JOSHUA LIND
Model record / V1–V3 V3 — Build 13
Back to the NHL Value Lab

NHL Value Lab / Release history

LAB
CHANGELOG

Significant model releases, features, and upgrades that shaped the NHL Value Lab. Dates mark published model, feature, and policy changes; daily data refreshes do not create new model releases.

Full Strength reflects players who can resume playing

  • Reviewed ended careers and medical cases where further play is not expected remain outside Current Lineup, Full Strength, future returns, and trade scenarios, even when a contract remains active.
  • Contract affiliation, statistics, and ratings are retained. Estimated injury dates cannot remove a career hold; affirmative comeback evidence requires a new review.

Timetables inform the whole league’s outlook

  • Current ESPN return estimates, with CBS fallback coverage, become provisional working dates. Illness can use the same sourced dates; missing or expired timetables retain absence until updated.
  • The daily updater notices date changes independently of injury labels and processes evidence in bounded, resumable steps. Current Lineup and Full Strength stay distinct; historical forecasts remain frozen.

Short-term injuries no longer mean a season-long absence

  • Verified day-to-day and week-to-week injury labels inform provisional return distributions learned from historical reports and actual NHL appearances. Simulations can restore players later without clearing today’s injury designation.
  • Current Lineup and Full Strength rankings stay separate from future Points Outlook and playoff chances. Unsupported labels and overlapping absence reasons retain the existing treatment; historical forecasts remain frozen.

Compare the roster and the possibilities

  • Current Lineup and Full Strength rankings distinguish today’s available lineup from the same organization with every eligible player available. These roster rankings stay separate from NHL standings and Points Outlook.
  • The Trade Simulator defaults to Full Strength, with an optional Current Lineup view and 500 matched seasons. OUT labels appear only in Current Lineup; shared trade links retain the selected view.

Verified signings and trades enter the Lab

  • The daily updater can discover missing players through official signings and trades, verify identity and contract coverage, and apply the existing historical-rating and availability rules. Unresolved cases remain visible for review.

Smaller, recoverable update steps

  • Reduced memory pressure and repeated processing across source collection, validation, and forecast finalization. Completed work is saved between stages, while publication still requires the existing data, model, and precision checks.

Confirmed absences change the lineup

  • Injured, suspended, and personal-leave players retain their ratings and appear in separate roster sections. Eligible replacements enter the Current Lineup; each forecast uses its saved availability snapshot.
  • Healthy scratches and DTD/GTD labels alone do not exclude a player. A confirmed absence requires affirmative return evidence to clear.

Broader historical ratings and depth coverage

  • Added a consistent activation path through archived ratings and recent NHL history, reconstructed Max14 ratings for returning veterans, and support for known depth players with complete current NHL statistics.
  • True newcomers can become rated after their first NHL game; zero-game newcomers remain NR. Ratings and contract eligibility stay separate. The unpublished Build 3 candidate was superseded by corrected Build 4 coverage.

A daily-updated, player-driven model

  • Extended historical player projections and combined them with current-season K40 evidence in one unified team model. The October 2 launch was refined on October 3 to use nine weighted player metrics and seven public team dimensions.
  • Projected lineups use 12 forwards and six defencemen. Completed results stay fixed while the remaining schedule is simulated, and frozen forecasts remain available for accuracy evaluation.

Explore roster changes

  • Introduced hypothetical trades with rebuilt lineups and before/after team outlooks, kept separate from official forecasts and accuracy records.

Measure forecasts against results

  • Added team and player accuracy evaluation against frozen, dated predictions, with separate forecast archives and owner navigation.

Simulate a shared league schedule

  • Replaced disconnected team forecasts with coherent standings and playoff simulations, then refined Points Outlook presentation while preserving relative team strength.

Test the model against past seasons

  • Introduced metric-specific historical weighting, small-sample protection, age adjustments, and roster replacement, supported by chronological model testing.

Player value, team strength, and comparisons

  • Established player-driven team views, positional rankings, filters, comparisons, and a published methodology—the foundation for the later live model.