Methodology

The whole framework, written down. Nothing computed in private.

Every Friday, shortly before the US close, five rules read the market and resolve to exactly one gear. There is no model you can't see and no discretion applied after the fact. Below is the entire logic — the five pillars, how they combine into a gear, the one leverage adjustment, and the design choices behind each. If you disagree with a choice, you can see precisely what to change.

The five pillars

Each pillar is a single boolean the rules read on the weekly frame. They are inputs, not opinions — none is "good" or "bad" on its own.

P1Macro trend

P1 = close(SPY, weekly) > SMA(SPY, 44 weeks)

Is the market above its long-term weekly trend line? The gate for everything else.

P2Slope

P2 = SMA(SPY, 44w) > SMA(SPY, 44w)[4 weeks ago]

Is that trend line itself rising? A flat or falling average is treated as no trend, even if price sits above it.

P3Volatility (VIX)

P3 = VIX(weekly close) < 25

Leverage is only taken when volatility is contained. Above 25, the rules will not run 3×.

P418-day confirmation

P4 = weeks_above(SPY, 44w SMA) >= 4   // ~18 trading days

A fresh cross must hold for roughly 18 trading days before it counts. This is the anti-whipsaw filter.

P5Credit stress

P5 = HYG(weekly close) < SMA(HYG, 40 weeks)   // stress present

High-yield credit is the early-warning system. When it breaks down with equities, the defensive gate engages.

How the pillars become a gear

The pillars are evaluated in a fixed priority order. The defensive gate is checked first — capital preservation outranks growth — then the growth gears. The first matching rule wins.

Every Friday, near the close — read P1–P5, then evaluate top-down

capital preservation is checked first

GATE 1Defensive

¬P1 (trend broken) AND P5 (credit stress)

both fail▶DEFEND

+ defense weakens▶CASH

gate closed → continue

GATE 2Growth

P1 ∧ P2 ∧ P3 ∧ P4 all true?

a pillar missing▶CRUISE

all true → TURBO, then trim check

M1aOver-extension trim

SPY > +8% above the 44-week line?

yes▶TURBO1×

no▶TURBO

DEFENDCASHDefensive gate (checked first)

If the macro trend is broken (¬P1) and credit is stressed (P5), the rules go DEFEND. If conditions deteriorate further — trend broken and the defensive sleeve itself weakens — they go fully to CASH. Nothing below runs while this gate is open.

TURBOTURBO — full growth

If the gate is closed and all of P1–P4 are true (trend up, slope up, VIX < 25, confirmed), the rules take 3× leverage.

CRUISECRUISE — cautious growth

If the trend holds but a pillar is missing (e.g. confirmation still building, or VIX elevated), the rules stay invested unleveraged rather than reaching for 3×.

Each gear maps to exactly one position

Positions are descriptions of what the rules computed, not recommendations to buy or sell.
Gear	Position	Why
TURBO	100% TQQQ	3× Nasdaq-100 — full growth
TURBO1×	100% QQQ	1× Nasdaq-100 — trimmed (see M1a)
CRUISE	100% QQQ	1× Nasdaq-100 — unleveraged
DEFEND	60% TLT / 40% SHV	Long bonds + T-bills
CASH	100% SHV	T-bills — out of the market

The M1a leverage trim

One overlay sits on top of TURBO. When the market is far above its trend line, a 3× position is most fragile — a snap-back hits leverage hardest. So when SPY is over-extended, the rules trim 3× TQQQ down to 1× QQQ. That trimmed state is TURBO-1×.

M1aOver-extension trim

extension = (close(SPY) − SMA(SPY, 44w)) / SMA(SPY, 44w)

if gear == TURBO and extension > +8%:
    gear     = TURBO-1x
    position = 100% QQQ   // 1x, NOT 3x TQQQ

TURBO-1× is a trimmed TURBO — still bullish, but holding 1× QQQ, not 3× TQQQ. This is why the per-signal page shows the gear and the position at equal weight: reading "TURBO" and assuming "3× TQQQ" would be a real and costly mistake.

Why these choices

Three decisions shape everything above. Each trades something away on purpose.

Why weekly, not daily

Daily signals on a 3× instrument generate constant whipsaw — and every round-trip pays decay and spread. The weekly frame deliberately trades responsiveness for far fewer, higher-conviction decisions. The cost: a fast crash mid-week can hit before Friday. We disclose that as the worst-week number rather than hide it.

Why a 44-week SMA

44 weeks is the high point of a favorable band on the weekly frame — its neighbors (42w, 46w) score lower, which is exactly why we publish the full SMA sweep on Evidence. It is the top of a robust range, not a number fit to one lucky path.

Why no synthetic data

TQQQ launched in February 2010. We could fabricate a longer track record by modeling 3× leverage before it existed; we don't, because modeled leverage is unreliable and would inflate the very number skeptics check first. The record is exactly as long as the real instrument — honestly bounded.

See the evidence How to trade it