University of Toronto Mississauga
Two self-directed research projects, written up in full. Both are reports on method rather than on results: one documents a screening system and the errors found while building it, the other documents a trading thesis that was killed by a rule written before the data arrived.
Design and deployment of a full-market screening system covering every listed US company above USD 1.5bn market cap. Sets out why the quality grade is built on return on invested capital rather than margin thresholds, how the nine-tier classification is assigned, and six documented errors with their corrections, including one case where the verification method itself was the source of the fault.
Chinese version: 中文版 · 23 页
A nineteen-day retrospective on an automated research system for prediction markets. Every kill criterion was pre-registered as executable code before its sample filled, four strategies were terminated on those lines, and the project closed under a cost ceiling that held. Reports a negative result and the decision rules that outlasted it.