Alchemy, TopDownHockey) use different methodologies for isolating individual contributions -- especially for defense and shorthanded play. The defensive component is the most difficult to measure and the biggest source of disagreement between models.