LECTURE · FRAMEWORK
AI-cycle rotation · playbook + backtest

The 8-category board: how to read it, and does buying the laggard pay?

Part 1 answers the method question — is daily return + RSI(13) enough to summarise a category and call the cycle stage? Part 2 puts the idea on trial against 180 weeks of data since 2023-01-13.

8 categories RSI(13) + relative strength backtest 2023-01-13 → 2026-06-19 momentum vs reversion
Part 1 — method

Does daily return + RSI(13) summarise a category?

Short answer: RSI(13) is a good within-category gauge, but it is the wrong tool for the rotation decision. The board answers two different questions, and they need two different reads.

▲ What RSI(13) is good for
  • Is a category overbought / exhausted (RSI>70) or washed out (RSI<30)?
  • Is its own trend up or down (RSI above/below 50)?
  • A clean single number per tile for a daily glance.
▼ Where it misleads
  • In a bull market almost everything sits >50, so "others all >50" is nearly always true — it carries little information.
  • Absolute RSI doesn't tell you which category to own — that's a cross-sectional question.
  • RSI(13) on noisy daily data whipsaws; the rotation signal is weekly.
The fix: rank, don't level

For "where do I rotate," use cross-sectional relative strengthrank the 8 categories against each other (by 13-week return or by RSI) — not each one's absolute RSI level. The signal is the leader, the laggard, and the spread between them, not whether a tile is above 50. In this sample the average category sat above RSI 50 about 74% of all weeks — proof the absolute level is a weak cross-sectional filter.

So read the board on three layers

Part 1 — stages

Board shapes → cycle stage

The same patterns the live Rotation Monitor flags, as visual shapes and what each implies. This is the health/stage read — whether it's also a profitable trade is Part 2.

Part 2 — the test

Does rotating actually pay? (2023→now)

Weekly rebalance, signal lagged 1 week (no lookahead). Equal-weight category indices. Established board only (recommended-adds excluded). Current-constituents -> mild survivorship bias. Costs: 5/10 bps per switch.

Headline

Who led, who lagged

Equal-weight category indices since 2023 (all rebased to 100). Leadership share = how often each category was the strongest by 13-week relative strength.

Part 2 — the decisive test

Forward return by relative-strength rank

The cleanest answer to "buy strength or buy weakness?" For every week, rank the 8 categories (1 = strongest by 13-week momentum, 8 = weakest), then measure each rank's average return over the next 4 weeks. If buying weakness pays, rank 8 should be tallest (reversion). If momentum persists, rank 1 is tallest.

Part 2 — strategies

Strategy results (net of 5bps/switch)

Weekly rebalance. Momentum = hold the strongest N; Reversion = hold the weakest N (the proposed idea); Buy-the-laggard = your exact rule (buy weakest, hold until it becomes strongest, rotate). Benchmarks: own all 8 equally, or buy-and-hold SOXX.

Part 2 — bottom line

How to actually use the board

Part 3 — leading or lagging?

Is the stage a leading or lagging signal?

You asked the sharpest question in the whole project. The honest answer: the confirmed stage label is lagging by design — the regime engine uses hysteresis (it waits for confirmation) to avoid whipsaw, so it tells you where you were, not where you're going.

The leading content isn't the label — it's the components: the demand/cloud divergence, equipment/WFE orders rolling, memory pricing, and credit/MOVE. That's exactly why the Rotation Monitor surfaces divergences rather than just a verdict. Use the board as a risk-posture gauge; treat the divergence tells as the early read.

Tested your exact idea

"Run the strategy once a stage is confirmed, hold until it flips" — that's the regime-conditional test below. Because you must wait for confirmation, you eat part of the move; the cross-era table quantifies exactly how much.

Part 3 — regime-conditional

Momentum's win was the boom regime

Splitting 2023→now by an exogenous regime (SOXX above/below its 40-week average, lagged one week so you only act after it's confirmed) and re-running each strategy within each regime. Your suspicion was right.

So the ranking flips: momentum dominates risk-on and falls to the bottom in risk-off, where buying weakness / owning the board win. The synthesis is a regime switch — momentum when the trend is up, defensive when it isn't:

Part 3 — cross-era

Eight bubbles, one shape

How a 40-week trend-follow (momentum-with-an-exit) behaved vs buy & hold around each historical peak. Index level, free data — sector/basket depth needs the TradeStation export.

Part 3 — the repeatable arc

The six stages every bubble repeats

Across railways 1840s, 1929, Nifty-Fifty, Japan 1989, dotcom, housing, crypto, and SPAC — the same arc, and a different dominant strategy in each stage.

The top-tells that say "momentum is now too late"

These lead the label. When this cluster lights up, stop adding and rotate up the quality curve.

Part 3 — where are we

Which stage is AI in now — and the hand-off rule

Part 4 — the Fed turnaround

The Warsh pivot: higher-for-longer as a framework

Part 4 — historical rhymes

Has the Fed done this to a frothy market before?

Seven past Fed regime-changes, scored for closeness to Warsh-2026 across six dimensions (new hawkish chair, framework/communication overhaul, balance-sheet/QT, higher-for-longer, frothy tech market, productivity-raises-r*). No single one matches all six — each rhymes on a different dimension.

Part 4 — the playbook

What rhymes when a hawkish Fed hits a frothy market

Part 4 — through our model

Mapping the turnaround to our regimes & housing

Rotation Playbook · built on the Lecture Theme Framework