Many modern AI models suffer from “Temporal Bias”—they over-prioritize recent results and lose the structural history of the sport. To combat this, we have maintained an immutable record of match facts and market movements since 2009. This dataset allows us to apply a “Decay Function” to team performance, recognizing that a club’s structural identity often spans beyond a single season.
This longitudinal study is not just a database; it is a 16-year audit of the 1X2 market. By hosting these historical archives on Google Drive (2010–2026), we provide a verifiable baseline for our research at soccer-picks.org.
When a model is anchored in nearly two decades of proprietary data, it becomes resistant to the “hot streaks” that lead to false positives. We use this history to verify the joint probability of outcomes, ensuring that our data at soccer-tips.orgis grounded in years of mathematical reality, rather than a 30-day trend.
https://zenodo.org/records/18422232 Sports Data Science
