Methodology

Momentum is the empirical observation that stocks which performed strongly over the past year have, on average, continued to perform strongly over the following months. “12-1” means the score uses each stock's return over the past 12 months while skipping the most recent month.

The effect was documented rigorously by Jegadeesh and Titman (1993), “Returns to Buying Winners and Selling Losers” (Journal of Finance), and has since been found in most equity markets studied, including India, across nearly a century of data. It is one of the most persistent anomalies in finance — and one of the least comfortable to hold, as the crash section below explains.

At the one-month horizon, returns tend to reverserather than continue — last month's sharpest gainers often give some of it back (short-term reversal, documented by Jegadeesh in 1990). Including the most recent month therefore contaminates the momentum signal with a reversal signal. Skipping it is standard practice in the academic literature and in most factor indices.

Two stocks can both be up 60% over the year: one grinding steadily higher, another swinging 10% a week. Dividing the 12-1 return by annualized volatility (the standard deviation of daily returns, scaled to a year) ranks the steady climber above the wild one for the same headline return.

This is a form of risk-adjusted momentum, related to what the literature calls volatility-scaled or Sharpe-style momentum. It tends to reduce — not eliminate — the strategy's tail risk, and it keeps the screen from being dominated by the most speculative names.

The same formula is applied within five NSE index segments: Nifty 50, Nifty Next 50, Nifty Midcap 150, Nifty Smallcap 250, and the full Nifty 500. A stock's score is identical everywhere — only the competition changes. Filtering to a segment answers “which large-caps rank highest against other large-caps” rather than mixing size categories.

List sizes scale with universe breadth — top 10 for the two 50-stock indices, top 15 for the Midcap 150, top 20 for the Smallcap 250 and Nifty 500 — roughly the top 8–20% of each universe. The sizes are fixed design choices, not optimized numbers, and they are not a statement about how many stocks anyone should hold.

The rankings page shows where the selected segment's benchmark index sits relative to its own 200-day simple moving average — the Nifty Next 50 page shows the Nifty Next 50 index, the S&P 100 page shows the S&P 100 index, and so on. Where no free daily series exists for a segment's own index (the Smallcap 250), the badge shows the market index instead and names it. Comparing an index with its 200-day average is one of the oldest published trend rules in finance: some momentum practitioners and academic studies (for example, work on trend-following overlays) use rules like it to study when momentum strategies have historically been most fragile — deep downtrends followed by sharp reversals are where momentum crashes have occurred. The robustness grid's pre-declared regime variants remain defined on the market index (Nifty 50), unchanged.

The badge is descriptive information about a published rule's current state — nothing more. It is not a buy signal, a sell signal, or a suggestion to change anything about your investments. Trend rules generate false signals routinely, and their historical usefulness is itself a contested research question.

Backtests simulate the past with the benefit of hindsight; the tracking page removes that benefit. Each time the screen publishes a list, that list is frozen. At every subsequent weekly refresh, an equal-weight basket of those names — priced at their publication-day adjusted closes and held untouched — is measured against the segment's benchmark index over the identical window (the Nifty 50 where no free segment series exists), along with the basket's deepest peak-to-trough fall.

Because the lists are published before the measurement happens, there is nothing to tune and no way to quietly drop a bad stretch. The numbers are price returns without transaction costs or taxes, and the record only becomes meaningful as it lengthens — a few weeks of tracking is noise, stated as such on the page itself.

Any single backtest can be a fluke of its exact settings. The robustness page re-runs the identical backtest under four spec variants declared in advance — plain 12-1 versus volatility-scaled, each with and without the 200-day-average regime rule — and reports all of them, every time. The point is not to find the best variant (that would be tuning); it is to show whether the historical result survives reasonable changes to the formula. The published spec stays what it is regardless of which row looks best in any given run.

The screen recalculates on the last trading day of each month, and the backtest assumes the simulated portfolio is rebalanced to the new top 20 (equal weight) at that day's adjusted close. Monthly is the standard cadence in the literature: frequent enough to track the signal, infrequent enough that costs don't swallow the premium.

Momentum portfolios turn over heavily — often 50–100% of names over a few months. That churn is why cost assumptions matter so much.

The backtest charges 0.4% per side — 0.8% on a round trip — on every trade at every rebalance. That is meant to approximate the all-in cost for an Indian retail investor: brokerage, STT, exchange charges, GST, stamp duty, and (usually the biggest piece) impact cost and slippage in mid- and small-cap names.

Taxes are not modelled. Monthly rebalancing in India generates mostly short-term capital gains, which would take a further meaningful bite out of realized results.

This is the part most momentum content skips. The factor's average is good; its worst episodes are horrific. Momentum's signature failure mode is the “momentum crash”: after a market collapse, the momentum portfolio is stuffed with defensive winners, and when the market snaps back violently, the beaten-down losers it just dropped rally hardest. In the US in 2009, academic momentum portfolios lost on the order of half their value in a few months — a pattern documented by Daniel and Moskowitz (2016), “Momentum Crashes”. Similar reversals hit in 1932, in 2009, and in gentler form after the 2020 crash.

Momentum is also cyclical on multi-year horizons: it has had long stretches — years, not weeks — of underperforming the index. A factor premium is compensation for exactly this kind of pain; historically it has accrued to those who held through it, and there is no guarantee it will persist at all.

Prices come from yfinance (free, adjusted for splits and dividends). Free data has gaps, occasional bad prints and imperfect corporate-action handling. The backtest universes are the currentindex constituents — the Nifty lists from NSE, the S&P 500 from Wikipedia's constituents table — which introduces survivorship bias: companies that collapsed or were delisted along the way never enter the simulation, flattering historical results. Backtests also benefit from hindsight in ways that are hard to fully remove.

Every assumption is listed on the backtest page. If a number here disagrees with your broker or data terminal, trust theirs.