RD-AGENT / FACTOR LOOP

The problem

Quant research is not just drawing an equity curve. The hard part is asking a better question, testing the hypothesis, and recording why an experiment was kept or discarded.

What it does

• Switch between momentum, quality, and mean-reversion hypotheses.
• Follow the state of hypothesis, factor, backtest, and feedback stages.
• Compare return, Sharpe, maximum drawdown, and benchmark performance.
• Play the next fixed iteration and watch the research record evolve.

Technical direction

The interaction follows Microsoft RD-Agent’s quant research scenarios and borrows the research-pipeline idea from Qlib. Everything runs on deterministic browser fixtures, with no live-market connection.

Outcome

It turns “I made a strategy” into “I know how to propose, test, and iterate on a strategy”—a workflow closer to real product and research engineering.