Every rotation strategy I have ever seen starts from the same clean idea. Rank things by relative strength, hold the top slice, drop the laggards, repeat. On paper it is beautiful. In practice it works in maybe a third of the calendar and quietly bleeds you the rest of the time, and the part that decides which is which has almost nothing to do with your ranking logic. It has to do with what the market as a whole is doing.
The failure mode is specific and worth naming. In a choppy, highly correlated tape, yesterday's leader is tomorrow's laggard and the day after's leader again. Nothing is actually trending, everything is just sloshing around the same mean, and a strength-ranked strategy buys the top of each slosh and sells the bottom. You are not rotating toward an edge, you are paying the bid-ask spread and slippage to whipsaw yourself. The ranking is fine. The regime is wrong.
I saw this most clearly running the same simple momentum rotation across a basket of sector ETFs and a basket of crypto majors. Same logic, hold the top few by trailing strength, rebalance monthly. Over the trending stretches it did what it was supposed to and rode the leaders. Over the sideways stretches it gave most of that back, and the drawdowns lined up almost exactly with the periods where everything in the basket was moving together. The strategy was not broken during those stretches. It was just being asked to find leaders in an environment that did not have any.
The four phases, kept boring on purpose
The old Wyckoff framing splits a market cycle into four phases, and I keep coming back to it because it maps cleanly onto whether rotation should be on or off. I am not going to pretend I can call these in real time with precision. I can usually tell which one I am probably in, and probably is enough to change what I do.
- Accumulation. Price is basing after a decline, range-bound, volatility compressing. Correlations are high and dispersion is low. Nothing is really leading yet, so relative strength rankings are mostly noise.
- Markup. The breakout regime. Price is making higher highs, breadth is expanding, and crucially dispersion widens so leaders separate from laggards and stay separated. This is the only phase where chasing strength is actually chasing something durable.
- Distribution. Topping, choppy, still elevated but no longer trending. Rankings churn violently here. This is the phase that eats rotation strategies alive because they keep buying the last thing that popped.
- Markdown. The downtrend. Correlations spike toward one, almost everything falls together, and the least-bad asset is still an asset that is falling. Relative strength tells you which thing loses slower, which is a terrible thing to be right about.
The pattern that matters across all four is dispersion. Rotation needs winners and losers to separate and stay separate. Markup gives you that. The other three do not, in different ways, and each one breaks a strength strategy for its own reason.
Classifying the phase without kidding yourself
You do not need an exotic model for this. A handful of coarse measures, combined, gets you most of the way, and coarse is a feature because it keeps you from flipping states every other day. Here is roughly the checklist I run.
- Trend. Is the broad benchmark above or below a slow moving average, say something in the 150 to 200 day range, and is that average sloped up or down. Above and rising leans markup. Below and falling leans markdown. Flat and crossing repeatedly is your choppy warning.
- Dispersion. Look at the spread of returns across your universe over a trailing window. Wide spread means leaders and laggards are genuinely different. Tight spread means everything is moving together and your ranking is measuring randomness.
- Correlation. Average pairwise correlation across the universe. When it climbs toward one, diversification and rotation both stop working, because there is really only one asset moving in different disguises.
- Volatility state. Rising volatility with falling price is the markdown signature. Compressing volatility after a decline leans accumulation.
Combine those into a simple state rather than a score you overfit. Trend up plus wide dispersion plus moderate correlation is your green light for aggressive rotation. Trend down or correlation spiking is your signal to stop rotating and get defensive. The messy middle, flat trend with churning rankings, is distribution or accumulation, and the honest move there is usually to do less.
A practical tell I trust more than any single number is rank persistence. Take your strength ranking from a few weeks ago and check how much of the top slice is still in the top slice today. In a real markup, leaders stay leaders for a while, so persistence is high and rotation has something to hold onto. When the top slice completely reshuffles every rebalance, that is the market telling you there are no durable leaders and your ranking is just chasing the most recent bounce. High turnover in your own holdings is a symptom, and it is one you can measure directly off the strategy you are already running.
Switching what you actually hold
The phase is only useful if it changes the position. I think of it as three gears rather than a dial, because a dial invites you to fine-tune noise.
In a confirmed markup, run the rotation the way it was designed. Hold your top strength slice, rebalance on your normal cadence, let the trend do the work. This is the environment the strategy was built for and the only one where the ranking is load-bearing.
In distribution or accumulation, the uncertain middle, downshift. Hold the benchmark instead of the leaders, or hold a smaller number of positions, or widen your rebalance interval so you stop reacting to every rotation of rankings. The goal is to stop paying transaction costs and slippage to chase strength that is not persisting. Holding the index here is not a cop-out, it is refusing to make a bet the environment will not pay off.
In markdown, the least-bad asset is still losing money, so the defensive holding is cash or short-duration bills, not the strongest-looking equity. Relative strength between falling things is a trap. Sitting out is a position, and in this phase it is often the best one you have.
The failure mode to watch for
The classifier itself whipsaws at the edges, and if you let it flip your entire book every time trend crosses flat you have just moved the whipsaw up a level. I lean on two things to keep that in check. First, require confirmation before switching gears, a few days on the new side rather than a single close. Second, size the transition, so a phase change scales you between rotation and defense over a short window instead of dumping the whole book on one signal. You will be late to some turns because of this. That is the price of not getting chopped up at every false one, and I would rather pay it.
The thing I keep relearning is that the rotation logic was never the edge. It is a lever, and whether pulling it helps or hurts depends almost entirely on the regime you pull it in. Spend your effort on knowing which phase you are probably in and on being willing to sit in cash when the answer is markdown. The ranking can stay simple. Knowing when to trust it is the part worth working on.