NHL RESEARCH

Every game gets a Projected score. Each team's goals and shots for and against, opponent-adjusted and shrunk toward the league early in the year, set the two scoring rates; last season carries in until a team has a body of work. Every rating is computed as of that game's puck drop, using only games already played, so no projection is built with hindsight.

Only A+ or A is a play. A posted play carries the side, our win odds and the two or three real numbers that back it — team scoring and shot rates with the league rank, the starter's save rate and form, rest. A game that grades under A is shown as a projected matchup: the Projected starters and the Projected score, no play.

Projected starters until confirmed. Clubs name their goalie on game day, often at the morning skate. Until then each card carries the Projected starter — the probable from the same feed our research vault uses, or our rotation read — and flips him to Confirmedwhen the club does. Each starter is rated on goals saved above the league-average save rate per 60 minutes, on the season, over his last ten starts and split home and road. We tested that hard: switching the goalie term off made almost no difference to the model's accuracy in either held-out season — so the starter is on the card because he matters to the matchup, not because he is the secret sauce.

Rest and home ice. The second night of a back-to-back and long road trips cost real goals; home ice is worth a small, measured bump.

Posted game-day morning. The card goes up the morning of the slate and the plays are frozen from then on — a late goalie change updates the starter on the card, never the call. Each play and projection comes down at puck drop; from then on the game shows only where it stands — in progress, then the final score. The DraftKings number is shown for context only; no line went into any projection.

Analysis and entertainment, not betting advice and never a guaranteed outcome. 21+. 1-800-GAMBLER.