Factor & Sector Rotation:
A Systematic Parameter Optimisation

Two sleeves, one signal

This strategy allocates capital across two independent sleeves each month: a Factor sleeve and a Sector sleeve. At the end of each calendar month, the strategy scores every ETF in each sleeve by its N-month total return. The single top-ranked ETF in each sleeve receives the full sleeve allocation for the following month. This is pure momentum rotation, applied separately to two distinct groups of ETFs.

The Factor sleeve covers four style-factor ETFs: momentum, quality, value, and small-cap. These capture systematic risk premia that have been documented extensively in academic finance literature. The Sector sleeve covers nine SPDR sector ETFs spanning the full S&P 500 industry landscape, from technology to utilities. The two sleeves are designed to capture different dimensions of market leadership at any given point in the economic cycle.

Capital is split between the two sleeves either with a fixed 50/50 allocation (Split A) or dynamically based on the prior month's actual sleeve returns (Split B). Each sleeve's weight equals its prior-month return as a fraction of the total, floored at zero; if both sleeves had negative returns the prior month, the allocation reverts to 50/50. A third dimension tests six methods for filtering out adverse market regimes using SPY as a proxy for broad market health. Monthly filters (F1, F2) check the signal once at the start of the month and hold the position for the full month. Daily filters (F3, F4, F5) check the SPY signal at each day's close and exit or re-enter at the next day's open, allowing intra-month participation when the trend recovers.

The strategy makes no fundamental judgements and no macroeconomic forecasts. It follows momentum mechanically, checks one market regime signal, and rebalances once per month. Execution is assumed at month-end closing prices with no transaction costs or slippage.

Thirteen ETFs, two sleeves

All price and total return data is sourced from Compustat Security Daily (comp.secd) via WRDS. Returns are computed using the total return index, calculated as prccd × trfd (the Compustat total return factor), which adjusts closing prices for dividend reinvestment. This ensures returns reflect what an investor actually received, not just price appreciation.

The common backtest start date of July 2014 is determined by the latest inception among all 13 ETFs (QUAL, July 2013), plus a 12-month lookback buffer for momentum scoring. SPY is used solely as a drawdown filter signal and benchmark; it is not part of the rotatable universe.

Three variables, sixty combinations

The parameter space spans three independent dimensions. Every combination is evaluated over the identical date range using the same data, so Sharpe ratios and drawdown statistics are directly comparable across the full grid.

Momentum Scoring

At each month-end, every ETF in a sleeve is ranked by its N-month total return. The ETF with the highest score receives the full sleeve allocation for the following month. There is no blending or smoothing across lookback periods. For the total return calculation, the Compustat trfd factor is used throughout.

Drawdown Filters & Sleeve Splits

Monthly filters (F1, F2) are checked once at the start of each month using prior month-end data. If triggered, both sleeves earn 0% for the entire month. Daily filters (F3, F4, F5) are evaluated at each day's close; if triggered, both sleeves exit at that day's close and re-enter when the signal reverses, capturing partial-month returns. No cash interest is applied to idle days in either case. Split A allocates 50/50 between sleeves each month. Split B weights each sleeve proportionally to its prior month's actual return, floored at zero; falls back to 50/50 if both were zero or negative.

Top 10 combinations, ranked by Sharpe

Top 10 of 60 combinations sorted by Sharpe ratio. Top-3 rows highlighted. Click any column header to re-sort.

Sharpe Heatmaps

The three heatmaps show how Sharpe ratio varies across the (lookback, filter) grid for each split method, plus the cell-wise best across both. Warmer green indicates higher risk-adjusted returns. A robust strategy shows a broad green region, not a single outlier cell.

L1, Filter 3, Split A

The highest-Sharpe combination uses a 1-month lookback, price < 200d sma as the drawdown filter, and 50/50 fixed sleeve allocation.

Equity Curve vs Benchmarks (Log Scale, Rebased to 100)

Annual Returns vs SPY

Monthly Return Heatmap

Regime Analysis

Performance across distinct SPY trend regimes over the backtest period.

Strategy vs benchmarks

All three series evaluated over the identical period: July 2014 to March 2026. Equal-weight buy-and-hold is rebalanced monthly across all 13 ETFs in the universe.

What this study does not show

What the data suggests

Across 60 parameter combinations, momentum-based ETF rotation produces Sharpe ratios ranging from 0.26 to 0.81 over the 2014–2026 period. The top combinations approach but do not exceed SPY's Sharpe ratio of 0.868. Short lookback periods (1 month) dominate the top of the ranking; longer windows of 6 months or more produce markedly weaker results over this period. The inclusion of daily drawdown filters (F3, F4) meaningfully reduces maximum drawdown, particularly during the 2022 bear market, while preserving most of the upside through intra-month re-entry. The difference between Split A and Split B is modest, suggesting that the rotation signal itself matters more than the weighting rule between sleeves.

This strategy is suited to an investor who wants systematic, rules-based exposure to relative equity strength without individual stock selection. It requires monthly attention and tolerance for periods of full cash when filters trigger. The results presented here should not be extrapolated without accounting for the data mining, short live-history, and execution-timing caveats described above. Walk-forward or out-of-sample testing on data after 2024 would be the appropriate next step before any live implementation.