NHL Projections

Create my rankings ⓘ

Customize the real scoring and roster rules below to match your own league — every ranking, VOR, and Auction $ on this page recomputes live from whatever's set here.

Categories league ranks your team across a chosen set of individual stats (the categories below), same as a real roto/head-to-head categories league. Points league combines every stat into one weighted total instead -- pick whichever matches your real league's actual format.

Moves rankings, VOR, and Auction $ (both league types) toward each player's real floor or ceiling instead of the projected midpoint, using the real spread across independent projection sources. A player without enough independent sources to measure a real spread is unaffected.

Sleepers & Busts ⓘ

Diff = ADP rank minus system rank under your current scoring settings. A sleeper's system rank is much better than where the market is drafting them; a bust's is much worse. Shown here: anyone the market and the model disagree on by a full round or more (one round = your league size) — a gap of a spot or two isn't a real signal, just where two reasonable rankings happen to differ.

How accurate is each system? ⓘ

Jackdaw: strengths & weaknesses ⓘ

Rebuilt 2026-09-12 from this site's own live numbers (nhl_category_rows()'s skill_vs_average, pooled 2020-2025) instead of hockeystats' own separate tool, which had computed every prior version of this panel. Real reason: hockeystats confirmed their own tool has a bug (`_score_season()` intersects the scored population against every registered source, so one narrow-coverage source silently caps everyone else's real population too — e.g. ESPN's ~260-skater universe capping Jackdaw's real ~675-skater reach), understating real skill by roughly 20-25% wherever it runs, and said a proper fix has no timeline on their side. This site's own accuracy pipeline was never subject to that specific bug (a separate computation, fixed to true leave-one-out 2026-09-08) — so rather than keep showing numbers known to be wrong, this panel now sources itself from what's already correct and already live here. Same real methodology as before: margin over the mean error across every scored system in that cell (leave-one-out, so Jackdaw is never scored partly against its own opinion) — a much harder bar than beating any one competitor, since the field average includes several skilled human analysts, not one naive model. Scoped to Jackdaw itself specifically, same as MLB/NBA/NFL's own panels — not the published Magpie blend, which can and does score some of these stats differently. + / ++ / +++ and − / −− / −−− — each tier is the number's rank within its own third of every positive (or negative) value on this page, not a fixed cutoff. Raw Saves and raw Goals Against aren't scored here: this site's own pipeline only scores the real per-60-minutes/shots-against RATE versions (GAA, SV%), deliberately, not the raw counts (which conflate accuracy with how much a goalie played).