IceSheet player evaluation

Here, G is roughly the number of games a single team has played so far (all league games divided by half the number of teams). Forwards need at least the larger of 3 games or 35% of G, and the larger of 45 rated minutes or 6 × G minutes. Defensemen use the same game threshold and the larger of 55 rated minutes or 7 × G. Goalies need the larger of 2 appearances or 15% of G, and the larger of 60 verified 5v5 shots or 8 × G shots. Fractional game and shot requirements round up. Every player needs 90% appearance and workload coverage. Goalies with no verified 5v5 shots cannot count as scored appearances under the current data contract.

Headline ranks require 30 qualified peers in the same position and scope. Below that minimum, results are provisional. Team and search filters leave the full peer group unchanged. Across forwards and defensemen, the players table orders season standing by each player’s positional peer percentile; their raw rates use different scales. When qualified standing is unavailable, the table falls back to raw-stat ordering. The selected date limits every season calculation.

A rank such as #16 F means sixteenth among qualified forwards. Forwards and defensemen are ranked within their own position group. Tied values share a rank; ranks use full precision, so rounded numbers can look equal without being tied. Goalies are ranked only against other goalies.

Skater season rank

Qualified workload and 90% appearance and workload coverage, as set out above

Peer minimum

At least 30 qualified skaters or goalies for ranking

Rolling skater radar

Valid observations; fixed priors and reference group, with visible weighted exposure

Rolling goalie radar

Valid axis observations; fixed approved priors, with visible weighted SOG or eligible saves

Goalie profile axes

Live: shot stopping, mid danger, low danger and rebound suppression; high danger is withheld. A separate five-spoke descriptive profile is under review.

Playoffs

Raw statistics available; no rank shown

Players below a threshold remain searchable with raw statistics and an unranked reason. Missing components never become zero or trigger redistributed weights. Games with missing shot attribution or xG are excluded from both the totals and their ice time. Early-season thresholds are not automatically reduced.

Each release uses one xG model generation and a reviewed data snapshot. A corrected game is left out of season totals and rated workload until it is replayed. A global invalidation withholds the checked evaluation. Missing data or a release awaiting review produces an explanation instead of a fallback rating. A missing or incompatible historical component reference withholds its percentiles while valid raw component rates remain available. Descriptive radar percentiles can remain available under their own eligibility rules; a reviewed axis whose own evidence fails is withheld and shown with its reason, and the five-spoke shape appears only when all five measures qualify.