Sector Rotation Strategy in India: How Rotation Is Measured and Analysed

Sector rotation is the observation that market leadership does not sit still: over any long period, different sectors take turns outperforming and underperforming the broad market. A sector rotation strategy is a rules-based attempt to systematise that observation, by ranking sectors on a defined measure and shifting exposure at set intervals rather than by judgment.

The concept is easy to state and unusually hard to test honestly. This article covers how rotation is actually measured, the four choices that determine any result, and the specific ways the analysis misleads in an Indian context.

What rotation is, and what it is not

Rotation is a description of a fact about return dispersion: sector returns differ from each other and from the index, and the ordering changes over time. That much is simply observable.

What rotation is not, by itself, is an explanation. Various narratives are attached to it, most commonly an economic cycle story in which certain kinds of businesses do better at different points in the cycle. Those narratives may or may not hold. The measurement of rotation is entirely separate from the explanation of it, and the two get run together constantly. You can measure that leadership shifted without knowing why, and you should be clear which of the two you are claiming.

The four choices that decide every result

Any rotation analysis, however it is dressed up, comes down to four decisions. Change any one of them and the answer changes.

1. Sector definitions. This is the choice most often skipped over and it matters more than the rest. India has several coexisting classification schemes: the exchange’s own sector labels, global classification standards, and the constituent lists of published sector indices. They disagree, sometimes substantially, on where a diversified company belongs. A conglomerate with a financing arm, a manufacturing arm, and a retail arm has to be assigned somewhere, and different schemes assign it differently.

Worse, classifications change over time. A company reclassified from one sector to another rewrites the history of both sectors. If your analysis applies today’s classification to past data, you have introduced a subtle look-ahead: you are grouping companies by information that was not available on the historical date. That is the same mechanism described in what is lookahead bias, and in rotation work it is easy to miss because the labels feel like static facts.

2. The ranking measure. The most common is relative strength: divide the sector index level by the benchmark index level to get a ratio, then look at how that ratio has changed over a window. A rising ratio means the sector outperformed. Variations rank on absolute return over the window, on risk adjusted return, or on the ratio’s rate of change rather than its level. Some approaches use fundamental measures such as aggregate sector earnings revisions instead of price. Each measure ranks sectors differently.

3. The lookback window. Relative strength over three months and relative strength over twelve months frequently produce different orderings, because the shorter window responds to recent moves while the longer one is dominated by what happened months ago. There is no correct lookback. There is only a chosen one, and the choice must be declared and defended rather than tuned until the result looks good.

4. The rebalance frequency. How often the ranking is refreshed and positions adjusted determines turnover, and turnover determines cost. Monthly rotation and annual rotation are different strategies with the same name.

How rotation is presented

Three presentations dominate.

Relative strength lines. For each sector, plot the ratio of the sector index to the benchmark over time. A rising line means outperformance, falling means underperformance. Simple, transparent, and hard to misread. See relative strength explained for the computation in detail.

Ranking tables. A grid of sectors by period, showing each sector’s rank or return in each month or quarter. These make the persistence question visible: does a top ranked sector tend to stay top ranked, or does the ordering reshuffle constantly? That is the empirical question any rotation approach lives or dies on, and a ranking table answers it more honestly than a chart of the winners.

Relative rotation graphs. An RRG plots relative strength on one axis against the momentum of that relative strength on the other, producing four quadrants and a trail showing how each sector has moved between them over recent periods. It is a compact way to view level and change together. It is also frequently over-interpreted, since the quadrant boundaries are arbitrary and a sector can drift across one on noise alone. How to read an RRG chart covers the mechanics and the caveats.

Building a defensible analysis

If rotation is being tested rather than merely narrated, the construction requirements are strict.

• Fix the classification as of each historical date, not as of today. If historical classifications are unavailable, say so and treat the study as indicative only.
• Use total return indices, not price indices, so that dividend differences between sectors do not masquerade as rotation. Sector dividend profiles differ a lot.
• Declare the lookback and rebalance frequency before running the test, and report results across a range of both. If the conclusion survives only at one specific setting, it is a tuning artefact rather than a finding. This is the calendar-and-parameter form of overfitting.
• Model costs. Rotation implies turnover by construction. Brokerage, statutory charges, and the gap between modelled and achieved prices all apply. Transaction costs in backtests covers the components. A rotation result that ignores costs is not a result.
• Compare against the honest baseline, meaning the broad benchmark held throughout, on the same total return basis and over the same window.
• Report the full sample, not the good part. A rotation approach’s behaviour during the periods it failed is the informative part.

What it does not tell you

It does not tell you the cause. Rotation measurement is descriptive. Any economic cycle explanation layered on top is a separate claim requiring separate evidence, and the historical record of mapping sector leadership onto cycle stages in real time is not encouraging.

A sector index is not the sector. Several Indian sector indices are highly concentrated, with a small number of companies driving most of the index’s movement. When such an index outperforms, that may reflect two or three companies rather than a sector-wide condition. Before drawing a conclusion about “the sector”, check whether the index breadth supports it. This is where market breadth indicators and the advance decline ratio are genuinely useful as a cross-check on a sector index move.

It says nothing about any company in the sector. Dispersion within a sector is routinely larger than dispersion between sectors. A sector ranking carries no information about whether a particular company in it is doing well, and substituting a sector view for company work is a category error.

Ranking is not timing. A ranking tells you what has already outperformed over the lookback window. It is a backward-looking statistic presented in the present tense. Reading it as a statement about the next period adds a forecast the data does not contain.

Reshuffling can be noise. When several sectors have similar relative strength, the ordering between them flips on small moves. A strategy that acts on rank changes will trade on that noise and pay costs for it. Checking the size of the gaps between ranks, not just the ranks, is basic hygiene.

Results are fragile to settings. This is the honest headline. The sensitivity of rotation results to classification, lookback, and rebalance frequency is large enough that a reader should treat any single reported result as one point in a wide range of possible results, and ask to see the range.

Historical studies need real constituent history. Sectors gain and lose companies through listings, delistings, and mergers. Reconstructing sector history from current membership drops the failures and improves every sector’s past, which is survivorship bias at the sector level.

Related reading

• Portfolio metrics explained: the hub for how portfolio and market measures are defined and read.
• Relative strength explained: the computation underneath most rotation rankings.
• How to read an RRG chart: the rotation visualisation and its limits.
• Market breadth indicators: checking whether a sector move was broad or driven by a few names.
• Common backtesting mistakes: the errors that make rotation results look better than reality.

This article is educational. Altys Labs is not a registered research analyst or investment adviser, and nothing here is investment advice or a recommendation to buy, sell, or hold any security.