The presenter configures Kimi K3 through OpenCode and gives it Polymarket API data for sporting-event markets. The workflow asks the model to inspect prices and historical inputs, calculate expected value and propose a small portfolio of possible outcomes.
A previous experiment produced two unusually large percentage returns, prompting a second test on a World Cup final. The model proposes several outcomes, including low-probability score lines, and assigns estimated value based on the supplied market data.
The presenter does not place the proposed trades and treats the exercise as an exploratory model test. A single profitable event, a model-generated estimate and a short backtest do not establish predictive reliability or a repeatable financial strategy. Closing channel promotion is omitted.
Watch on YouTube
