AL- 0-2 AL- · 60% AL- 2-0 ALD · 60% IFE 2-1 MAL · 53% VAS 2-0 HAL · 62% RAN 1-2 BRO · 46% ALA 1-0 KOC · 50% GOA 2-1 TEL · 54% AL- 1-2 AL- · 43% VFB 2-0 BOR · 62% STA 2-1 NIC · 50% TOR 2-1 MON · 52% KOR 0-1 KVM · 52% IPS 2-1 NOT · 47% ALA 1-0 MAL · 60% AL- 0-2 AL- · 60% AL- 2-0 ALD · 60% IFE 2-1 MAL · 53% VAS 2-0 HAL · 62% RAN 1-2 BRO · 46% ALA 1-0 KOC · 50% GOA 2-1 TEL · 54% AL- 1-2 AL- · 43% VFB 2-0 BOR · 62% STA 2-1 NIC · 50% TOR 2-1 MON · 52% KOR 0-1 KVM · 52% IPS 2-1 NOT · 47% ALA 1-0 MAL · 60%
Thursday 22 October 2026

AI NHL picks yesterday

Yesterday's record: the AI consensus picked the right result in 4 of 5 games (80%). We publish the misses too, and the AI leaderboard grades every model.

  1. ✓ Result right

    Final score: Toronto Maple Leafs 5–4 New York Rangers. Before kickoff, 10 of 10 AI models pick Toronto Maple Leafs to win (100%); the consensus was right. The AI predicted score was 4–3.

    Full breakdown, form and head-to-head →
  2. ✓ Result right

    Final score: Pittsburgh Penguins 4–0 Florida Panthers. Before kickoff, 8 of 10 AI models pick Pittsburgh Penguins to win (80%); the consensus was right. The AI predicted score was 4–3.

    Full breakdown, form and head-to-head →
  3. ✓ Result right

    Final score: Boston Bruins 6–0 Nashville Predators. Before kickoff, 9 of 10 AI models pick Boston Bruins to win (90%); the consensus was right. The AI predicted score was 3–2.

    Full breakdown, form and head-to-head →
  4. ✓ Result right

    Final score: New Jersey Devils 4–1 Anaheim Ducks. Before kickoff, 10 of 10 AI models pick New Jersey Devils to win (100%); the consensus was right. The AI predicted score was 4–2.

    Full breakdown, form and head-to-head →
  5. ✗ Result wrong

    Final score: Ottawa Senators 2–6 San Jose Sharks. Before kickoff, 10 of 10 AI models pick Ottawa Senators to win (100%); the consensus was wrong. The AI predicted score was 4–2.

    Full breakdown, form and head-to-head →
  6. 9 of 10 AI models pick Los Angeles Kings to win (90%). The mean predicted score is 3–4 at 59% average confidence (medium).

    Full breakdown, form and head-to-head →

How these NHL computer picks are made

Every fixture on this page was run through the same panel of independent AI models. Each model returns a winner and a scoreline; we publish the vote split, the mean predicted score and the average confidence, then grade every model once the result is in. Open any match for recent form, head-to-head history and the full model table, or ask the council your own question.

Think of them as computer picks built from language models instead of a single formula: ten AI models read the same form, injuries and head-to-head data and pick each game straight up. When they all agree, the pick has historically been far more reliable than a split vote. Every model's record is public on the AI prediction leaderboard. Not betting advice.