Hockey results Monday 19 October 2026: AI predictions review

The ten-model consensus finished Monday 19 October 2026 with a record of 2 of 4 across four NHL fixtures. Two favourites delivered as expected, and two went the other way in games the models had backed heavily.

The biggest surprise came in Calgary, where the Flames beat the Carolina Hurricanes 3-2 despite 90% of the ten models siding with Carolina. Philadelphia's 3-2 win over Colorado was the second upset, also against a 90% consensus. The night's sharpest call was Toronto over San Jose, where 80% of models landed on the right side.

Full-time ice: two 90% consensus calls went the wrong way on Monday night.

Where the AI got it right

New York Rangers vs Anaheim Ducks prediction: the Rangers won 5-2, but this was the least settled call of the night. Only 50% of the ten models backed New York, with the most common scoreline sitting at 4-3 and average confidence at 58%. The correct winner was a split-consensus pick rather than a firm one, and the models got the direction right while undershooting the margin.

Toronto Maple Leafs vs San Jose Sharks prediction: Toronto's 3-1 win was the consensus's most comfortable correct result. Eight of ten models picked the Maple Leafs, with a most common score of 4-2 and average confidence of 58%. The winner was right; the goal count ran one short on each side of the projection.

Where it got it wrong

Calgary Flames vs Carolina Hurricanes prediction: nine of ten models backed Carolina, with a most common score of 3-4 and average confidence of 58%. Calgary won 3-2 instead. This was the night's clearest miss by weight of consensus, and the models were on the wrong side of a one-goal game they had read as a Carolina win.

Philadelphia Flyers vs Colorado Avalanche prediction: the same 90% split favoured Colorado, with a most common score of 2-4 and average confidence of 59% — the highest average confidence of any match on the card. Philadelphia won 3-2. The models had the projected margin close, but pointed it the wrong way, and the highest-confidence call of the day was one of the two that failed.

Across the four games, average confidence sat between 58% and 59% throughout. The consensus was never emphatic in numerical terms, but the two 90% splits carry the weight of the record: on the two games where nine of ten models agreed, the group went 0 for 2.

Models of the day

Claude Sonnet 5 was the only model to finish above .500, going 3 of 4 correct. No other model in the field of ten matched that tally.

Five models finished at 2 of 4: GPT-5.6 Luna, DeepSeek V4 Pro, Gemini 3.8 Flash, GLM 5.3 and Qwen 3.8 Max.

Four models finished at 1 of 4: MiMo V2.5 Pro, Grok 4.6, Kimi K3 and Mistral Medium 3.5.

Mistral Medium 3.5 was the worst performer of the day, going 0 of 4 and missing every result on the card, including both games the consensus got right.

Full results

• NHL: Calgary Flames 3–2 Carolina Hurricanes — consensus Carolina Hurricanes win (90%) WRONG
• NHL: New York Rangers 5–2 Anaheim Ducks — consensus New York Rangers win (50%) RIGHT
• NHL: Philadelphia Flyers 3–2 Colorado Avalanche — consensus Colorado Avalanche win (90%) WRONG
• NHL: Toronto Maple Leafs 3–1 San Jose Sharks — consensus Toronto Maple Leafs win (80%) RIGHT

The full slate for the next matchday, with every model's pick and confidence level, is available on the today's hockey predictions page.

Next up: today's hockey predictions from the same panel, and the results archive with every graded call. Forecasts are informational only.