Fantasy Hockey Analytics: Key Metrics and Data Points
Fantasy hockey analytics sits at the intersection of on-ice performance measurement and roster construction strategy, drawing on a substantially different statistical vocabulary than the other major North American sports. The metrics that drive value in NHL-based fantasy formats — Corsi, Fenwick, expected goals, and zone deployment data — are largely invisible to fans who rely on traditional boxscore stats alone. This page defines the core analytical categories, explains how they are calculated and applied, and identifies the decision boundaries where one metric class outperforms another in specific roster contexts.
Definition and scope
Fantasy hockey analytics encompasses the collection, processing, and application of NHL player performance data for the purpose of predicting future fantasy point output. The scope spans two primary format types: season-long rotisserie (roto) and head-to-head (H2H) leagues, plus daily fantasy sports (DFS) contests governed by contest rules from operators licensed under individual state frameworks. The distinction matters analytically because roto formats reward cumulative category totals across a full season, while H2H formats introduce weekly variance as a critical variable.
The public data ecosystem for NHL analytics is anchored by two primary non-commercial sources. Natural Stat Trick (naturalstattrick.com) provides line-level and player-level shot-quality data updated after each game. Evolving Hockey (evolving-hockey.com) publishes goals-above-replacement (GAR) and expected goals models derived from play-by-play data released by the NHL under its public data feed. The NHL itself maintains an official stats portal at NHL.com/stats, which serves as the baseline record for counting stats.
For a broader orientation on the sport-agnostic framework governing how fantasy analytics data is sourced and classified, the Fantasy Analytics Authority home organizes the full analytical domain.
How it works
NHL fantasy scoring systems assign point values to counting categories — goals, assists, shots on goal, hits, blocks, plus/minus, penalty minutes, and goaltender-specific stats (wins, saves, goals-against average, save percentage). Analytics-driven roster construction works backward from those categories to identify which underlying on-ice metrics are predictive of future category production.
The core analytical pipeline involves four discrete phases:
- Raw event data ingestion — Play-by-play files are pulled from the NHL public API, which logs every shot attempt, goal, penalty, and faceoff with zone location, time, and score state embedded.
- Shot-model scoring — Each unblocked shot attempt is assigned an expected goal (xG) value based on shot location, shot type, and traffic indicators. Models from Evolving Hockey and MoneyPuck (moneypuck.com) are the two most-cited public implementations.
- Rate-stat normalization — Raw totals are normalized to per-60-minutes rates to control for ice time variation. A player logging 14 minutes per game and a player logging 22 minutes per game are not directly comparable on counting stats alone.
- Context adjustment — Metrics are filtered by score state (5v5 close, power play, shorthanded) and zone start percentage (offensive zone vs. defensive zone draw percentage) to isolate skill from deployment effects.
The most widely used possession metrics in this pipeline are Corsi For % (CF%), defined as shot attempts for divided by total shot attempts while the player is on ice, and Fenwick For % (FF%), which excludes blocked shots from the denominator. At 5v5 score-close situations, both metrics show meaningful correlation with future goal-scoring rates over samples larger than approximately 500 minutes of ice time (Evolving Hockey methodology documentation).
Common scenarios
Identifying high-value power play contributors — Power play time on ice (PP TOI) is the single strongest predictor of power play point production. A forward averaging 3.2 PP minutes per game on a top unit with an xG rate above 0.30 per 60 PP minutes represents a substantially more reliable category asset than a player with comparable ES production but 0.8 PP minutes.
Goaltender streaming in H2H formats — Goaltenders in DFS and weekly H2H leagues are evaluated on projected save count (shots faced) and quality-start probability. Quality-start rate, tracked publicly by sites including Corsica Hockey, requires a save percentage of at least .917 on any given night to qualify. Two goaltenders with identical win rates may have quality-start rates that differ by 15 percentage points or more depending on team defense and shot volume allowed.
Shot-heavy defensemen — Blueliners ranked in the top 20 by shots on goal per game consistently outperform their draft position in points-per-game leagues. The 2022–23 NHL season saw defensemen in the top-20 shots-on-goal tier average approximately 42 fantasy points per game more than mid-tier blueliners in standard ESPN scoring (Natural Stat Trick season summary data).
For context on how these hockey-specific metrics compare structurally to metrics in other sports, advanced statistics in fantasy sports provides a cross-sport reference framework.
Decision boundaries
The critical analytical decision in fantasy hockey is selecting the right metric class for the right time horizon and format. The following contrast defines three primary boundary conditions:
Corsi/Fenwick vs. goal-based stats — Over short samples (fewer than 20 games), goal-based stats exhibit high variance driven by shooting percentage fluctuation. Corsi and Fenwick stabilize faster and are preferred for early-season and in-season trade evaluation. Over full-season samples, both converge.
xG vs. actual goals — Players persistently outperforming their xG model (high PDO: shooting % + save % combined above 1.000) are candidates for negative regression. Players underperforming xG are buy candidates. This is among the most actionable boundaries in waiver wire and trade contexts.
Rate stats vs. counting stats — In roto formats, counting stats determine final standings; rate stats are used only to project future counting-stat volume. In DFS, where roster slots are fixed and salary cap constraints apply, rate stats per dollar of salary determine lineup construction efficiency.
The regulatory framing that governs how fantasy operators present and apply this data — including state-level skill-game classification standards — is documented at regulatory context for fantasy analytics.