Peer value

Why role-relative

Ranking every skater on one list buries a shutdown third-pair defenseman under a first-line scorer. That happens even when the defenseman is excellent at his job. A depth forward will never out-produce a franchise center. Holding him to that bar tells you nothing about how well he plays his actual role. Peer value fixes the comparison. Each player is ranked and given a percentile only against others in the same season who play the same role. The question changes. It is no longer whether he is the best player in the league. It becomes whether he is the best player at this job.

How the peer group is drawn

A peer group is a position group crossed with a usage tier, drawn fresh each season. Skaters are split into forwards and defensemen. Each position is then cut into thirds by season-long ice time. Forwards become Top-6, Middle-6 and Bottom-6. Defensemen become Top-4, 2nd-pair and 3rd-pair. Goalies split into Starters and Backups by workload. That makes eight separate groups, each ranked from best to worst on its own scale. Within a group, skaters are ordered by their season value (GAR). Goalies are ordered by their shot-quality-adjusted goaltending (GSAx). The percentile runs from 0 to 100, with 100 the best in the group.

How it stays honest

Two guardrails do the work. The first is an eligibility gate. Only players with at least 20 games in a season (20 starts for goalies) are ranked. So a hot handful of games can't produce a peer-group crown. The second is that every group stands alone. A player's rank runs 1 through N inside his own tier and nowhere else. The reported group size is exactly the number of players ranked against him. Nothing bleeds between tiers or across seasons. The comparison set is fixed to same-season, same-role peers before anyone is ranked.

What the groups actually look like

This is a ranking rather than a prediction. So whether it holds up comes down to how carefully the groups are drawn. That makes it worth showing exactly how they sit. The check below is redone from the published rows every time the numbers update. It covers 2023-24 through 2025-26. It includes 2025-26, the season the boards are showing. Across those 24 groups the tier sizes are stable and sensible. The rules hold with zero exceptions. Every group's rank runs 1 through N. Every reported group size is exactly the number of players ranked against that player.

Take 2025-26 as an example. The table shows how many players each of the eight groups held.

Within every one of those tiers the rank ran cleanly from 1 to the tier's size. No player slipped under the 20-game floor.