Kristian Fagerlie contrasts a chatbot's confident guess with a structured forecasting workflow that gathers data, calculates probabilities and compares results against market pricesEvaluation measures how well an AI system performs against defined tasks, criteria and failure conditions using repeatable evidence.. His football example illustrates margin removal, data cleaning and ratings based on historical play-by-play information.
Kristian Fagerlie uses a coding model to build and run a forecastCode generation uses AI or another automated system to create source code from instructions, examples, schemas, or higher-level specifications., then creates an HTML overview of the results. He reports that his historical backtestBacktesting applies a strategy to historical data to estimate how it would have behaved under past market conditions. performed worse than the market and shows a no-position outcome at the quoted price, emphasizing that a model prediction is not proof of an exploitable advantage.
Kristian Fagerlie discusses fees, predeclared decision rules, sizing and asking why the market might be right while the model is wrong. The demonstration is an educational example with uncertain assumptions, not independently validated evidence of profitability or a recommendation to place a trade.
Watch on YouTube
