Quant for breadth, fundamentals for depth
How we divide the labour between a systematic engine and fundamental judgement: data finds the opportunity, fundamentals build the conviction, and the whole loop is re-run every day.
Quantitative and fundamental investing are often presented as rivals. We treat them as two halves of one process, each doing the job it is best at. The systematic engine gives us breadth: it reads big data, machine-learning and alternative signals across every name in the universe, every day, without fatigue or favouritism. Fundamental analysis gives us depth: company quality, earnings and valuation, industry and competitive position, and management insight — the things a model can rank but cannot fully understand.
The engine is not a static model. Each day the data is refreshed, signals are re-ranked, model weights are updated, holdings are re-ordered under risk constraints, and the results of execution feed back into the next cycle. Research, holdings, execution and feedback form a closed loop; markets change every day, so the strategy learns every day.
Between the engine and the portfolio sits a fundamental fusion and review stage. Financial quality, industry trends, valuation and event context are used to filter or weight the quantitative signals, and a fund manager with more than fourteen years' experience reviews every algorithm-selected holding against the investment logic and the risk controls. Breadth proposes; depth disposes.
None of this is credible without honest history. Ordinary backtests see the revised past. Our research replays, as-of each decision date, exactly what was truly visible at the time: point-in-time data slices, bounded forward-fill, run-id lineage for every result, and walk-forward evaluation. A strategy that looks good under those rules has earned the right to be reviewed by a human — and then to be traded.
That is what we mean by certainty, traceability and sustainability in practice: a defined stock pool and trading plan, a per-trade evaluation system that can explain every position, and drawdown control that keeps the process alive through the periods when any single view is wrong.