Super Signal
Adaptive Momentum
for NSE Equities

The base model selects stocks using a composite signal — the product of a stock's annualised 90-day price slope and the R² of that slope. This rewards stocks rising smoothly, not just stocks rising. Combined with eligibility filters (price above 50 and 200-day moving averages, positive 12-month momentum, daily turnover above ₹5cr, volatility below 4%), it creates a clean eligible universe ranked by trend quality daily. Each position was initially given a calibrated fixed target and stop based on the stock's own volatility, with a 9-month time stop. Capital was redeployed monthly. This baseline produced a 22.33% CAGR with a −24.11% max drawdown over the 9-year window.

The first experiment kept everything identical — same signal, same eligibility, same monthly cadence — and changed only the exit mechanism. Fixed targets were replaced with a trailing stop: an initial stop set at 1.5×ATR below entry, ratcheted upward daily as the price rose, never moving down. No profit target of any kind. The position exits only when the daily close drops to the trailing stop level.

The result was immediate and clear: CAGR rose from 22.33% to 24.02% — a +1.69pp improvement from a single mechanical change. Win rate fell from 54.3% to 36.6%, as expected (a trailing stop catches many small reversals before the position becomes profitable), but winners grew proportionally larger, producing a better overall expectancy. The key insight: fixed profit targets cut off positions that could have run for months. A stock in a genuine multi-month trend was being sold at an arbitrary percentage ceiling, leaving most of its move on the table.

The second experiment kept the trailing exit from Stage 2 and changed only the entry cadence — from monthly rebalancing (scan on the last trading day of each month) to daily scanning (scan every trading day, enter up to 3 new positions per day when slots are available).

Monthly rebalancing creates a structural inefficiency: when a stop-loss exits a position on the 5th of the month, that capital sits idle for up to three weeks waiting for the next rebalance date. Over a 9-year compounding system, that dead time is a consistent drag. Daily entry redeploys freed capital into the current top-ranked opportunity within one day. The result was a further +0.67pp of CAGR, with minimal increase in drawdown. The model also became more responsive — catching strong setups that appeared mid-month rather than having to wait for the calendar.

The third and largest experiment introduced the adaptive regime layer. The previous stages used a binary gate — either entries were allowed or they weren't, based on whether the Nifty 50 was above a single moving average. This is a blunt instrument: it treats a Nifty 50 that just crossed its 200-DMA the same as one that has been in a confirmed bull trend for six months.

The question was: why should position size, slot count, and stop width all be fixed when the market is telling you something different every day? In a confirmed, broad-participation bull market, you should deploy maximum capital with wide stops that let winners breathe. In a choppy, uncertain market, you should be smaller and tighter. In a bear, you should stop opening new risk entirely.

A four-signal regime score (0–4) was constructed from signals already computable from available data: the index's 50-DMA trend, 200-DMA trend, its own Trend Quality reading, and market breadth (what percentage of Nifty 500 stocks are above their own 50-DMA). The breadth threshold was calibrated at 40% after diagnostic analysis showed that 50% incorrectly classified genuine bull years (2019–2021) as "Choppy." The raw score is smoothed over a 3-day rolling average to prevent single-day whipsaw in parameters.

This single addition produced the largest improvement of the three experiments: the model's best year in the holdout window (2024) delivered +41.53% versus +10.16% for the fixed-gate daily baseline — a +31pp difference in one year, attributable entirely to the regime correctly identifying a strong environment and deploying maximum capital with wide trails, while the fixed gate deployed the same fixed position size regardless of conditions.