Serving tech enthusiasts for over 25 years. TechSpot means tech analysis and advice you can trust.

Editor’s take: A coach under pressure turned to a publicly available chatbot to search for fresh perspectives when conventional analysis and scouting reports failed to produce wins. It’s an unusual reflection of how accessible language models have become: a multimillion-dollar sports organization resorting to the same free online tool used daily for essays, code snippets, and recipes.

Calgary’s professional hockey team, the Flames, is facing one of the worst starts in franchise history. Fourteen games into the National Hockey League season, they have managed only three wins – a record that has left them buried near the bottom of the standings and searching for solutions. Head coach Ryan Huska’s latest idea to stop the slide came from an unexpected source: ChatGPT.

The revelation appeared in The Chase, a team-produced behind-the-scenes series that documents life around the club. During a recent team meeting, Huska admitted he had spent the previous night “going down a rabbit hole” with the AI chatbot. He explained to his players that he had fed the system a batch of data from their previous five games – from shooting percentages to shot volume and season projections – to make sense of the team’s scoring drought.

According to Huska, ChatGPT processed these inputs and estimated that if the trends continued, the Flames could average about 2.36 goals per game over the season. For context, that figure would rank among the NHL’s lowest-scoring teams. Huska framed the number as a challenge rather than a forecast, using it to urge players to focus on generating more scoring chances, even without a superstar goal scorer on the roster.

The hockey world saw another story beneath the motivational message. Analysts and technologists raised eyebrows when Huska used ChatGPT for statistical projections, since the system cannot perform numeric computation or predictive analytics. Unlike tools such as R, Python’s pandas, or commercial sports machine-learning frameworks, ChatGPT interprets prompts in natural language and predicts likely text completions. It can approximate reasoning but cannot verify numbers step by step.

In practical terms, if Huska entered player stats and asked for projections, ChatGPT would likely estimate using language-pattern correlations rather than processing the numbers mathematically. Such an approach can produce plausible-sounding results but may contain hidden logical or arithmetic errors. Hallucinations are well-documented in generative AI systems, and OpenAI has repeatedly warned against using ChatGPT for analytical or quantitative decision-making.

Many coaches across sports have already embraced data analytics, but those systems typically rely on vetted statistical models, sensor inputs, and machine learning explicitly trained for game analysis. By contrast, ChatGPT handles natural language better than numerical reasoning, making it an unconventional – and technically unreliable – choice for performance forecasting.