What a Win Costs on the 2026-27 NHL Market

Cap-efficiency study on the projected 2026-27 rosters (projected_rosters_2026_27_v2.csv, 640 slots) priced with the PuckPedia 2026-07-06 cap scrape (cap_hits.csv).

Definitions

• Total projected wins/82 — skaters: ProjValue82 (relWPA wins/82) + (GAx82 + PenW82)/6; goalies: ProjValue82/6 (GSAx is in goals per 82 starts). The standard conversion 1 win ≈ 6 goals is used for the goal-denominated components; the season simulator's stage-1 actually maps these through a fitted regression, but 6 g/w is the conventional exchange rate and is what these dollar figures assume.
• Availability-weighted WAR = (wins/82 − replacement) × ExpGPFrac: when a player sits, his slot produces replacement-level value, so only games played earn the premium. For goalies ExpGPFrac is the expected starts share, which converts per-82-starts GSAx to a team-season contribution.
• Replacement level = 25th percentile of rostered wins/82 per position group — the marginal 13th forward / 7th defenseman / 3rd goalie is by construction around the bottom quartile of the 640 rostered slots. Result: F -0.21, D -0.53, G -0.12 wins/82. The projection's NODATA level for skaters (−1.5 goals ≈ −0.25 wins) sits close to these, which is the cross-check that the quartile is measuring 'freely available' rather than something stricter; the quartile is preferred because it prices F/D/G separately from the actual 2026-27 talent pool.
• League minimum = $850k (2026-27). Market rate and surplus both work in cap dollars above the minimum, since a roster body costs the minimum regardless of quality. With that convention, non-ELC surplus sums to $0 league-wide by identity.
• ELC proxy — cap hit ≤ $1.00M and age < 25 (or a 2026 draftee with no bio row). 130 of 640 slots. ELCs are excluded from the market-rate fit (their price is CBA-set, not market-set) but included in the surplus rankings.
• Term is not in the dataset (single-season cap-hit scrape), so contract-length context in the commentary below is inferred from public knowledge, not computed.

The market price of a win

$5.37M per win above replacement (non-ELC fit: $1,796.39M of above-minimum cap buys 334.6 WAR across 510 contracts).

Sanity: including ELCs the league average is $5.16M/win. Implied team check — a cap-ceiling team ($104.00M) has ~$84.45M above 23 minimums; at $5.37M/win that buys ~16 wins above a replacement-level team. With a median team at ~46 wins, that puts the all-replacement team at ~35 wins (avg roster holds 10.9 WAR).

That is above the naive $2-4M/win folk range, for two identifiable reasons. (1) Universe: only the 20 projected roster slots are priced — $73.37M of cap per team; the remaining ~$30.63M (scratches, buried deals, LTIR, bonus overages) buys ~zero WAR, so a full-cap accounting would price a win higher, not lower. (2) Compression: the shrunk projections put the all-replacement team at ~35 wins vs the conventional ~21; relWPA-based WAR understates true spread (the simulator's stage-1 regression re-inflates it), which mechanically inflates $/win. The number is robust to the replacement-level choice: refitting with the projection's NODATA levels (−0.25 wins skaters, −1.33 goalies) moves it only to $5.21M/win. Treat $5.37M as the within-model exchange rate — rankings and team totals, which is what this study is for, are insensitive to it.

Top 15 surplus contracts (best value)

Term context (public knowledge, not in the scrape): McDavid's $12.5M is the famously short 2-year bridge (through 2027-28) — maximal surplus, minimal security. Draisaitl ($14M through 2033) and Bouchard ($10.5M through 2029) make Edmonton's top-heavy core the single biggest pile of locked-in value; Aho (through 2032) and Jarvis (through 2033) do the same for Carolina on longer horizons. Hyman at $5.50M and Zuccarello at $1.00M are expiring-vintage deals whose value cannot be re-signed at this price. Note the list is all ages 25-39 on market deals — the biggest stars out-earn even max-cap-share contracts at this model's exchange rate.

Bottom 15 (the albatrosses)

Term context: the classic aging-UFA albatrosses are here — Nurse ($9.25M through 2030), Hronek ($7.25M through 2032), Pionk (through 2031), Lindholm (through 2031), Huberdeau (through 2031) — long tails on declining players. Krug, Couture and Seguin are at least expiring (final year 2026-27). A second, different failure mode also shows up: brand-new second contracts that pre-pay unproven production — Coronato ($6.5M x 7) and Nazar ($6.6M x 7) start this season, and Carlsson's $18M and Power's $8.35M are priced far above their current projection; these are bets on development, not sunk costs, and the aging curve is on their side (both bands still improving) unlike the UFA group.

Team cap efficiency (total surplus)

Age bands: what teams pay vs what they get next

The aging column is the adopted within-player curve (age_curve_relwpa.csv) difference from this season's age to next (goalie GSAx deltas ÷ 6). Read the two rightmost columns together: $/win rises monotonically from 25-28 onward while the expected aging delta deteriorates — teams pay the most per win exactly for the players who will be worth less next season, the classic pay-for-past-production signature, and the 32+ band alone is a nine-figure sinkhole of negative surplus. One nuance: 21-24 market (post-ELC) deals are not cheap either — second contracts price projected future production, so the only systematically underpaid cohorts are the price-controlled ELCs and the 25-28 prime band.

The ELC subsidy

Entry-level deals hold $72.94M of total surplus (130 slots, avg $561k).