You can't time the market, but...

The Rotation Signal

There is a particular kind of article a senior operator reads twice. The first read is for content. The second is to figure out what to do.

This week it was a Telegraph dispatch where a long respected British investor, who has called every major bubble of the last forty years, said we are now inside the largest one he has ever seen. The S&P 500's Shiller price-to-earnings ratio entered the year above 40, marking its second-priciest valuation since 1871, with the only prior instances at this level followed by drawdowns of 49 percent and 25 percent. OpenAI has committed to spending 1.4 trillion dollars over eight years on data centers against 13 billion in revenue, while a National Bureau of Economic Research study found that 90 percent of firms report no productivity impact from AI. The structural setup is clear. The timing is not. The Motley FoolWikipedia

The interesting question is what an operator at our stage actually does with this. The answer is not to sell. The answer is to build an early warning system and let the data tell you when to move.

So I built one. Well, Claude did. A weekly tracker that grades five specific signals: the QQQ to XLP ratio (tech versus consumer staples momentum), RSP to SPY (equal weight versus cap weight breadth), the put-call ratio on tech, the ten-year yield direction, and net dollar flows in and out of tech ETFs. Each indicator scores either HOLD (tech-dominant) or ROTATE (defensive rotation) week over week. The composite is a score out of five.

The premise is that bubbles do not announce themselves. They leak. Smart money rotates first, the financial press explains the rotation second, the retail investor reads about it third. By the time the third group acts, the first group is already positioned for the next regime. The five-indicator dashboard is an attempt to listen for the leak before it becomes the headline.

Last Friday's reading was 1 out of 5. The QQQ to XLP ratio flipped to ROTATE as consumer staples rebounded faster than tech advanced, even with the Nasdaq making a new high. One indicator out of five is not a signal. It is noise with a heartbeat. The other four held: equal weight breadth still lagging cap weight, put-call ratios still showing aggressive call buying, the ten-year at 4.39 percent not yet enough to dislodge tech leadership, ETF flows into QQQ and XLK still positive on both the five-day and one-month windows.

The rule I set is deliberately boring. A composite score of 3 or higher, sustained for five to six consecutive weeks, is the threshold at which I trim tech meaningfully and rebalance into the cohort the market is actually voting for. Anything shorter is a head fake. Anything weaker than 3 of 5 is institutional fidgeting, not regime change.

What I am protecting against is not the crash. The crash is unmanageable in real time. What I am protecting against is being correctly bearish six months early or correctly bullish four weeks late. The dashboard is a discipline, not a prediction. It removes my opinion from the decision and replaces it with a checkable threshold.

The operators who came through 2000 and 2008 with capital intact were rarely the ones who saw it coming first. They were the ones who had a pre-committed exit signal and the discipline to act when it arrived.

Five weeks at 3 of 5. Then we move.