Eight published strategies, implemented from their authors’ own stated rules and left to compete on a live record. Each publishes a portfolio of twelve names; from that day the site tracks what those names did against the market over the same window. Nothing here is back-fitted — the picks are recorded before the outcome is known. Open any row for the rule set and its full position ledger.
Tracking since 21 August 2026 across 15 published sessions. Ranked on the 10d window — the longest one that has had time to resolve. That is far too short to separate skill from luck — treat the table below as a scoreboard that has just started, not a verdict.
Leaderboard
equal-weighted portfolio minus the equal-weighted universe, averaged over every published day
Booked is the average excess return of positions the rules have already closed — opened the day a name entered the portfolio, closed the day it dropped out or the holding window ran out, measured against the equal-weighted universe over the identical days. It is a record of what happened, not a forecast.
Only own stocks already in a confirmed uptrend, near their highs, outperforming the market.
Where our version differs: Relative strength is measured as a percentile of one-year return across our universe, in place of the proprietary IBD RS rating.
Rank by the steepness of the trend multiplied by how straight it is, so a single lucky gap cannot buy its way to the top.
Where our version differs: Position sizing by volatility parity and the index-above-200-day gate are part of the published system and are not applied here; this is the ranking only.
A year of return, ignoring the most recent month because it tends to reverse, and preferring a steady climb to a jumpy one.
Where our version differs: The smoothness filter is the share of up-days over the window, a simpler cousin of the published information-discreteness measure.
Current quarterly earnings growing hard, price near a new high, and the stock leading rather than lagging its market.
Where our version differs: C, N and L only. Institutional sponsorship and share-supply data are not available for Indian retail feeds, so the S and I of CAN SLIM are missing.
A committee of language-model agents reads each company — four analysts, a bull and a bear researcher who argue it out, a trader, a risk desk — and votes buy, sell or hold.
Where our version differs: Run exactly as released, on DeepSeek, over the twenty most liquid names each afternoon. We record its buy list before the next session and score it like every other row here. Its written reasoning is generated prose about real companies and is not published anywhere on this site.
Buy good businesses cheaply: rank every company on both return on capital and earnings yield, then add the two ranks together.
Where our version differs: Profit before tax stands in for EBIT, and equity plus borrowings for capital employed. Greenblatt excludes financials and utilities; we do not, and that matters most for the banks.
A business that earns well on its capital, owes little, has been consistently profitable, and is not priced above its own history.
Where our version differs: This is the loosest attribution on the page. Buffett has never published a screen, and the qualitative core of his approach — the durability of a business — is exactly what a screen cannot see. Read it as a quality filter, not as his portfolio.
Growth is only worth paying for up to a point. Compare the price-earnings multiple with the growth rate behind it.
Where our version differs: PEG uses trailing earnings growth rather than forecast growth, because consensus estimates are not freely available for most Indian mid-caps.
Score a company on simple binary tests of profitability, leverage and cash, and prefer the ones that pass most of them.
Where our version differs: Piotroski uses nine signals across two years of annual accounts. India publishes machine-readable balance sheets half-yearly and only since 2025, so five signals are computed and the score is out of five, not nine.
These are our implementations, not these investors’ portfolios. Every one of them describes judgement a screen cannot carry — which businesses are durable, when a market is hostile, how much to hold. Where Indian disclosure does not support a rule, it is approximated, and each card says where. None of these people are associated with this site.
Portfolios are equal-weighted, held for the stated window, and compared against the equal-weighted universe over the identical period. No costs are deducted here — see Are the scans working? for what trading actually costs.
How these numbers were made. Computed here from NSE and BSE end-of-day bhavcopy, adjusted for splits and bonuses, not supplied by anyone being measured. Portfolios are recorded on the day they are published and scored forward from that date, so no figure on this page could have been chosen after the outcome was known. The code that produced them is stamped in build.json.