Strategy library

Start with a question. A screen finds matching stocks; a strategy adds portfolio rules and needs evidence after costs. Being implemented does not establish a profitable edge.

Three kinds of strategy live here. 🧑 means the founder thought of it. 🏠 means we designed it in-house. 📚 means it is a public, textbook idea represented by the implementation described here. A familiar name does not establish exact reproduction, complete data or a profitable edge.

Measured verdicts (returns, drawdowns, whether a family passed or failed) live in the validation record and each strategy's own reference page, so there is exactly one copy of every number.

A holding stays until it falls below rank 35; the family rosters draw only on names trading at least ₹5 crore a day.

These are the rules implemented here, including simplified versions. Read each definition before treating it as the named original. A missing roster does not mean zero matches or an unbuilt calculation.

Famous published screens, re-run point-in-time on our own exchange data. Where we cannot reproduce a rule exactly we say so and label it a proxy, never silently substituted.

Where these numbers come from. Price, momentum, volatility, the tradeable universe and the Nifty 500 benchmark come from NSE's own files, and Low-Volatility reads nothing else, so it is shown in full. Six screens that select on company fundamentals are rebuilt on exchange filings and show today's roster.

Retired from this site: the reconstructed records of the fundamentals screens and the Coffee Can screen. Both were built on a fundamentals archive mostly collected from Screener.in, and the exchange filings, which begin in 2019, are too shallow to rebuild them under the primary-sources rule.

Each built screen's rule, re-run as a 25-name book on our own data from its first rebalance, beside Nifty 500 over the same span. Under five years is not a track record, so no per-year figure is printed for it. A reconstruction, not a record of money.

The famous factor families, measured head to head on our own NSE record and shown with what they delivered here, never with textbook claims. The families that failed are shown with their numbers, because the menu includes what does not work.

Return / volatility ratio — not a Sharpe ratio: no risk-free rate is subtracted and costs are modelled flat. Flat cost charges 0.3% of turnover; realistic cost adds the spread and slippage the Nifty 500 hurdle is measured with.

A family with a leg read from company accounts (quality, value, earnings or book yield) rests on the vendor-collected fundamentals archive before 2019, so its figures and its verdict are withheld under the primary-sources rule, until it is rebuilt on exchange filings.

Momentum's edge here is a known risk premium, not stock-picking skill. Quality momentum as measured here blends company quality with strength; the CRAFTSMAN-25 book is a different, price-only rule.

Every entry and exit the factor families have made, newest first. The rule decides these — nobody overrides them, which is the point.

These are variants of one idea — a book that buys stocks either turning up out of an oversold patch or already showing confirmed strength. They are shown here because they exist and you should be able to see them, NOT because they are usable. None of them holds real money. The 2026-10-03 checkpoint starts the review; at least eight completed forward quarters are required before the frozen family adjudication, which graduates at most one.

Read the three return columns as three different questions. Gross is what the rule earned before real trading friction — the number that flatters. Net is the same rule after brokerage, slippage and impact on a real broker's cost model. Figures are shown only where admitted to this board. Four other variants have earlier in-sample measurements whose current basis is unverified; those cells show that status. Other absent studies say not measured. Adjusted estimate discounts historical results for selection from many trials; it is neither an observed forward return nor a forecast. Four siblings have no forward figure computed yet and say owed. None of the three columns is net of TAX, which is yours and not modelled anywhere in this estate.