◆The plain month-end edge died after 2017. What survives is a rare, vol-gated version: it only works when a month's stock-vs-bond move is extreme (≥3.8%) AND VIX ≥ 20. Then fading it into T−3/T−2 paid +193 bps, 67% hit, significant every era incl. 2018–26. Fires ~2×/year; otherwise this stays silent.
Today's call
—
Checking the gate…
① Flow size · divergence
—
need |move| ≥ 3.8% (this month, stocks vs bonds)
② Volatility · VIX
—
need VIX ≥ 20
—
−3.8%
+3.8%
← stocks crashed (long-equity zone)stocks ran up (short-equity zone) →
The validated trade — vol-gated extreme fade, SPY vs AGG
Sample
mean / trade
t
hit
n
Gate = |month-to-date stock−bond| ≥ 3.8% AND VIX ≥ 20, checked at ~T−4 (before the trade). Trade = fade the move on T−3/T−2 (short the winner / long the loser, dollar-neutral), exit at the month-end close. Without both conditions the historical edge is ~0.
How to trade it
When the gate opens (~2×/yr). On T−3/T−2: if stocks ran up ≥3.8% vs bonds → short equity / long bonds; if they fell ≥3.8% → long equity / short bonds. Dollar-neutral, exit at the month-end close.
Instruments. Futures ES vs ZN/ZB (lowest friction); ETF SPY vs AGG/IEF. Follow-on: long equity on T+1/T+2 after a gated up-month.
When closed (most of the time). Stand aside. The edge is absent in calm vol or small divergence — don't force it.
Caveats
This is a regime/turbulence tool, not a monthly strategy. It overlaps high-vol mean-reversion (your fear-capitulation playbook is a cleaner expression).
Recent ex-crisis sample is small (n=7) — strong (t=2.5, 86% hit) but not bulletproof. Size accordingly.
Quarter-end months add noise; the quarterly signal itself has no edge and is excluded.