We tried to make a momentum strategy win every year. The momentum had to go.
Strota has been testing a residual momentum strategy on Indian stocks - a screen that ranks companies by how far they have run after the market's own move is stripped out. The version we had made money, but it made it lumpily: two outstanding years carried an entire decade. The question this round asked was blunt. Could the lumpiness be engineered away? It could. But only by taking the momentum out of the momentum strategy.
That is the finding, and it deserves stating before any of the arithmetic. Holding only the top 20 momentum names is what produced the two spectacular years, 2023 and 2024, and it is also what produced the loss in 2025. The signal was not a steady edge with a couple of unlucky patches attached. The signal was the swings.
Measured as the share of total profit earned in the two best years, concentration fell from 51% to 42%. The strongest configuration ended positive or flat in every one of the ten years tested, its worst year a loss of 3%. Gains now sit spread across 2017, 2020, 2021 and 2023 rather than piling up at one end of the decade. Worth being precise, though: this is not a literal ten-out-of-ten record, and the entry itself calls the goal substantially met rather than met.
What came out ahead is barely a stock-picking strategy. It is an equal-weighted basket of the 100 most liquid names, paired with a 0.7 times short against the Nifty that is opened and closed inside the trading session. Year by year on the test panel it reads 45% in 2017, a 1% loss in 2018, 13% in 2019, 44% in 2020, 51% in 2021, flat in 2022, another 51% in 2023, 34% in 2024, a 3% loss in 2025 and 1% in the stretch of 2026 covered. Its Sharpe ratio - return per unit of volatility, where above 1 is generally thought good - came out at 1.48, on a compound growth rate of roughly 23% a year.
Six rounds of iteration and about 35 separate configurations sit behind that result, all run on the same ten-year panel and all net of fees, meaning brokerage and taxes were deducted rather than waved away. Costs are not a detail for this design. It trades daily: the basket is carried overnight, the index hedge lives and dies inside market hours.
The engine underneath is no longer momentum. It is the overnight anomaly - the long-noticed habit of Indian equities doing their earning between one close and the next open, then giving part of it back during the day. Widening the long book from 20 names to an equal-weighted 100 stripped out the momentum tilt and left that pattern in plain view. Described honestly, the survivor is: own the liquid market overnight, short the index lightly by day. Filing it under momentum would flatter it.
A compromise was tested too, the top 60 ranked on a diversified version of the signal, and it behaves the way compromises usually do. Some signal is retained. Concentration climbs back to 46%, and 2025 turns into a 4% loss. A different variant bolts on a drawdown brake, halving the long book whenever its own trailing 21-day return is negative. That flips 2018 and 2025 into the green and trims the worst year to a 2% loss, but pushes concentration up to 48%. The two variants disagree about which goal matters more, and neither is obviously right.
Three ideas failed outright and are written down so nobody spends a week on them again. Volatility targeting levered up into the calm 2023 bull run, which made concentration worse instead of better. An overlay using the 200-day moving average to step aside in downtrends was whipsawed in 2018 on every attempt, because 2018 was a year when the index rose while midcaps sank - a divergence an index-level filter is blind to by construction. Running the same filter stock by stock did nothing for 2025.
Both stubborn years want opposite medicine. 2018 punishes anything that cuts exposure while the index looks healthy; 2025 punishes anything that keeps momentum exposure on. Chasing a strictly positive record every single year simply rotated the small loss around between 2018, 2022, 2025 and 2026, which is curve-fitting - tuning the dials to fit history rather than to anything that repeats. Testing was stopped at that point, deliberately.
Now the part that matters most, and it cuts against everything above. The evenness looks real. The level is fantasy. A Sharpe near 1.5, and years in the 20 to 50% range, are inflated by survivorship bias: the liquid pool is drawn from today's Nifty500, so firms that were delisted or shrank out of the list are quietly missing from a decade of history. Equal weighting and overnight holding maximise that distortion rather than diluting it. Layered on top, the overnight edge is measured at exact opening and closing marks, which are not prices anybody actually transacts at.
So read the figures as a statement about the past - a past that was kind to them. With real fills on the index leg, and without the invisible survivors, a fraction of these numbers is the sensible expectation. The structural property, flat years and no single blow-up, has a fair chance of surviving contact with a live account. The size of the returns does not. Settling that takes a paper run against real fills, which is the only arbiter that counts, and not another round of tuning.
More money stories
- Low-volatility stocks, not momentum, produced a book positive in every completed year
- Fading morning gap-ups paid in liquid stocks and lost in illiquid ones
- Widening the stock pool, not the signal, gave this hedged low-volatility test a positive decade
- A short-volatility test on Nifty returned 11.3% a year, inside a 15% drawdown
- Momentum on India's most liquid stocks beat the Nifty — after we found the bug in our own test
- Boring stocks held up better than winners in our re-test of Indian factor books