All Products
Building2025

OpenQuant India

Honest, open-source quant infrastructure for Indian markets. Your backtest is lying to you — OpenQuant fixes that.

GitHub
Open SourceQuant FinancePythonNSEBuilding in Public

The Problem

Indian retail quants run backtests that look like this:

"+22% CAGR. Sharpe 1.8. Strategy: 12-month momentum on Nifty 500."

Then they go live and bleed money.

The gap between gross and net is where retail quants get destroyed. A backtest that doesn't model STT, brokerage, exchange charges, GST, SEBI fees, stamp duty, STCG/LTCG, slippage, survivorship bias, and corporate actions isn't a backtest — it's a fantasy.

OpenQuant India shows you the right number, by default.

What It Is

A composable, pip-installable Python stack for Indian markets:

Package Status What it does
oq-core Alpha Shared primitives: Instrument, NSE TradingCalendar
oq-data Planned NSE/BSE bhavcopy ingestion, adjusted prices, point-in-time index universes
oq-backtest Planned Vectorized portfolio backtester with an honest Indian cost engine
oq-mcp Planned MCP server exposing data and backtests to LLM clients
oq-broker Planned Unified async broker abstraction (Kite / Upstox / Fyers / Dhan)

Each package is independently installable. Use one piece or all of them.

Quickstart

pip install oq-core

from datetime import date
from oq_core import Instrument, TradingCalendar

reliance = Instrument(symbol="RELIANCE", isin="INE002A01018")
# NSE:EQ:RELIANCE

cal = TradingCalendar()
cal.is_session(date(2024, 1, 26))          # False — Republic Day
cal.next_session(date(2024, 8, 14))         # date(2024, 8, 16)
cal.session_count(date(2024, 1, 1), date(2024, 12, 31))

Principles

  • Honesty over excitement — net-of-cost numbers are the default. Gross requires an explicit flag.
  • Data correctness is sacred — a wrong adjusted price is a P0 bug.
  • Paper-first, safety-default — live execution is opt-in, loud, and guarded.
  • Composable, not monolithic — separate packages with a shared core.
  • Community is the product — built in public, contributions welcome.

Roadmap

  • Phase 0 — Foundation (in progress): oq-core, repo scaffolding, CI, license.
  • Phase 1 — Data Layer: NSE bhavcopy ingestion, corporate actions, point-in-time index universes.
  • Phase 2 — Backtester: vectorized engine + honest cost engine + broker presets + tearsheet.
  • Phase 3 — MCP Server: data and backtests exposed to LLM clients (Claude Desktop, etc.).
  • Phase 4 — Execution Layer: paper + live broker abstraction with SEBI-2026 compliance built in.
  • Phase 5 — Ecosystem: docs site, strategy zoo, Discord, cohorts.

Licensed under Apache 2.0. Watch the repo to follow along.