This report is provided by MacroTerms.com
Data as of: 2026-11-06 · Settlement: Exchange final daily settlement
Soybeans (ZS=F) Deep Dive | Bearish — Oversold Bounce Fails at NEXT P 1185.75 | 2026-11-07
SHORT ZS=F (November 2026 delivery, ZSF27.CBT). Invalidation: a daily settle above 1198 (NEXT R2) plus a reclaim of the 20-day low-band would shift the structure from breakdown to base; until then every rebound is a lower-high into a descending channel. Three reasons: (1) trend — settle 1189 [2026-11-06] is down 7.97% on 20 days and sits at the very bottom of the 20-day channel (1179.5–1299.25 ex-today, position 9.2%), with five consecutive lower daily closes from 1204.75 to 1189; (2) carry — the M1–M2 spread is in contango at −8 cents (−0.67%) with a roll yield of −4.01%, so longs pay to hold while shorts are paid to wait, and there is no premium anywhere in the curve to squeeze; (3) margins — the US crush margin has fallen to 2.0148 USD/bu [2026-11-05] from 2.4634 twenty sessions earlier, a 1-year percentile of 29.37%, meaning the marginal demand engine is throttling back exactly as supply pressure builds. Volatility is elevated (ATR14 16.34, RV20 9.2%) but it is trending volatility, not event volatility — the VIX at a 95th percentile and OVX at 58.29 confirm a risk-off macro tape; DXY at 102.28 and 10Y at 4.86% add a strong-dollar headwind. Seasonality is mildly supportive (+0.77% mean, 9 of 15 up) but is context, not a catalyst. The desk stays short; NEXT S1 1181.25 and S2 1173.5 are the downside waypoints.
This brief is published in Chinese. The English translation is being prepared.
Today's Risk Map
Macro · Micro · Cross-assetVIP members unlock thresholds & transmission chains. Upgrade
Full brief is for VIP members
The body, key levels, charts and PDF are for active VIP members.
Single-name briefs Deep dive
Previous briefs
⟠ This month's strategy backtest
Daily AI brief ideas settled day-by-day with objective stop/target matching
No settled signals yet - sample building up.
* Backtest settles each daily trade idea: stop/target whichever first, flat at close when window ends. Past performance does not predict future results.