A more ambitious investor rebalances on a signal rather than on drift — shifting toward assets that look cheap, that are trending, or that suit the current macroeconomic regime. The appeal is obvious. The track record is not.
Macro-Regime Switching The pitch is to read the economy and rotate accordingly: risk-on when growth accelerates, defensive when it stalls. Candidate signals abound — the yield curve, manufacturing surveys, credit spreads, the copper-to-gold ratio (copper for industrial demand, gold for fear; together a serviceable coincident gauge of growth, and a decent tracker of bond yields). As a dashboard, these are fine. As a switch, they are treacherous: a single noisy ratio, fitted to the handful of recessions in the historical record, yields a strategy that backtests beautifully and disappoints live. The one tactical family with more durable support is trend-following — time-series momentum, or a 200-day moving-average filter of the kind Meb Faber popularized — and even that is a costly overlay that spends years lagging a plain buy-and-rebalance portfolio in exchange for cutting the worst drawdowns. Treat any regime-switch rule as an expense to be justified, never as a free improvement.
The Evidence Dimensional tested this head-on:82 720 timing strategies built on valuation ratios, mean reversion, and momentum, across the market, size, value, and profitability premiums in several regions. The overwhelming majority underperformed simply staying invested. Thirty appeared to win — but their edge was acutely sensitive to the chosen breakpoints and rebalancing dates, and it did not survive a change of region or premium. The winners, in other words, were indistinguishable from luck. Consistent exposure to a premium beat trying to time it, nearly every time.
Machine-Augmented Overlays A more recent line of work pushes back on the wholesale-skeptic conclusion. “Machine learning-based portfolio optimization: comparative analysis with the all-weather portfolio strategy”83 runs penalized linear regressions (LASSO, Elastic Net) on equity returns and gradient-boosted trees (LightGBM, XGBoost) on long-duration Treasuries, dynamically rebalancing across SPY/TLT/GLD. Out-of-sample is modest — around 3.7% on equities — but by the Campbell–Thompson identity, the Sharpe lift from that edge is substantial: the back-tested Sharpe approaches 0.70 against 0.58 for a static all-weather benchmark, with 13% annualized returns and meaningfully smaller drawdowns than buy-and-hold SPY. The result is more credible than the typical timing study — it uses only price-history features, with no obvious in-sample leakage — but it is still one paper and one sample; treat the headline numbers as an upper bound on what the strategy delivers live.
Even granting the gross edge, taxable execution will eat most of it. ML overlays churn positions, and at a top bracket of 37% federal plus state, NIIT, and any high-earner surcharges, half the alpha is paid to the IRS before you see it. Repeated round-trips through SPY trigger wash-sale disallowance under IRC §1091, locking up the very losses the rebalances need to harvest. If you want this exposure, isolate it inside a Roth IRA, defined-benefit plan, or self-directed solo 401(k) where turnover is invisible to the tax code. In a taxable account, express the equity and gold signals through broad-based index options (SPX) or futures (/ES, /GC), which under IRC §1256, “Section 1256 contracts marked to market” are wash-sale-exempt and receive 60/40 long-term/short-term treatment regardless of holding period. As an unboxed taxable ETF strategy the math does not work.
Seasonal Tendencies Calendar effects are the most-cited anomalies: “sell in May” (equities have historically done most of their work in November–April), the Santa Claus rally and turn-of-year strength, the January effect (beaten-down small caps rebounding once tax-loss selling subsides), and the volume spike of quarterly triple-witching expirations. The patterns are visible in long histories and have plausible mechanical causes — tax-loss harvesting, year-end bonus flows, the timing of 401(k) contributions. They are also weak, unstable, and largely arbitraged away; the January effect has been roughly neutral since 2000. As a primary strategy they do not survive trading costs. Their honest use is at the margin: if you already intend to rebalance or harvest losses near year-end, the calendar is a reasonable tiebreaker for when inside that window — not a reason to make a trade you would not otherwise make.
The thread running through all of this: rebalancing earns its keep as risk control and behavioral discipline, not as a return engine. The approach proven most efficient is unglamorous — a fixed allocation, a hybrid calendar-plus-threshold trigger with wide bands, trades funded first from cash flows and routed through tax-advantaged accounts, losses harvested alongside. Get those mechanics right and you have captured essentially all of the dependable benefit. The tactical overlays promise the rest; the evidence says they mostly deliver turnover.