Nifty sector rotation's advantage over equal weight fails Holm adjustment

Holding two of three Indian sector indices earned 16.7% a year from April 2018 to September 2026. Bank, IT and Auto individually earned 9.8% to 11.0% a year. Timing accounts for the gap. A single test window offers thin support for expecting it to persist.

How Hattali chooses the sectors

Hattali encountered the top-down framework during an internship at a Pune portfolio manager. It favours sectors whose earnings are growing, valuations are low relative to their own history, and prices are trending upward. For the trend component, the paper draws on the industry-momentum evidence of Moskowitz and Grinblatt, its closest cited analogue.

Three signals put the premise into practice. Earnings growth (EG) measures the log 12-month change in EPS, inferred by dividing the index level by P/E. Valuation comfort (VAL) subtracts the percentile of P/B from 100, using the sector's own expanding history and requiring at least 24 months. Momentum measures the price return from t-12 to t-1. The signals are ranked across sectors, then combined at 40/30/30. Strategy A, as the paper names the rule, holds the top two sectors at equal weights. It rebalances quarterly and charges 25bp one-way.

Nifty Bank, Nifty IT and Nifty Auto make up the universe; the benchmark is the Nifty 100. Hattali set the rules using January 2011 to March 2018, a development sample of 87 months. He then froze the code and ran it once over April 2018 to September 2026, covering 102 months.

Over that test period, Strategy A earned a 16.7% CAGR with a 0.56 Sharpe. The Nifty 100 price index returned 10.1% with a 0.27 Sharpe. In the abstract, the author acknowledges that none of the primary hypotheses survives Holm adjustment. The point estimates do exceed his pre-set economic thresholds: at least 2% active return, an information ratio of 0.3 and a 50bp cost break-even. Strategy A delivers 6.6%, 0.76 and about 433bp, respectively. Even so, he describes the statistical evidence as "remains inconclusive".

We could not reproduce the run. It requires National Stock Exchange (NSE) sector-index prices alongside the index-level P/E and P/B histories published by NSE. We have no Indian-market coverage, and substituting US sector ETFs would test a different market.

Does leaving one sector out pay?

Equal weight in the same three sectors is the relevant benchmark. Hattali identifies this comparison as the central test. Equal weight earned 11.5% with a 0.33 Sharpe, leaving rotation ahead by +0.23 on Sharpe. Its 95% interval spans -0.05 to +0.45. With block length 6, the bootstrap p is 0.076, rising to 0.21 after Holm adjustment. At block length 12, it falls to 0.044. The author discloses that result and rightly declines to rely on it.

A top-two-of-three rule makes one decision: which sector to exclude. Bank appeared in the portfolio 25 times across 34 rebalances, Auto 24 and IT 19. IT was left out most often, on 15 occasions. As the paper acknowledges, about 8.5 years of monthly data, or 102 observations here, cannot detect small differences in Sharpe ratios.

Clearing the economic thresholds does little to resolve this. The 5.3% active return over equal weight and IR of 0.65 make an appealing estimate. But an interval spanning zero leaves that estimate short of evidence.

Momentum alone: 0.47 Sharpe versus the composite's 0.56

The momentum-only rule returned 14.5% CAGR, with a 0.47 Sharpe and -29.6% drawdown. Adding the fundamental signals improves Sharpe by +0.09, with p = 0.601. The interval stretches from -0.28 to +0.37. It is also too wide to establish equivalence between the rules. Hattali accepts that the fundamental signals' incremental contribution remains unestablished.

The fundamental-only result is harder to defend. Its 11.4% return and 0.32 Sharpe nearly match equal weight's 11.5% and 0.33. Strategy A leads it by +0.24, with bootstrap p = 0.041. The Jobson-Korkie test, however, gives 0.107.

The fundamental series have breaks, too. Implied book value jumps on 29 September 2023 by 25% for the Nifty 100, 19% for IT, 24% for Auto and 4.5% for Bank. A change in accounting basis appears to explain the jumps. Hattali splices the series, although leaving it unspliced produces a higher Sharpe: 0.59 versus 0.56. Another break affects P/E on 31 March 2021. Growth windows crossing that date are marked missing. Our reading of the rule leaves EG missing for every sector at the 2021 quarter-ends. For Bank, it is also missing through most of 2020 because of a separate P/E break on 25 September 2019.

Momentum requires only prices.

Built in 2026, evaluated over one 2018 to 2026 window

The ranking changes in the development sample. Strategy A has a 0.14 Sharpe against equal weight's 0.18. Before June 2013, however, A chooses from only two sectors and therefore matches equal weight. Momentum-only ranks last in that period. Its test-period Sharpe of 0.47 follows a 0.00 Sharpe over 2011 to 2018.

Much of the test-period advantage comes from 2020: A returned 52.8% against the Nifty 100's 14.9%. In 2021, the comparison was 36.5% against 25.0%. Earlier, A lagged in April to December 2018, returning 4.5% against 6.0%, and in 2019, returning -1.1% against 10.4%. The path came with a -38.6% drawdown. That compares with -28.8% for the benchmark and -36.6% for equal weight in the same three sectors.

The weight tests deepen the doubt. Test Sharpe varies from 0.30 to 0.56 across 36 weight combinations; the baseline 40/30/30 lands at the maximum. Hattali says he chose the weights before running the grid. He also acknowledges that the outcome "could reflect a favourable configuration by chance". He built the framework in 2026 knowing the history, and describes the split as "procedural rather than a fully independent, prospectively specified out-of-sample test." Choosing only the top sector yields a 0.80 Sharpe and -27.7% drawdown. Reported solely as a sensitivity, that variant shows how heavily individual choices weigh in a three-asset universe.

Costs leave a 0.50 Sharpe at 100bp. Can it be held?

Annual turnover is 1.43 times portfolio value. Raising costs from zero to 100bp reduces Sharpe only from 0.57 to 0.50.

Hattali flags the unresolved implementation problem: investors cannot hold sector indices directly. Sector funds or baskets of constituents would be needed for live trading. Either could introduce tracking error beyond what the 25bp charge captures. The regression suggests exposure also matters, with beta of 1.05 and R-squared of 0.82. Alpha is 6.2% (t = 1.98, p = 0.048). These figures are unadjusted, and the author does not regard the alpha as conclusive.

A wider cross-section would change my view. The paper proposes adding FMCG, Pharma, Metals and Energy. Across that broader universe, the composite would have to beat momentum as well as equal weight using an implementable vehicle. Until then, the result amounts to a momentum tilt across three indices, with earnings and valuation inputs adding +0.09 Sharpe over momentum alone, indistinguishable from zero (p = 0.601).