https://ajmclab.com/Recent content in Home on The Decision LabHugoenThu, 30 Jul 2026 00:00:00 +0000https://ajmclab.com/about/Mon, 01 Jan 0001 00:00:00 +0000https://ajmclab.com/about/<p>This is a personal notebook, currently focused on a single project: an open, reproducible <strong>FPL decision laboratory</strong> — comparing deterministic analytics, mathematical optimisation and AI-agent orchestration on point-in-time evidence, with auditable outcomes.</p> <p>Posts are written by me. Tools assist with drafts and diagrams; they do not get a vote. Diagrams are drawn as wireframes in <a href="https://tldraw.dev">tldraw</a> and embedded as SVG.</p> <p>The project lives in the <a href="https://github.com/">FPL repository</a> </p>https://ajmclab.com/data-and-confidence/Thu, 30 Jul 2026 00:00:00 +0000https://ajmclab.com/data-and-confidence/<h2 id="1-the-set-up">1. The Set Up</h2> <figure> <a href="https://ajmclab.com/diagrams/data-confidence-setup-light.svg" title="Open full size" class="diagram-light"> <img src="https://ajmclab.com/diagrams/data-confidence-setup-light.svg" alt="Decision lab quick overview: source registry and unstructured evidence feed capture and reconcile, then engine output, ending in a Gameweek Decision Record." loading="lazy"> </a> <a href="https://ajmclab.com/diagrams/data-confidence-setup-dark.svg" title="Open full size" class="diagram-dark" aria-hidden="true"> <img src="https://ajmclab.com/diagrams/data-confidence-setup-dark.svg" alt="" loading="lazy"> </a> <figcaption><small>Lab at a glance — registry and evidence into capture → reconcile → engine → Gameweek Decision Record.</small></figcaption> </figure> <p>This is an overview of what the data engine currently looks like, what it takes as inputs, a map of those inputs, how we manage the data and how it is dated, how much is trusted.</p>https://ajmclab.com/why-an-fpl-decision-lab/Wed, 29 Jul 2026 00:00:00 +0000https://ajmclab.com/why-an-fpl-decision-lab/<p>This site will document the building of an <strong>agentic decision laboratory</strong>: an open, reproducible environment that compares deterministic analytics, mathematical optimisation, single-agent reasoning and multi-agent orchestration, using point-in-time evidence and auditable outcomes.</p> <p>The test environment is Fantasy Premier League. That choice is deliberate as a test sandbox for pulling together analysis of structured and unstructured data, using an algorithmic engine, and then overlaying AI and agentic models to interpret, in particular, the unstructured data, and align it with the structured data outputs. This stems from my own work, reframed into an entertaining format with some competition and clear outcomes!</p>