The yen short capitulated —
the tech short doubled down.

Every market as a share of its own 1-year positioning extreme, as of the Tue 4-Aug COT close. Green = the crowd was long; red = short. Compare it with last week and one whole block has drained: the FX row is no longer a wall of −100s, because seven of the eight G10 shorts covered and the yen went −100 → −28. Two red cells did NOT drain and they are this week’s trades — Nasdaq at a fresh −100 (leveraged funds pressed a record short into a +7.1% melt-up) and CAD at −100 (the only currency short that added). At the top, rates stay pinned deep green at +99 while the long end keeps rising, and a genuine new green cluster has appeared in metals, where copper hit a RECORD +100 but silver (+31) and platinum (+52) still have fuel. One convention, everywhere: positioning is shown per currency — red = specs are SHORT that currency (CAD −100 = record-short the loonie). A whole board of currency-shorts is the report’s one big long — the US dollar (the FX bar sits at the bottom of the compass: currencies net short −45, up from −61, = a shrinking dollar long). The trade is always the pair. Hover any tile for the current call.

How the conclusions moved versus the 28-Jul print — the call changes first (last week → this week), then the biggest positioning shifts that drove them. This is the week-over-week delta; the detailed same-week read follows below.

The edge isn’t the chart — it’s how the crowd’s positioning moved versus how price moved in the SAME week (Tue 28-Jul → Tue 4-Aug). Agree → a CONFIRM = a TREND (ride it); fight → a DIVERGE = a trapped crowd = a reversal/squeeze brewing (fade it). Both are actionable — the §03 trades are split into exactly those two playbooks. Price since the print and the technicals below are a secondary timing layer — not the driver.

Three sessions, and one number reset the macro. July payrolls fell −23,000 on Friday 7-Aug — the first outright monthly job loss of the cycle, against a +83k consensus — and traders cut the odds of a September Fed HIKE to about 42%. The front end rallied (2Y to 4.19%, its lowest since 17-Jul) while the long end kept climbing (30Y 5.20%): the same bear-steepener, now with a growth crack in it. That print is why gold ran +6.0% to ~$4,400, its strongest week since January, with silver +5.4% alongside. Equities took the soft data as pure relief: the Nasdaq-100 added roughly $3.5trn of market value in four sessions on hyperscaler AI earnings, the S&P closed at a record 7,757 and the Dow cleared 54,000 for the first time — all while leveraged funds were pressing shorts. In commodities, US–Iran talks to reopen the Strait of Hormuz began 3-Aug and Trump paused a strike, which knocked crude −4.4% into the print; the strait then stayed shut, tankers kept being turned back, and crude recovered +3.2%. Sugar ran +9.6% as the 2026/27 balance was cut to a 3.17MMT deficit. Price + optional technicals since the close — useful for timing the entry, not the thesis:

→ Every actionable row above becomes a trade in §03 What To Do, sorted into the two playbooks: Trend (the confirms — ride) and Reversal / squeeze (the diverges — fade on the trigger).

Two kinds of trade, because there are two kinds of signal (from §02). TREND = a CONFIRM (crowd & price agree) → join / ride, exit by trailing — let it run. REVERSAL / SQUEEZE = a DIVERGE (a trapped / maxed crowd turning) → wait for the trigger, then fade, and take profit into the snapback (time-boxed, tighter stop). Within each, ▲ LONG / ▼ SHORT sort by conviction — a YOUNG sub-extreme crowd outranks an EXTREME ⚑ one (the 1-yr fuel is spent → a ride, not a fresh max). FX is always the tradeable pair.

★ Highest conviction this week is ▲ LONG SILVER (6.0) — the framework’s premium setup, a YOUNG confirmed build: +2.8k added at only rel +31 while silver rose +4.7% in-week and +5.4% since, in the same trend that took gold to ~$4,400 on the −23k payrolls miss. Buy the leg whose crowd has not arrived: gold at +80 ⚑ is a hold, silver is the entry, and ▲ Platinum (5.0, rel +52, the week’s best metal at +8.2%) is second. The squeeze trade is ▲ LONG Nasdaq/S&P (5.5) — leveraged funds pressed a RECORD −100 short into a +7.1% week and $3.5trn of melt-up, the board’s biggest DIVERGE. Then ▲ MSCI EM (4.5, the youngest position anywhere at rel +2) and ▲ EUR/USD (4.0 — still rel −80, so unlike the yen its fuel is intact). On the short side the best is dated: ▸ SHORT corn (4.5) into the 12-Aug USDA yield print against trapped longs, plus ▸ SHORT USD/CAD (4.5) below 1.3855 — the one G10 short that never covered — and ▼ Soybean Oil (4.0), the cleanest unfinished ag liquidation. Closed this week: USD/JPY, by its own written exit — the yen short covered from −100 to −28. Every table below is conviction-sorted — the top row is the strongest.

Honesty box. Three things must be said plainly. First, the call that worked, worked — and the discipline is to close it. SHORT USD/JPY was carried at ★★★ 6.0 with a written exit: "close if the yen short covers back toward −60". A joint US–Japan intervention forced +117,939 contracts of covering to −28, and the pair has gone flat at 157.8. Honouring that exit while the trade still feels good is the whole point of writing it down in advance; the remaining dollar-long risk is in CAD, not JPY. Second, we under-rated the metals last week and said so too late. Gold was called a "stalling hold" at rel +73 and silver "no edge" at +24 — then payrolls printed −23k, gold ran +6.0% to ~$4,400 and silver +5.4%. The framework was not wrong about the crowd, it was late to the sponsor; the correction is to buy the young leg (silver +31, platinum +52) rather than chase the one everyone can now see. Third, last week’s sugar veto and grains downgrade both paid — sugar +9.6% after the balance flipped to a 3.17MMT deficit, and the ag complex liquidated exactly as the sponsor-less build warned. Open risks on this week’s book: the equity squeeze is the fade of a record short, so it is time-boxed by definition — if leveraged funds simply keep pressing and the melt-up stalls at the 50-day, there is no snapback to harvest. Silver is a 4.4%-ATR instrument: correct thesis, wrong size, still loses. Corn’s short depends on a single dated print (12-Aug); a sub-183 yield or an escalation of the EU drought that cut the bloc’s corn −4.6MMT kills it. USD/CAD needs its 200-day to give way, and a Hormuz deal capping crude at $70–75 would defend it. Never short the physically tight names — COMEX copper, LME aluminium at century-low stocks, cattle at 75-year-low herds, coffee into El Niño flowering risk, and sugar into a deficit. Never chase a spent squeeze (JPY, cocoa) or press an oversold trend (Lean Hogs at RSI 28).

Rates, equities and VIX — the coiled extremes. Each is Now / Waiting for / If it fires, so “no trade” is never the end of the story: it tells you the exact trigger that would create one.

The relationships that mattered — one record crowd capitulating while another doubled down, a real precious-metals trend arriving with a macro sponsor, and the physical world overruling the crowd for a second week running. Each was adversarially stress-tested; the ones that broke are traps to avoid.

CFTC Disaggregated — Managed Money, positioning to Tue 4 Aug 2026 (self-pulled). The core read: Net / ΔNet / Flow (how the crowd moved) against Px · 28 Jul→4 Aug (how price moved in the SAME week) → CONFIRM or DIVERGE. Crowd & fuel = spec net as a % of its 1-yr extreme. · since = the secondary move since the close.

LME weekly COTR, Investment Funds bucket (the managed-money equivalent) — the read is the fund net move vs the same-week price. Mind the calendar: the LME COTR is as of Fri 31-Jul (released the next Tue), so these rows use the London COT week, Fri 24-Jul→31-Jul, with · since = 31 Jul→latest. Crowd & fuel = fund net as a % of its 1-yr extreme. The complex is confirming and physically tight: LME aluminium stocks are at their lowest level this century as consumers draw down exchange inventory, and zinc closed at its highest since 2022 — which vetoes every short here. Zinc (+98 ⚑) and aluminium (+66) are confirming holds with no fuel left, nickel is the odd one out (the biggest fund add of the week went INTO a −0.7% fall), and lead stays closed after last week’s squeeze call.

Source: LME MiFID COTR (per-metal weekly XLSX), Investment-Funds long/short totals; net = funds long − short, rel-to-max over the trailing ~52 weeks. Tin omitted (thin — 87 fund holders); LME copper omitted (COMEX copper is tracked above, for the Sec-232 tariff story). Prices: LME 3-month (TradingView).

CFTC reports each currency, not the dollar — so − = specs are SHORT that currency (EUR −100 = record-short the euro). We translate every line into the tradeable pair so you never convert in your head: short a currency quoted XXX/USD = SELL the pair (short GBP → SHORT GBP/USD); short one quoted USD/XXX = BUY the pair (short CAD → LONG USD/CAD). A whole board of currency-shorts = long USD (the USD row is the implied mirror, now +$37.2bn after −$12.0bn of covering). The extreme drained — unevenly, and that is the trade. Seven of the eight G10 shorts covered this week, led by the yen’s outright capitulation (+117,939, rel −100 → −28) after a rare JOINT US–Japan intervention — which by our own written exit CLOSES the USD/JPY short. Two positions did not follow: EUR still sits at −80 (covering, fuel intact → stay LONG EUR/USD) and CAD was ADDED to at a fresh −100 (−179,095, the largest single position on the board, paid nothing) → ▸ SHORT USD/CAD on a close below its 200-day at ~1.3855. The line for the whole board is the DXY 200-day at ~99.18, which price is now scraping at 99.60.

USD row = implied aggregate dollar position (−Σ of the others’ $bn): specs are net long $37.2bn vs the basket (−$12.0bn on the week) — the record +$49.2bn long now bleeding out (DXY 99.60, below the 20- and 50-day, RSI 36, scraping its 200-day at ~99.18). Seven of eight G10 shorts have started covering; only CAD was added to.

CFTC Traders-in-Financial-Futures. Rates read via Asset Managers (real money; the Lev-Fund net is the basis trade, not directional). Equities/VIX via Leveraged Funds. Each row reads the positioning move against the same-week price/yield move (28 Jul→4 Aug), then the move since.