Hockey results Friday 9 October 2026: AI predictions review

Friday night's NHL slate produced six games and one of the more instructive evenings of the young season for the TuringStats consensus. The ten-model panel finished 4 of 6, a solid but unspectacular return that hinged almost entirely on two results nobody saw coming.

The headline surprise was St. Louis. All ten models backed the Blues at home against San Jose, and the Sharks won 3-2. Detroit's 6-3 home defeat to Seattle was the other miss, and it carried the same uncomfortable distinction: a unanimous panel, a wrong answer.

The night's results hinged on two unanimous calls that went the wrong way.

Where the AI got it right

The panel's most convincing call came in Calgary, where Calgary Flames vs Colorado Avalanche prediction went the way of all ten models. Colorado won 7-3, a wider margin than the most common projected score of 2-4, with average confidence at 60 per cent. The direction was never in doubt.

Vegas delivered a tighter but equally clean result. The Vegas Golden Knights vs Toronto Maple Leafs prediction was unanimous for the home side, and the Golden Knights edged it 4-3 — exactly the scoreline the panel had flagged as most likely, at 58 per cent average confidence.

Washington was the panel's split decision of the night, and it landed. Only 50 per cent of models backed the Capitals, with a 4-5 scoreline the most common projection. The actual result was a 3-1 Capitals win, proving the majority direction right even if the scoreline was not.

Columbus rounded out the correct calls. The Columbus Blue Jackets vs Pittsburgh Penguins prediction drew 70 per cent support for the Blue Jackets at 57 per cent average confidence, and Columbus won 3-2. The most common projected score of 4-3 was close without being exact.

Where it got it wrong

St. Louis was the night's clearest failure of consensus. All ten models picked the Blues to beat San Jose, with a 3-2 scoreline the most popular projection and average confidence at 57 per cent. The Sharks won 3-2 — the same score, the wrong winner. A unanimous panel producing the exact scoreline for the losing side is the kind of result that makes the pre-match percentages look less like certainty and more like a coin flip wearing a suit.

Detroit was the second unanimous miss. Every model backed the Red Wings at home against Seattle, most commonly at 4-2, with average confidence at 58 per cent. The Kraken won 6-3. Not a near miss, not a late collapse — a four-goal swing against a panel that had no reservations whatsoever.

Models of the day

Five models shared the best return of the evening. Claude Sonnet 5, DeepSeek V4 Pro, Qwen 3.8 Max, MiMo V2.5 Pro and Grok 4.6 each finished 4 of 6, matching the consensus record exactly. None of them found a way through the St. Louis or Detroit traps, but none of them fell into any additional ones either.

At the other end, three models finished 2 of 6: Mistral Medium 3.5, Gemini 3.8 Flash and Kimi K3. Each of them missed both unanimous calls plus one further game, leaving them two games adrift of the leading group. GPT-5.6 Luna and GLM 5.3 sat in between at 3 of 6.

Full results

• NHL: St. Louis Blues 2–3 San Jose Sharks — consensus St. Louis Blues win (100%) WRONG
• NHL: Calgary Flames 3–7 Colorado Avalanche — consensus Colorado Avalanche win (100%) RIGHT
• NHL: Vegas Golden Knights 4–3 Toronto Maple Leafs — consensus Vegas Golden Knights win (100%) RIGHT
• NHL: Washington Capitals 3–1 New York Rangers — consensus Washington Capitals win (50%) RIGHT
• NHL: Detroit Red Wings 3–6 Seattle Kraken — consensus Detroit Red Wings win (100%) WRONG
• NHL: Columbus Blue Jackets 3–2 Pittsburgh Penguins — consensus Columbus Blue Jackets win (70%) RIGHT

Saturday's slate is already processed. The full model-by-model breakdown for every game on the schedule 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.