Project H
An independent quant research engine · live since day one

A systematic edge in Indian markets, built in the open.

Project His an independent research engine for Indian F&O — dozens of systematic strategies, forward-tested on paper every session. The goal: find durable, repeatable edge, and prove every claim in the open before a rupee is ever at risk.

Paper-traded forward tests · no real capital · research, not investment advice.

Play with the engine
illustrative

Set a scenario — watch the signals become a view.

News
Volatility
Trend
the engine leansCALL
conviction100%
risk budget~9% of book · defined-risk
News +2Trend +1Volatility ×1.0

News and trend agree, and options are cheap — so it leans in harder.

Illustrative only — a toy of how the engine weighs signals into a view. Not the real model, not a prediction, not investment advice.

How it thinks

A disciplined stack, held to an honest bar.

The edge isn't any single model — it's the discipline of combining many independent signals, refusing to trust what can't be measured, and letting risk lead. The specifics stay private; the principles don't.

Signalsmany, independentEnsembleranked into one viewRisk gatesized · capped · vetoedPaper bookreal prices, no capitalScoreboardpublic, every day

Many weak signals, one view

No single model gets to decide. Independent, loosely-correlated signals are combined into one conviction — diversification at the level of ideas, not just names.

Measured, never trusted

Nothing earns weight until it survives out-of-sample, corrected for how many things were tried. A good-looking backtest is a hypothesis, not a result.

Risk before return

Position sizing, stop discipline and defined-risk structures come first. The P&L is what's left after the risk has been respected.

News as signal, not noise

The news cycle is read for what actually moves price — filtered for materiality and novelty, attributed to the right name, and discounted when the market has already priced the move.

In the open

The scoreboard is public. Most aren't.

Anyone can claim a strategy works. This project publishes its paper scoreboard live instead — every experiment, marked against real market prices, refreshed through the session. The evidence accrues in public or it doesn't count.

01

Every run reports

No cherry-picked screenshots. The live board lists every experiment, its capital and its P&L — including the ones having a bad day.

02

Paper until proven

No real capital is deployed. Every strategy trades a simulated book marked against real market prices, and has to earn its keep out-of-sample before anything changes.

03

Research, not advice

This is the public log of a forward-testing experiment — not a signal service, not a tipsheet, and never investment advice.

The scoreboard, in the open

The structure is public. The P&L stays private.

Here's the shape of the live book — how many strategies are under test, how long it's been running, what's been retired, and how many have earned real capital. No rupee figures here; the live P&L sits behind a login.

loading the public scoreboard…
Research log

Latest entry

Cover image for Ranking, not classification — framing the ML problem so it can be honest
Latest·

Ranking, not classification — framing the ML problem so it can be honest

The most consequential modelling decision in Project H isn't an architecture or a hyperparameter — it's the question the model is asked. 'Will this stock go up?' is the wrong question. 'Which names are likely to outperform their peers?' is the right one. Why the framing matters, and why honest validation matters more.

Harish Subramanian
Harish Subramanian
About

One person, one honest experiment.

Project H is an independent research engine for the Indian F&O market, built and run end-to-end by a single developer. Machine learning, a news-intelligence layer and event-driven mechanics feed one systematic, risk-led book — and every idea is forward-tested on paper, marked against real market prices, with the scoreboard published in the open.

There's no fund here, no product, no signal service. Just a narrow question, tested honestly: can a disciplined, systematic approach to Indian F&O hold up out-of-sample — and can the working be shown in public, wins and losses alike? When something doesn't earn its keep, it stays on the board anyway.

Universe
Indian F&O
Capital
Paper only
Cadence
Every session
Team
Built solo