Building a multi-market equity backtesting & research platform

What this is

An in-house research project, not a client engagement and not a product we sell. We build serious research infrastructure, and projects like this are how we keep that engineering sharp. It’s representative of the backtesting and research systems we build for clients.

What we built

A single Python platform that backtests equity strategies across two very different markets, Indian (NSE) and US, from one codebase, with the market-specific handling each needs.

• A sector-ranking research pipeline. A four-factor composite, momentum, relative strength versus the benchmark, risk-adjusted return, and a technical score, that ranks sectors so research only looks for candidates in the strongest ones.
• A volume-based signal module. Research into whether unusual trading volume can flag accumulation by larger market participants (described at the research-question level, the logic itself stays in-house).
• Portfolio & risk controls. Position sizing drawn from current equity rather than a fixed number; per-market-cap allocation buckets; a hard cap so no single sector is ever over-allocated; and a risk-reward check that must be in favour before any position opens.
• Serious data engineering. Corporate-action-adjusted price series so a stock split can’t create a phantom “cheap entry”; per-stock eligibility gating; and alignment to each market’s own benchmark trading calendar.

The engineering rigour, the real point

• Auditable and reproducible. Indicators pre-computed and close-based so there’s no look-ahead, and a full audit trail, per trade, per day, per year, per symbol, so any result can be traced back to the chart.
• AI-assisted, with a hard integrity guard. We use AI tooling to surface areas worth improving, then modify and re-run, but under a strict rule that the actual entry and exit logic is never silently changed or “hallucinated”. A sample of trades is checked against the charts every time, to confirm the system does exactly what the written rules say and nothing else.
• Code-quality discipline. Modular architecture, deterministic rounding, reproducible runs.

This is an engineering and research exercise. It is not a product, not a signal service, not investment advice, and we make no claim about how it performs.

It is the same discipline behind our backtesting software and quantitative research work.

← All case studies