@sdelsad/commodity-desk-daily 1.0.72 → 1.0.74

package/covered.md

@@ -26,3 +26,4 @@ Running log. Read before writing a new episode: avoid repeating material, and on
 - **Ep 22** (Fri) — *Risk Management and Why Hedges Are Never Perfect*: Ep 22 - Risk management and why hedges are never perfect: the six channels a hedge leaks through once flat price is removed - basis, timing, quality, currency, cross-hedge and quantity - each quantified on one cargo; worked example 27,000 t of 46 percent protein soybean meal sold CFR Rotterdam at 402 EUR/t for November arrival at 1.1750 = 472.35 USD/t, freight 38.00 and finance-insurance-outturn 6.00, planned purchase FOB New Orleans at the October board plus 8.00 per short ton with the board at 368.70, 27,000 t x 1.102311 = 29,762.397 short tons = 297.62 lots rounded up to 298, planned margin 13.11 USD/t = 353,954 dollars; the six leaks - basis -178,574 (paid plus 14 not plus 8), timing -89,287 (hedged October while the physical priced against December at 3.00 over), quality -129,600 (outturn 45.2 against 46.0 on a 6.00 per point protein allowance = 4.80/t), currency -162,810 (10,854,000 euros left unsold as EURUSD went 1.1750 to 1.1600, of which 19,537 was forward points and 143,273 avoidable), cross-hedge -94,500 (FFA index route returned 2.00/t against actual freight up 5.50/t), quantity -628 (37.603 short tons of excess futures on a 16.70 board fall) - total leakage 655,399 for a realised -301,445 or -11.16 USD/t against a planned +13.11, with the board contributing nothing; TRADER/TREASURY dialogue on forward points where minus eighteen points is an adjustment to spot and not a price; three limits over a physical desk - position in tonnes and lots by month, loss as VaR and a stress number, liquidity as days to liquidate - plus the usually unwritten concentration limit on counterparty, port and origin; value at risk worked on 2,000,000 bu of corn basis at 1.2 c/bu daily volatility, 1.645 sd = 1.97 c/bu = 39,480 dollars one-day 95 percent, scaled by root 21 = 180,920, against a 25 c/bu three-week Gulf basis stress = 500,000 dollars, 2.8 times the monthly and 12.7 times the daily figure; three reasons VaR flatters a physical book - marks are assessments rather than trades, no screen exists so the liquidation horizon is fiction (ep 15 days-to-liquidate callback), and the covariance matrix is estimated on quiet days so correlations break on the resolving event (ep 16 callback); risk reports to the CFO because the owner of the P and L cannot also mark and size it, and a limit is a statement about fundable mistake size rather than a forecast; depth point that a chosen risk has a nameable price while an inherited risk does not, that inherited risk is deleted or converted rather than hedged harder, and that the irreducible residue - lot rounding and cross-hedge basis - belongs in the quoted margin rather than the risk report. Pulse: Thu 17 Sep settles Dec corn 530.50 -3.75, Nov beans 1319.75 -0.75, Oct meal 368.70 +7.80, Oct bean oil 68.68 -51 pts, Dec Chi wheat 727.00 -3.75, with Dec KC 799.50 +3.25 from Wednesday; meal up about 7 percent over six sessions while oil lost 2 percent, a product-spread transfer inside a crush that barely changed, used as the bridge into the lesson; export sales w/e 10 Sep beans 1,702.0 kt mostly China and unknown, corn 1,026.7 kt a three-week low read as South American competition, wheat 325.9 kt to the Philippines and Mexico; corn harvest 6 percent, winter wheat planting 12, rice harvest 50; FranceAgriMer cut French soft wheat exports outside the EU to 6.3 Mt from 7.0 and intra-EU to 7.1 from 7.4, soft wheat ending stocks 3.01 Mt from 3.65, maize ending stocks 1.46 Mt -26 percent y/y and the smallest crop since the 1970s; GEO escalation of the Black Sea thread from the scope-of-the-truce reading in ep 21 to the NEGOTIATION CHANNEL - Ukraine floated talks with Russia on restarting grain exports from both countries and Russia called the proposal impractical while shipping stays very limited at both origins, transmission named as insurance rather than announcement because a corridor reopens when underwriters reprice war risk on hulls and cargo and not on a communique, evidenced by Chicago wheat falling on the day the headline printed; the policy channel the board did pay for was Washington-Beijing with about 1 Mt of US beans bought in the week, roughly half of a 25 Mt per year commitment running to 2028, a 10 percent Chinese tariff on US farm goods in play and a 24 September meeting.
 - **Ep 23** (Mon) — *Trade Finance, Contracts and Counterparty Risk*: Ep 23 - Trade finance, contracts and counterparty risk: working capital as the binding constraint on a merchant's size rather than a cost line, worked on one Panamax of corn 66,000 t at Dec 527.50 plus 70 FOB Gulf = 597.50 c/bu, 66,000 t x 39.368 = 2,598,288 bu hedged 520 lots, cargo value 15,524,771 dollars leaving on the day the bill of lading is issued; the cash-flow timeline load day 0, presentation +2, document check +3, letter of credit at 30 days sight running from acceptance = 35 days out, financing 35 days at 6 percent = 90,562 dollars = 3.49 c/bu against an 8 c/bu execution margin so 44 percent of the trade consumed by the calendar, plus a five-day discrepancy at 0.50 c/bu = 12,937 dollars for a wrong certificate of origin; the unit moment on at 30 days sight - sight is the bank's acceptance of conforming documents so the tenor starts on a date the bank controls, from the bill of exchange on which the party owing the money wrote the day he first saw the draft, making a document error interest rather than administration; the borrowing base as a continuously resized secured line, advance rates 85 percent hedged inventory, 90 percent insured receivable from an approved buyer, 0 percent unhedged stock, 0 percent receivable over 90 days, so 85 percent of 15,524,771 = 13,196,055 of bank money against 2,328,716 of equity per cargo, meaning 50m dollars of trading equity carries 21 such cargoes at once and not 22; the depth point that the advance rate is a property of the paperwork rather than of the grain, so a buyer downgrade or a sanction removes a receivable from the base overnight and the line shrinks while the position does not, which is why a credit event somewhere unrelated shows up on a desk as a forced seller of something liquid it liked, and why financing capacity is an edge requiring years of audited inventory and clean receivables rather than a hiring decision; the paper layer - GAFTA forms for grains and feed and FOSFA for oils, oilseeds and meals, traded as a form number plus a handful of variables with everything else settled law, and the default clause closing an unperformed contract at the market price ruling on the day of default; washout computed as the consensual version of that clause, 30,000 t soft red winter wheat FOB Gulf December shipment sold at Dec Chicago +85, 30,000 t x 36.744 = 1,102,320 bu hedged 220 lots, identical parcel now quoted Dec +62, settlement 23 c/bu x 1,102,320 bu = 253,534 dollars paid by buyer to seller with the flat price cancelling out entirely because both legs reference the same December board, so a washout is a pure basis settlement and the cleanest proof that the differential is the trade (ep2 callback); BUYER/SELLER dialogue arguing two cents of basis = 22,046 dollars without either side naming a price; the trap that the washout extinguishes the physical obligation and does not touch the futures, so an unlifted long of 220 lots leaves you outright long 1,102,320 bu and Friday's 12.75c move is 140,546 dollars, more than half the settlement just collected, on a market you never had a view on; strings and circles - each contract bilateral, a circle settling differences in cash while the physical never moves, and a failure in the middle not closing the chain up, so one insolvency in a string of six produces five separate disputes rather than one gap; the depth point that a default clause gives a calculated number and a right to arbitrate rather than cash, converting price risk into a legal recovery that cannot be sized, funded or hedged, arbitration taking months and enforcement of an award taking a further year in the jurisdiction where the buyer keeps assets, which is why the binding limit on a physical desk in a crisis is the credit limit and never the position limit (ep22 concentration-limit callback). Pulse: Fri 18 Sep settles Dec corn 527.50 -3.00, Nov beans 1303.50 -16.25, Dec Chi wheat 714.25 -12.75, Dec KC 783.75 -10.75, Dec MGE 741.25 -11.25, Oct meal 354.60 -14.10, Oct bean oil 67.70 -98 pts; meal -3.82 percent against corn -0.57 percent after setting a new two-year high in every session of that week, read as a crowded long unwinding in one session rather than any change in the meal balance sheet; beans still higher on the week with China buying 111,000 t of US beans before the Friday open and Sinograin auctioning 543,000 t of imported beans out of state reserve on Tuesday 22 Sep, reserve selling read as making physical room to buy ahead of the 24 Sep Washington meeting, with 30 Sep Grain Stocks and Small Grains behind it; GEO escalation of the Black Sea thread from ep22's negotiation-channel and insurance-repricing reading to the UNDERWRITERS ACTUALLY MOVING - the Joint War Committee in London extended its listed war-risk areas on 16 Sep to almost the entire Black Sea, excluding only the territorial waters of Turkey, Georgia, Bulgaria and Romania, effective 19 Sep, a listing being a bill rather than a ban because it triggers separate war-risk cover priced vessel by vessel, with cover on those calls having gone from about 0.2 to about 1 percent of hull value within weeks after four vessels were hit in twenty days in August and bulk carrier availability at affected ports down 21 percent in a month, transmission named as insurance then freight then origin differential with flat price reached last if at all, evidenced by Chicago wheat falling 12.75c on a day of bullish-sounding headlines; Ukraine's farm unions asking their government the same week for a vessel insurance mechanism and for lending secured on grain in certified warehouses, used as the bridge into the trade-finance lesson
 - **Ep 24** (Wed) — *Data Sources and Positioning*: Ep 24 - Data sources and positioning: the two axes that rank any data series - latency (gap between the period described and publication) and distance (how far the measured thing sits from the question), with the trap that they run in opposite directions so the fastest series in a category is usually the least relevant; the export calendar laid out as a latency table - flash sale same or next day above 100,000 t to a single buyer, export inspections 1 day measuring grain loaded against business sold weeks earlier, crop progress 1 day as enumerator judgment, COT 3 days at release, weekly export sales 8 days measuring new commitments by destination, customs data about 5 weeks and definitive; worked as a question test - did Chinese demand improve last week is answered by the 8-day-old sales number and not by the 1-day-old inspections number because inspections count execution and sales count appetite; revision as a third axis with export sales revised and crop progress rewritten by the following week, so the first print is a draft the market trades and conviction is sized to the series rather than the headline; the fashionable sources ranked the same way - satellite measuring greenness not yield and failing in the years that matter, a crop tour as a once-a-year point estimate with no time series, Santos vessel line-ups as daily nominations rather than loadings; the COT unit moment - report taken as of Tuesday close and published Friday 3:30pm New York so it is 3 days old on release and 8 by the next Wednesday, everyone receiving the identical photograph at the same second so there is no edge in the level; managed money net long 426,842 corn contracts converted to physical at 5,000 bu a Chicago lot = 2,134,210,000 bu = about 54.2 Mt or roughly an eighth of a US crop held by people who never take delivery, beans +244,710 (-21,321 w/w) and Chicago wheat flipped back to net short -4,706 on a swing of nearly 9,000 in seven days; TRADER/BROKER dialogue on who is left to buy, in which neither party names a direction; the central distinction that positioning is a risk indicator and never a signal - a large net long has never obliged a market to fall and crowded markets stay crowded for months, but a heavily long market reacts violently to bad news and barely at all to good news because the marginal buyer is gone, so the asymmetry is tradeable and the direction is not; reading both sides because every long faces a commercial short holding physical, and under stress the commercial delivers while the fund sells; the crowded long unwinding worked on real tape - Oct soybean meal made a new two-year high in every session of the week of 14 Sep then fell 3.8 percent on Fri 18 Sep against corn -0.5 percent with nothing in the balance sheet changing, a 100 short ton contract so 14.10 dollars off = 1,410 per lot and 2,000 lots = 2.82m dollars lost in an afternoon on no news, defining a crowded position as one whose exit has no natural buyer rather than one that is wrong; the physical consequence worked on a desk long 2,400,000 bu Gulf corn basis hedged short 480 lots (480 x 5,000 = 2,400,000 bu) - board -30c gives the short futures +720,000, Gulf cash -18c gives the physical -432,000, net +288,000 = basis +12 c/bu, the flat price contributing nothing because a fund liquidation is transacted entirely on a screen while the exporter with a vessel arriving cannot reprice his need at the same speed (ep2/ep19/ep21 basis-is-the-trade callbacks); the depth point that the information edge is speed (now milliseconds and bought rather than earned) and interpretation, never secrets, since every desk gets inspections, sales and the COT in the same second, and that the only genuinely private dataset is a merchant's own flow - what it was shown, what it bid on and lost and by how much, which buyer went quiet - so a tender lost by a dollar a tonne says more than any published series and good houses log and review their own flow as data rather than letting it evaporate as gossip. Pulse: Tue 22 Sep settles Dec corn 536.75 -6.25, Nov beans 1325.50 -2.50, Dec Chi wheat 717.25 -9.50, Dec KC 781.25 -13.25, Dec MGE 736.25 -10.00, Dec oats 414.50 -1.75, Nov canola 824.30 C$/t -10.90; Monday export inspections w/e 17 Sep total grain 3.03 Mt against 2.71 Mt prior with corn 1.94 Mt, beans 759,193 t, wheat 335,253 t, used as the pulse-to-lesson bridge because strong loadings are execution of business sold weeks ago rather than appetite, and corn fell anyway with wheat hardest; board bid for a fortnight on expected Chinese demand news around the 24 Sep Washington meeting, framed as an expectation being a position rather than a fact, with 30 Sep Grain Stocks and Small Grains a week out; GEO escalation of the Black Sea thread from ep23's underwriters-actually-moving to the PREMIUM DECAYING INTO THE BASELINE - the sequence insurance then freight then volumes is now complete with both Russian and Ukrainian export paces running more than half below last year, so the market has stopped paying for the headline because the shortfall is in the baseline, evidenced by Chicago wheat losing 9.5c on a day of grim regional reports, with the asymmetry named that only a change in the LEVEL of disruption moves price now and a negotiated reopening would move it hard because nothing in the current price assumes one.
+- **Ep 25** (Fri) — *China, and Reading Market Colour*: Ep 25 - China and reading market colour: the distinction between a quantity instrument and a price instrument, worked on Thursday's prints - one Panamax of 66,000 t of soybeans is 2,425,104 bu or 485 Chicago lots, Nov CBOT 1317.50 plus the Gulf premium of 82 over = 1399.50 c/bu = 13.9950 dollars a bushel x 36.744 = 514.23 USD/t FOB, plus Gulf-Qingdao Panamax freight of 75.95 = 590.18 CFR, grossed up by 13 percent (3 percent Chinese MFN duty of 17.71 plus 10 percent retaliatory duty of 59.01) = 666.90 USD/t landed, against which the Brazilian indifference price is found backwards, 666.90 divided by 1.03 = 647.48 CFR less Santos-Qingdao freight of 56.53 = 590.95 FOB Santos, so Brazil can charge 76.72 USD/t or 209 c/bu more than the Gulf and still be the cheaper bean at a Chinese plant while the entire US export premium under discussion is 82 c/bu, making the policy line two and a half times the market line; the ad valorem unit moment - a percentage of cargo value rather than a fixed amount per tonne, so the duty grows in a rally, versus a specific duty which does not; the depth point that an ad valorem duty is neutral inside an origin and brutal between origins - an eight dollar freight rally requires exactly an eight dollar FOB fall because freight and price sit inside the same duty base, while a ten-point gap in duty rate is charged on one cargo's whole value and not the other's; the VAT footnote that input VAT credits against output VAT on meal and oil so it broadly cancels in an origin comparison while the duty does not; the two Chinese buyers - a commercial crusher running a derived bid marked backwards from meal and oil in yuan at his own plant gate net of conversion cost, which dies when the destination crush margin closes and can stay dead for weeks, against a state buyer managing days of cover against a policy commitment with the crush margin absent from its objective, so it buys into a negative crush and into a rally, making a China-bought headline uninformative until the buyer type is known; reserve rotation read the right way round - Sinograin auctioned 543,000 t of 2022-24 crop imported beans on 22 Sep for Nov-Jan delivery and sold 338,674 t at 4,280-4,450 yuan/t with a further 514,000 t offered on 28 Sep across twelve provinces, and selling old beans empties the tanks a new import needs so days of cover are roughly unchanged, with the 62 percent clearance rate as the real demand report because a third went unsold and domestic buyers declined discounted four-year-old beans; market colour as mostly about who rather than how much - a BROKER/TRADER dialogue in which the shape of the order identifies the buyer (four cargoes across two months with no argument about the level is a reserve shape and not a crusher shape), the trader asks whose the bid is rather than what the market is because the level was already on his screen, and nothing offered above plus ninety is the information rather than its absence, since a bid level with nothing showing against it means replacement value has moved above the last print and marking a book to that print is how a hedged position quietly stops being hedged. Pulse: Thu 24 Sep settles Dec corn 527.50 -1.50, Nov beans 1317.50 -0.50, Dec Chi wheat 707.00 -1.50, Oct meal 376.10 +5.80, Oct bean oil 66.93 -32 pts, so meal +1.6 percent and oil -0.5 percent with everything else inside a cent and a half on the day of the Washington meeting the trade had waited a fortnight for; the summit print itself as the escalation of the thread named forward in eps 23 and 24 - truce extended to 10 January 2027, the 25 Mt per year soybean commitment to 2028 reaffirmed with 9.858 Mt of 2026/27 commitments already booked as at 10 Sep, and the 10 percent Chinese retaliatory tariff on US farm goods left in place; weekly export sales w/e 17 Sep soft - beans 582,400 t of which China 311,800 t or 54 percent, corn 838,300 t, wheat 267,600 t -18 percent w/w and 13 percent below the four-week average; Gulf soybean bid 81-82 over November, up half a cent; 30 Sep Grain Stocks and Small Grains next; GEO rotation off the Black Sea thread to the POLICY-RETALIATION CHANNEL - a quantity pledge versus a price instrument, transmission named as state buyers executing tonnage they are indifferent to while commercial buyers who would have to bid the basis are priced out by the duty, evidenced by half a cent on the Gulf bid against a reaffirmed 25 Mt pledge, with the unpriced asymmetry that removing the 10 percent would be worth far more to the US basis than any increase in pledged tonnage.

package/ep25.md

@@ -0,0 +1,228 @@
+# Market pulse
+
+**The meeting the trade had waited a fortnight for moved soybeans half a cent.**
+
+| Contract | Thursday 24 Sep settle | Change |
+|---|---|---|
+| Dec corn | 527½ c/bu | −1½ c |
+| Nov soybeans | 1317½ c/bu | −½ c |
+| Dec Chicago wheat | 707 c/bu | −1½ c |
+| Oct soybean meal | $376.10 /short ton | +$5.80 |
+| Oct soybean oil | 66.93 c/lb | −32 pts |
+
+The Washington meeting produced what was expected and very little more. The trade truce now runs to 10 January 2027. The commitment to buy 25 Mt of US soybeans a year, running to 2028, was reaffirmed, and just under 10 Mt of that — 9.858 Mt of 2026/27 commitments as at 10 September — is already on the books. The 10% Chinese retaliatory tariff on US farm goods stays in place.
+
+Weekly export sales for the week to 17 September came in soft. Soybeans booked 582,400 t, of which China took 311,800 t, or 54%. Corn did 838,300 t. Wheat did 267,600 t, down 18% on the week and 13% below its four-week average. The Gulf soybean bid finished at 81–82 over November, up half a cent on the day.
+
+Meal was the only real mover, up 1.6% while oil lost 0.5%, continuing the product-spread transfer inside a crush that has barely changed. Next on the calendar is the 30 September Grain Stocks and Small Grains pair.
+
+```chart
+{"type":"bar","unit":"% change","title":"Summit day, priced in advance",
+ "caption":"The one market that moved on the day of the meeting was the one the meeting was not about. Beans finished four hundredths of a percent lower on a reaffirmed 25 Mt commitment, because the price had already been paid for the expectation.",
+ "source":"CBOT settlements, Thursday 24 September 2026, versus Wednesday 23 September, from Brownfield closing futures",
+ "x":["Corn","Beans","Chi wheat","Meal","Bean oil"],
+ "series":[{"name":"24 Sep","values":[-0.28,-0.04,-0.21,1.57,-0.48]}]}
+```
+
+## The geopolitical read
+
+Today the mechanism is not the Black Sea but a tariff line that did not move, and the distinction that matters is between a quantity instrument and a price instrument. A pledge says how much will be bought. A tariff says at what price, and by whom.
+
+Twenty-five million tonnes a year can be executed by a state agency that does not need a margin. Ten percent of the value of a cargo cannot be absorbed by a commercial crusher who does. So the tonnage arrives while the differential stays where it was: the flow is bought by buyers who are indifferent to it, and the market that sets the basis never has to bid. Half a cent on the Gulf bid was the arithmetically correct response to a reaffirmed pledge that came with the duty attached.
+
+The asymmetry to hold on to is the one nobody is positioned for. Removal of the 10% would be worth far more to the US basis than any increase in the pledged tonnage, because it would put commercial Chinese demand back into a market that currently only sees state demand.
+
+# Key takeaways
+
+- A quantity pledge and an ad valorem tariff are different instruments and are not measured in the same units. On Thursday's numbers the tariff was worth 209 c/bu of origin advantage to Brazil, against a total US export premium of 82 c/bu.
+- There is no single Chinese buyer. A crusher's bid is derived backwards from meal and oil at his own plant gate and dies when the margin dies. A state bid is derived from days of cover and a policy commitment and does not.
+- Which means a headline saying "China bought three cargoes" is not information until you know which China bought them.
+- An ad valorem duty is neutral inside an origin and brutal between origins. A dollar of freight costs exactly a dollar of FOB price, because both sit inside the same duty base — but a ten-point gap in duty rate is charged on one cargo's whole value and not on the other's.
+- A reserve auction is a demand report before it is a supply event. Selling old crop empties the tanks a new import needs, so rotation leaves days of cover roughly unchanged.
+- The clearance rate is the number worth keeping. Sinograin cleared 62% of what it offered on 22 September, which says more about demand at the Chinese plant gate than any purchase headline.
+- Market colour is mostly about who, not how much. The level is already on the screen; what the broker adds is the shape of the buyer.
+- When a level is bid and nothing shows against it, replacement value has moved above the last print and the last print is stale. Marking a book to that print is how a hedged position quietly stops being hedged.
+
+# Vocabulary
+
+| Term | What it means |
+|---|---|
+| **Ad valorem tariff** | A duty charged as a percentage of the declared value of a cargo rather than as a fixed amount per tonne, so the cash cost of the duty rises and falls with the market price |
+| **Specific tariff** | A duty charged as a fixed amount per unit of quantity, which is invariant to price and therefore falls in relative terms during a rally |
+| **MFN duty** | The most-favoured-nation rate a country applies to imports from any WTO member with which it has no preferential agreement, which for soybeans entering China is 3% |
+| **Landed cost** | The all-in delivered cost of a cargo at the buyer's port, being the CFR value plus duties and port charges, and the only number on which two origins can honestly be compared |
+| **Indifference price** | The price at which a buyer is economically neutral between two origins or two options, computed by holding the landed cost equal and solving backwards for the FOB |
+| **State buyer** | A government-owned purchaser such as a reserve agency, whose bid is set by a stockholding or policy target rather than by a processing margin, and which is therefore insensitive to the crush |
+| **Commercial buyer** | A privately accountable buyer — a crusher, a feed mill, an importer — whose bid is derived from the margin it can compute on its own products and who stops buying when that margin closes |
+| **Reserve rotation** | The practice of auctioning older stock out of a state reserve while buying newer stock in, which changes the age of the inventory rather than its size and frees tank space for imports |
+| **Auction clearance rate** | The proportion of a reserve offering actually sold, read as a direct measure of domestic willingness to pay and therefore of demand rather than of supply |
+| **Destination crush margin** | The processing margin available to a crusher in the importing country, computed in local currency from local meal and oil values against landed bean cost, and the leading indicator of whether commercial import buying continues |
+| **Derived bid** | A bid computed backwards from the value of the products a buyer will make, as distinct from a bid set by a budget, a target or a mandate |
+| **Nothing behind it** | Broker shorthand that a quote has no supporting interest at or near the level, meaning the print on the screen cannot be repeated in size |
+| **Subject** | A qualifier making a quote or agreement conditional on something not yet done — approval, allocation, a licence — and therefore not yet firm |
+
+# Quiz
+
+**Q1.** Thursday's landed-cost chain at a Chinese crusher, one Panamax of 66,000 t of soybeans for November shipment, runs like this. November Chicago settled at 1317½ c/bu. The US Gulf premium is 82 c over November. A tonne of soybeans is 36.744 bushels. Panamax freight settled at $75.95/t on the US Gulf–Qingdao route and $56.53/t on the Santos–Qingdao route. China charges a 3% MFN duty on soybeans from any origin, plus a further 10% retaliatory duty on US origin, both ad valorem on the CFR value. Now suppose Gulf–Qingdao freight rallies from $75.95 to $84.00 while Santos–Qingdao is unchanged and the retaliatory duty stays in place. By how much, in dollars per tonne, must the US Gulf FOB price fall to leave that crusher exactly as indifferent between the two origins as he was before?
+
+**Q2.** A Chinese crusher's margin on November-arrival beans is negative, and he has stopped bidding. In the same week Sinograin books three cargoes of US beans. Explain why those two facts are not in contradiction.
+
+**Q3.** A market's managed-money net long has just been halved in a single week. What does that do to the market's likely reaction to a piece of bad news?
+
+**Q4.** You sold 27,000 t of soybean meal CFR Rotterdam in euros and left €10,854,000 of the proceeds unsold. EURUSD then moved from 1.1750 to 1.1600. What did that cost in dollars?
+
+**Q5.** Conversion drill. A container is loaded with 320 coffee bags of 69 kg each. What is the cargo in pounds?
+
+# SOLUTIONS (spoilers)
+
+**A1.** $8.05 per tonne — exactly the rise in freight, and not a cent more.
+
+Build the chain first, because the indifference level has to be established before anything can be flexed.
+
+The US FOB price is the board plus the premium: 1317.50 + 82 = 1399.50 c/bu, or $13.9950/bu. Converting to metric multiplies by the bushels in a tonne: 13.9950 × 36.744 = $514.23/t FOB Gulf. For the record on the cargo itself, 66,000 t × 36.744 = 2,425,104 bu, which is 485 lots of 5,000 bushels with 104 bushels left over.
+
+| Line | US Gulf origin |
+|---|---|
+| FOB | $514.23 /t |
+| Freight, Gulf–Qingdao | $75.95 /t |
+| CFR Qingdao | $590.18 /t |
+| Duty, 13% ad valorem | $76.72 /t |
+| **Landed cost** | **$666.90 /t** |
+
+Now find the Brazilian FOB that matches it. Brazil pays the 3% MFN duty and nothing else, so its landed cost is its CFR grossed up by 1.03. Setting that equal to $666.90 and working backwards: 666.90 ÷ 1.03 = $647.48/t CFR, less $56.53 of freight, gives $590.95/t FOB Santos. That is the indifference price.
+
+Flex the freight. Gulf–Qingdao goes to $84.00. Brazil is untouched, so the crusher's alternative still lands at $666.90. The US cargo must therefore still land at $666.90, which means its CFR must still be 666.90 ÷ 1.13 = $590.18/t. Freight now takes $84.00 of that, leaving $506.18/t of FOB. The fall required is 514.23 − 506.18 = **$8.05/t**, which in Chicago terms is 8.05 ÷ 36.744 = 21.9 c/bu, taking the Gulf premium from +82 to roughly +60 over November.
+
+The trap is to gross the freight rise up by the duty, and answer $9.10. That is wrong, and the reason is worth more than the arithmetic. An ad valorem duty is charged on the CFR value, and freight and FOB price are both inside it. The duty therefore scales the two identically, and a dollar of freight costs exactly a dollar of price. What the duty does amplify is the *difference between origins*: the ten-point gap in rate applies to the whole value of one cargo and not at all to the other's, which is where the $76.72 came from in the first place.
+
+**A2.** Because the two buyers are solving different problems, and only one of them has a margin in its objective function.
+
+The crusher's bid is derived. He marks meal and oil at his own plant gate in yuan, deducts his conversion cost, and the residual is the most he can pay for a bean. When that residual falls below the offer, he does not bid — and he can go without bidding for weeks, because his alternative is to run at reduced rates or stop.
+
+Sinograin does not crush. It stores. Its objective is days of cover against a policy commitment, and the crush margin appears nowhere in that calculation. It will buy into a negative crush and it will buy into a rally, because the number it is managing is a tonnage target, not a P&L.
+
+The trap the question is testing is the habit of treating "China bought" as a demand signal. State buying tells you about policy. Commercial buying tells you about the crush. Only the second one carries information about consumption, and only the second one has to pay the basis to get its cargo — which is precisely why the Gulf bid moved half a cent on a week of confirmed state purchases.
+
+**A3.** It makes the reaction to bad news smaller, not larger.
+
+A crowded long is dangerous because its exit has no natural buyer. Bad news into a heavily long market produces a violent move, because the people who would ordinarily step in have already bought and the only flow left is liquidation. Halve the net long and you have removed most of that fuel: much of the selling that bad news would have triggered has already happened, and there is now room above the market for a marginal buyer to reappear.
+
+The symmetry is worth stating. The same market, having shed half its long, becomes *more* reactive to good news, because there is a buyer left to do the reacting. Positioning never tells you direction. It tells you the shape of the response to whatever arrives, and the answer flips depending on which side has been cleared out.
+
+**A4.** $162,810.
+
+Unsold proceeds in a foreign currency are an outright position in that currency, whether or not anyone on the desk decided to take one. The exposure is €10,854,000. The rate moved 1.1750 − 1.1600 = 0.0150 dollars per euro against the holder, so 10,854,000 × 0.0150 = **$162,810**.
+
+The part that matters on a desk is what happens to that figure in attribution. It is not a trading loss and it is not a basis loss: it is a currency leak on a cargo whose commodity risk was fully hedged. A portion of it — the forward points between the sale date and the settlement date — was never avoidable, because selling euros forward was always going to cost the interest differential. The rest was avoidable and is the number the post-mortem should carry. A hedge on the board and a hedge on the protein spec do not make a book flat when the invoice is denominated in a currency nobody sold.
+
+**A5.** 48,576 lb by the fast method, 48,678 lb exactly.
+
+The cargo is 320 × 69 = 22,080 kg. Kilograms to pounds is double it, then add ten percent: 22,080 doubled is 44,160, ten percent of that is 4,416, and the sum is **48,576 lb**. The true factor is 2.20462, which gives 48,678 lb — the fast method runs about 0.2% light, which on a container is a rounding error and on a Panamax is not.
+
+# The written edition
+
+## Two governments, one number, half a cent
+
+A trade meeting in Washington on Thursday reaffirmed a Chinese commitment to buy 25 Mt of US soybeans a year through 2028 and extended a tariff truce to 10 January 2027. November soybeans settled half a cent lower.
+
+That is not a market failing to react. It is a market reacting correctly, and understanding why requires separating two instruments that get discussed as though they were one. A **quantity pledge** commits a buyer to volume. A **tariff** sets the price at which that volume is worth buying, and to whom. They are not measured in the same units, and on Thursday's numbers they were not even the same order of magnitude.
+
+## Ad valorem, and why the word does the work
+
+A duty can be charged two ways. A **specific** duty is a fixed amount per tonne: twenty dollars is twenty dollars, whatever the market does. An **ad valorem** duty is a percentage of the declared value of the cargo, which means its cash cost rises with the price of the goods. It was designed that way so that customs revenue keeps pace with inflation. The consequence for a trader is that the duty gets larger exactly when the market is rallying.
+
+China's ordinary MFN duty on imported soybeans is 3%, applied to any origin without a preferential agreement. On US origin there is a further 10% retaliatory duty. Thirteen percent, charged on the CFR value landed at the Chinese port.
+
+## The landed cost, both origins
+
+Take Thursday's prints and land one Panamax — 66,000 t, or 2,425,104 bushels, hedged with 485 Chicago lots — at Qingdao for November shipment.
+
+| Line | US Gulf | Brazil (indifference) |
+|---|---|---|
+| FOB | $514.23 /t | $590.95 /t |
+| Freight to Qingdao | $75.95 /t | $56.53 /t |
+| CFR | $590.18 /t | $647.48 /t |
+| Duty | 13% = $76.72 | 3% = $19.42 |
+| **Landed** | **$666.90 /t** | **$666.90 /t** |
+
+The US side is built forwards. November Chicago at 1317½ plus 82 over is 1399½ c/bu, and 13.9950 × 36.744 = $514.23/t. Add the Gulf–Qingdao Panamax rate of $75.95, gross up by 1.13, and the cargo lands at $666.90.
+
+The Brazilian side is built backwards, because the interesting question is not what Brazil charges today but what it *could* charge. Hold the landed cost equal, divide by 1.03, take off the Santos–Qingdao freight of $56.53, and the answer is $590.95/t FOB.
+
+**Brazil can charge $76.72 a tonne more than the Gulf and still be the cheaper bean at a Chinese plant.** That is 209 cents a bushel. The entire US export premium — the number the whole basis market argues about, the one that moved half a cent on Thursday — is 82 cents. The policy line is two and a half times the market line.
+
+```chart
+{"type":"waterfall","unit":"USD/t","title":"Landing one Panamax at Qingdao",
+ "caption":"The retaliatory duty alone adds fifty-nine dollars a tonne, against an export premium of thirty. The tariff is the larger half of the trade, and no basis move can offset it.",
+ "source":"Worked example, episode 25, from Nov CBOT 1317.50 c/bu and Gulf basis +82 (24 Sep 2026), Panamax P7 route $75.95/t, Chinese MFN 3% plus 10% retaliatory duty",
+ "steps":[{"label":"FOB Gulf","value":514.23,"kind":"base"},
+          {"label":"Freight","value":75.95},
+          {"label":"MFN duty 3%","value":17.71},
+          {"label":"Retaliatory 10%","value":59.01},
+          {"label":"Landed Qingdao","kind":"total"}]}
+```
+
+### The subtlety in the arithmetic
+
+Suppose Gulf freight rallies eight dollars a tonne. How far must the Gulf FOB price fall to keep that crusher indifferent?
+
+Eight dollars. Not nine. Freight and FOB both sit inside the CFR value on which the duty is charged, so the 13% scales them identically, and a dollar of freight costs exactly a dollar of price.
+
+What the duty amplifies is not the components but the comparison. A ten-point gap in duty rate is charged on the whole value of one cargo and on none of the other's, which is where the $76.72 came from. An ad valorem tariff is neutral inside an origin and brutal between origins.
+
+One honest footnote: China also levies VAT on imported soybeans, but a crusher credits input VAT against the output VAT on his meal and oil, so it broadly cancels in an origin comparison. The duty does not cancel. It is a cost that stays in the cargo.
+
+## There is no single Chinese buyer
+
+The reason the tariff and the pledge can coexist is that they are aimed at different buyers.
+
+A **commercial** buyer — a crusher, a feed mill — runs a **derived bid**. He marks meal and oil in yuan at his own plant gate, deducts conversion cost, and the residual is his maximum bean price. That is the **destination crush margin**, and it is the single most useful leading indicator of whether commercial Chinese import buying continues. When it closes, the bid disappears, and it can stay disappeared for weeks.
+
+A **state** buyer runs no such calculation. Sinograin is managing days of cover against a policy commitment. The crush margin is not in its objective, so it will buy into a negative crush and into a rally.
+
+Same tonnage on the screen, entirely different information content. Which is why "China bought three cargoes" is not yet a fact about demand.
+
+## The reserve, read the right way round
+
+On 22 September Sinograin auctioned 543,000 t of imported soybeans out of state stock — 2022 to 2024 crop — for November-to-January delivery. It sold 338,674 t, at 4,280 to 4,450 yuan/t. A further 514,000 t is offered on 28 September, spread across twelve provinces.
+
+The instinctive reading is that a reserve sale is bearish: supply appearing from nowhere. On a desk it reads the other way round. Selling old beans empties tanks, and empty tanks are what a new import needs. That is **reserve rotation** — old crop out through the auction, new crop in through the port, days of cover roughly unchanged.
+
+```chart
+{"type":"bar","unit":"'000 t","title":"Rotation, not liquidation",
+ "caption":"Two thirds of the 22 September offering cleared, and another half-million tonnes was put up six days later. The unsold third is the number that matters: domestic buyers would not take discounted four-year-old beans, which is soft demand at the plant gate.",
+ "source":"Sinograin auction results, 22 September 2026, and the scheduled offering for 28 September 2026",
+ "x":["22 Sep offered","22 Sep sold","28 Sep offered"],
+ "series":[{"name":"Imported soybeans","values":[543,339,514]}]}
+```
+
+The clearance rate carries the information. Sixty-two percent cleared means a third of the offering went unsold, and domestic buyers declined discounted four-year-old beans. That is a direct read on willingness to pay at the Chinese plant gate, and a better one than any purchase headline, because the bidders were spending their own money.
+
+## Market colour: mostly about who
+
+The second half of reading China — or any destination market — is language rather than arithmetic. Market colour is what the broker tells you that is not the price. The level is already on the screen. What the broker adds is the shape of the buyer.
+
+> BROKER: Someone's bid November Gulf. Plus eighty-two.
+> TRADER: Whose?
+> BROKER: Won't say. Wants four cargoes. December, January.
+> TRADER: Then it's not a crusher. A crusher takes one and argues about the level.
+> BROKER: Nothing offered above plus ninety.
+> TRADER: Nothing offered, or nothing you'd show me?
+
+Three things happen in six lines.
+
+The **shape of the order** identifies the buyer. Four cargoes across two months with no argument about the level is a reserve shape, not a crusher shape. A crusher buys what he can grind and fights over every cent, because every cent comes out of a margin he can compute.
+
+The **question asked** is not what the market is but whose it is. The trader already had the level. What he needed was the buyer's type, because that determines whether the bid will still be there tomorrow.
+
+The **absence** is the information. "Nothing offered above plus ninety" is not a gap in the broker's knowledge. If a level is bid and nothing shows against it, replacement value has moved above the last print, and the last print is now stale. A desk that marks its book to that print is carrying a position it believes is hedged and is not — the physical leg has revalued and the mark has not followed.
+
+Silence is a price. It is simply not a printed one.
+
+## The thing to carry away
+
+Two instruments, two buyers, one habit worth breaking.
+
+The tariff is bigger than the basis, and on Thursday's numbers it was bigger by a factor of two and a half. A pledge that arrives with a duty attached delivers tonnage without delivering a bid.
+
+And a tonnage headline is not information until the buyer is identified. One China computes a margin. The other executes a policy. Only one of them ever has to pay up for a cargo, and it is not the one in the headline.

package/ep25.script.txt

@@ -0,0 +1,69 @@
+Two governments met in Washington yesterday and reaffirmed a twenty-five million tonne soybean commitment. The soybean market moved half a cent. ||| 0.6
+That is not a failure of the meeting. It is the market telling you something exact about who actually buys soybeans in China. ||| 0.5
+This is Soft Commodity Trading, episode twenty-five. China, and how to read market colour. ||| 0.8
+Thursday's settles first. December corn five twenty-seven and a half, down a cent and a half. November beans thirteen seventeen and a half, down half a cent. December Chicago wheat seven oh seven, down a cent and a half. ||| 0.4
+October soybean meal three seventy-six ten, up five dollars eighty. October bean oil sixty-six ninety-three, down thirty-two points. ||| 0.5
+So meal up one and a half percent, oil down half a percent, and everything else inside a cent and a half. On the day of the meeting the trade had been waiting a fortnight for. ||| 0.6
+Here is what came out of it. The truce runs to the tenth of January. The twenty-five million tonne a year soybean commitment is reaffirmed, and just under ten million tonnes of that is already booked for this crop year. ||| 0.5
+And the ten percent retaliatory tariff on American farm goods stays. ||| 0.7
+That last sentence is worth more than the other three. ||| 0.6
+Export sales for the week to the seventeenth. Beans five hundred eighty-two thousand tonnes, of which China took three hundred and twelve thousand. Corn eight hundred thirty-eight thousand. Wheat two hundred sixty-seven thousand, down eighteen percent on the week. ||| 0.5
+And the Gulf soybean bid moved up half a cent, to eighty-two over November. ||| 0.7
+The geopolitics today is not the Black Sea. It is a tariff that did not move. ||| 0.4
+A quantity pledge and a tariff are not the same instrument. A pledge says how much. A tariff says at what price, and to whom. ||| 0.5
+Twenty-five million tonnes can be bought by a state agency that does not need a margin. Ten percent of the value of a cargo cannot be absorbed by a crusher who does. ||| 0.5
+So the flow arrives, and the differential does not pay for it. That is the transmission mechanism, and it is why half a cent was the correct answer. ||| 0.8
+Start with the buyer, because there is not one Chinese buyer. There are two, and they read completely differently. ||| 0.5
+A crusher's bid is derived. He looks at meal and oil in yuan at his own plant gate, takes off his conversion cost, and works backwards to what he can pay for a bean. If that number sits below the offer, he does not buy. He does not buy for a week if he has to. ||| 0.5
+A state bid is not derived from anything. Sinograin is managing days of cover and a political commitment. The crush margin is not in the calculation. ||| 0.5
+Same tonnage on the screen. Completely different information. ||| 0.6
+Which is why a headline saying China bought three cargoes is not yet information. ||| 0.7
+Now the tariff, because the word matters. Ad valorem means the duty is a percentage of the value of the cargo, not a fixed amount per tonne. ||| 0.4
+China's ordinary duty on imported soybeans is three percent. On American origin there is a further ten. Thirteen percent, charged on the cost and freight value landed at the port. ||| 0.5
+Which means it grows when the market rallies. A specific duty of twenty dollars a tonne is twenty dollars whatever happens. Thirteen percent of a cargo is a moving number, and it moves against you in a bull market. ||| 0.8
+Take Thursday's numbers and land one Panamax at Qingdao. Sixty-six thousand tonnes of beans. ||| 0.4
+November Chicago at thirteen seventeen and a half, the Gulf premium at eighty-two over. Thirteen ninety-nine and a half cents a bushel, free on board. ||| 0.4
+Convert. A tonne of soybeans is thirty-six point seven four four bushels, so that is five hundred fourteen dollars twenty-three a tonne. ||| 0.5
+Freight. The Panamax route from the Gulf to Qingdao settled at seventy-five ninety-five a tonne. Santos to Qingdao, fifty-six fifty-three. ||| 0.5
+So the American cargo lands, cost and freight, at five hundred ninety dollars eighteen. Add thirteen percent. Six hundred sixty-six dollars ninety a tonne. ||| 0.7
+Now price the Brazilian against it. Same destination, three percent duty, nineteen dollars forty-two a tonne cheaper on the freight. ||| 0.4
+Work backwards from the same landed number. Six hundred sixty-six ninety divided by one point oh three is six hundred forty-seven forty-eight, cost and freight. Take off the fifty-six fifty-three of freight. Five hundred ninety dollars ninety-five, free on board Santos. ||| 0.6
+Five ninety ninety-five against five fourteen twenty-three. ||| 0.5
+Brazil can charge seventy-six dollars seventy-two a tonne more than the Gulf and still be the cheaper bean at a Chinese plant. ||| 0.6
+Seventy-six seventy-two a tonne is two hundred and nine cents a bushel. ||| 0.5
+The entire American export premium, the number the whole basis market argues about, the one that moved half a cent yesterday, is eighty-two cents. ||| 0.6
+The policy line is two and a half times the market line. ||| 0.8
+There is a subtlety inside that arithmetic worth ten seconds, because it is where people get it wrong. ||| 0.4
+Suppose Gulf freight rallies eight dollars a tonne. How far must the Gulf price fall to keep that crusher indifferent? ||| 0.5
+Eight dollars. Not nine. Freight and price sit inside the same duty base, so the thirteen percent scales both of them equally, and a dollar of freight costs exactly a dollar of price. ||| 0.5
+But the ten point gap in duty between the two origins is amplified, because it is charged on one cargo's whole value and not on the other's. ||| 0.5
+An ad valorem tariff is neutral inside an origin and brutal between origins. ||| 0.8
+Which brings in the reserve, and the part of the Chinese market most people read backwards. ||| 0.5
+On the twenty-second, Sinograin auctioned five hundred forty-three thousand tonnes of imported soybeans out of state stock, two thousand twenty-two to twenty-four crop. It sold three hundred thirty-nine thousand of it, at four thousand two hundred eighty to four thousand four hundred fifty yuan a tonne. ||| 0.5
+Sixty-two percent cleared. And another five hundred fourteen thousand tonnes is offered on the twenty-eighth, across twelve provinces. ||| 0.7
+The obvious reading is that a reserve sale is bearish. Supply appearing out of nowhere. ||| 0.4
+On a desk it reads the other way round. Selling old beans empties tanks, and empty tanks are what a new import needs. ||| 0.5
+That is reserve rotation. Old crop out through the auction, new crop in through the port, and the state's days of cover roughly unchanged. ||| 0.6
+And the sixty-two percent is the number to keep. A third of the offering went unsold. Domestic buyers would not take discounted four-year-old beans even at a discount. That is soft demand at the plant gate, and it is a better read on the Chinese crush than any purchase headline. ||| 0.8
+The second half of this is language rather than arithmetic. ||| 0.5
+Market colour is what the broker tells you that is not the price. Most of it is about who is buying rather than how much, and a desk will spend real effort on that question before it quotes anything at all. ||| 0.5
+BROKER: Someone's bid November Gulf. Plus eighty-two. ||| 0.25
+TRADER: Whose? ||| 0.25
+BROKER: Won't say. Wants four cargoes. December, January. ||| 0.25
+TRADER: Then it's not a crusher. A crusher takes one and argues about the level. ||| 0.3
+BROKER: Nothing offered above plus ninety. ||| 0.25
+TRADER: Nothing offered, or nothing you'd show me? ||| 0.7
+Three things happened there. ||| 0.4
+Four cargoes across two months with no argument about the level is a reserve shape, not a crusher shape. A crusher buys what he can grind, and he fights over every cent, because every cent comes out of a margin he can compute. ||| 0.5
+Second, the trader never asked what the market was. He asked whose it was. The level was already on his screen. ||| 0.5
+Third, and this is the one that takes years. Nothing offered above plus ninety is not an absence of information. It is the information. ||| 0.6
+If a level is bid and nothing shows against it, replacement value has moved above the last print, and the last print is now a lie. ||| 0.5
+Marking a book to the last trade in a market where nobody is offering is how a hedged position quietly stops being hedged. ||| 0.6
+Silence is a price. It is just not a printed one. ||| 0.8
+What to keep from this. ||| 0.4
+A quantity pledge and an ad valorem tariff are different instruments, and the tariff is usually the bigger number. Seventy-six dollars a tonne against eighty-two cents a bushel. ||| 0.5
+There are two Chinese buyers. One computes a margin, one executes a policy. Until you know which is bidding, the tonnage tells you nothing. ||| 0.5
+A reserve auction that clears two thirds is a demand report, not a supply event. ||| 0.5
+And the hard one. When nobody is offering, that is the quote. ||| 0.7
+Next time, the Black Sea and policy shocks. Corridors, export taxes, and how a desk positions for a decision that is binary and unforecastable. ||| 0.5
+There are four questions in today's notes and a conversion drill. The first is a full landed cost problem, and it is worth twenty minutes. ||| 0.6

package/feed.xml

@@ -18,6 +18,18 @@
       <title>Soft Commodity Trading</title>
       <link>https://storage.googleapis.com/podcast-audio-2647223968/index.html</link>
     </image>
+    <item>
+      <title>Ep 25 — China, and Reading Market Colour</title>
+      <link>https://storage.googleapis.com/podcast-audio-2647223968/commodity-desk-daily/ep25.html</link>
+      <description><![CDATA[<p>Why a reaffirmed 25 Mt soybean pledge moved the market half a cent, and how a 10% ad valorem tariff hands Brazil 209 cents a bushel of origin advantage. Then the two Chinese buyers, reserve rotation, and what a broker's silence is actually quoting.</p><p><a href="https://storage.googleapis.com/podcast-audio-2647223968/commodity-desk-daily/ep25.html">Read this episode, with the charts, the glossary and the quiz &rarr;</a></p>]]></description>
+      <itunes:summary>Why a reaffirmed 25 Mt soybean pledge moved the market half a cent, and how a 10% ad valorem tariff hands Brazil 209 cents a bushel of origin advantage. Then the two Chinese buyers, reserve rotation, and what a broker's silence is actually quoting.
+
+Read this episode: https://storage.googleapis.com/podcast-audio-2647223968/commodity-desk-daily/ep25.html</itunes:summary>
+      <enclosure url="https://storage.googleapis.com/podcast-audio-2647223968/commodity-desk-daily/ep25.mp3" length="0" type="audio/mpeg"/>
+      <guid isPermaLink="false">https://storage.googleapis.com/podcast-audio-2647223968/commodity-desk-daily/ep25.mp3</guid>
+      <pubDate>Fri, 25 Sep 2026 05:00:00 GMT</pubDate>
+      <itunes:duration>0</itunes:duration>
+    </item>
     <item>
       <title>Ep 24 — Data Sources and Positioning</title>
       <link>https://storage.googleapis.com/podcast-audio-2647223968/commodity-desk-daily/ep24.html</link>

package/glossary.md

@@ -5,6 +5,7 @@ Units, conventions and desk expressions, accumulated as the show introduces them
 - **45Z** — the US clean fuel production credit, one of the two policy levers that sets American soybean oil demand  _(ep 8)_
 - **abandonment** — planted area never harvested for grain, lost to drought, flood or a switch to silage  _(ep 6)_
 - **ABCD** — the four historic majors, Archer Daniels Midland, Bunge, Cargill and Louis Dreyfus  _(ep 2)_
+- **ad valorem tariff** — a duty charged as a percentage of the declared value of a cargo rather than as a fixed amount per tonne, so the cash cost of the duty rises and falls with the market price  _(ep 25)_
 - **advance rate** — the percentage of an asset's value a lender will advance against it, high for hedged and insured assets and zero for unhedged stock or stale receivables  _(ep 23)_
 - **against actuals (AA)** — the softs market's name for an exchange for physical  _(ep 18)_
 - **anhydrous ethanol** — near-water-free ethanol blended into petrol under a mandate, taking 1.7651 kg of ATR per litre  _(ep 14)_
@@ -18,6 +19,7 @@ Units, conventions and desk expressions, accumulated as the show introduces them
 - **ATR** — Acucar Total Recuperavel or total recoverable sugar, the kilos of sugar recoverable from a tonne of cane, the unit in which Brazilian growers are paid and the unit in which a mill compares sugar against ethanol  _(ep 14)_
 - **attribution** — decomposing a finished trade into flat price, basis, calendar spread, freight, currency and financing, so the result can be explained rather than merely counted  _(ep 21)_
 - **attribution bridge** — the line-by-line reconciliation from planned margin to realised margin, which has to close to the dollar or a line is missing  _(ep 21)_
+- **auction clearance rate** — the proportion of a reserve offering actually sold, read as a direct measure of domestic willingness to pay and therefore of demand rather than of supply  _(ep 25)_
 - **B50** — a blending mandate requiring 50 percent biodiesel in the diesel pool, the level Indonesia moved to in 2026  _(ep 9)_
 - **bag (coffee)** — 60 kg, how the coffee trade counts volume  _(ep 1)_
 - **balance sheet** — the one-page supply and demand statement for one crop and one marketing year, built so that supply minus use equals ending stocks and the page closes  _(ep 7)_
@@ -61,6 +63,7 @@ Units, conventions and desk expressions, accumulated as the show introduces them
 - **Coffee C (KC)** — the ICE arabica futures contract, 37,500 lb quoted in US cents per pound with a 0.05 cent tick worth 18.75 dollars  _(ep 12)_
 - **collar (fence)** — buying a put and selling a call against the same position so the price is bounded on both sides, the standard hedging structure around unpriced physical  _(ep 17)_
 - **Commercial (COT category)** — merchants, processors, elevators and end users who hold futures against physical grain, and who are normally short when the speculative crowd is long  _(ep 24)_
+- **commercial buyer** — a privately accountable buyer such as a crusher or feed mill whose bid is derived from the margin it can compute on its own products, and which stops buying when that margin closes  _(ep 25)_
 - **Commitments of Traders (COT)** — the weekly CFTC report that splits open interest in each futures market by category of holder, taken as of Tuesday's close and published the following Friday afternoon  _(ep 24)_
 - **concentration limit** — a cap on how much of a book may sit with one counterparty, one port or one origin, the limit most often left unwritten  _(ep 22)_
 - **confirming bank** — a second bank, usually in the seller's country, that adds its own undertaking to pay, so the seller no longer carries the issuing bank's or the country's risk  _(ep 23)_
@@ -93,7 +96,9 @@ Units, conventions and desk expressions, accumulated as the show introduces them
 - **delivery notice period** — the window in which shorts may tender certified stock against the expiring contract  _(ep 12)_
 - **demand-to-supply ratio** — the Baltic measure of tonne-mile demand growth against fleet growth, above 1.0 when cargo is outrunning ships  _(ep 10)_
 - **demurrage** — the penalty owed when a vessel is held beyond the agreed laytime  _(ep 2)_
+- **derived bid** — a bid computed backwards from the value of the products a buyer will make, as distinct from a bid set by a budget, a target or a mandate  _(ep 25)_
 - **despatch** — the reward paid when loading beats laytime, customarily half the demurrage rate  _(ep 4)_
+- **destination crush margin** — the processing margin available to a crusher in the importing country, computed in local currency from local meal and oil values against landed bean cost, and the leading indicator of whether commercial import buying continues  _(ep 25)_
 - **destination refinery** — a standalone refinery at the consuming end that buys raws on the water and sells whites locally, earning the white premium less its costs rather than a crop margin  _(ep 14)_
 - **differential** — the premium or discount to a named futures month, as in November plus 80, the negotiated part of a physical quote  _(ep 1)_
 - **differential (basis)** — the premium or discount to a named futures month, quoted as plus 80 or minus 20  _(ep 1)_
@@ -169,6 +174,7 @@ Units, conventions and desk expressions, accumulated as the show introduces them
 - **Incoterms** — the standard three-letter trade terms that allocate cost and risk between buyer and seller  _(ep 4)_
 - **independent risk function** — the reporting line that puts marking and position sizing outside the trading book, usually under the chief financial officer, because the owner of a profit cannot also be its scorekeeper  _(ep 22)_
 - **indication** — a guide price that is not firm  _(ep 1)_
+- **indifference price** — the price at which a buyer is economically neutral between two origins or two options, computed by holding landed cost equal and solving backwards for the FOB  _(ep 25)_
 - **inherited risk** — an exposure that arrived attached to a trade rather than being chosen, recognisable because nobody can name the price they were paid to take it  _(ep 22)_
 - **initial margin** — the deposit the clearing house takes per lot when a position is opened  _(ep 3)_
 - **inter-exchange spread** — the price gap between two exchanges pricing related but different goods, such as Kansas City over Chicago  _(ep 5)_
@@ -176,6 +182,7 @@ Units, conventions and desk expressions, accumulated as the show introduces them
 - **joint product** — two outputs produced in fixed proportion from one input, so that neither can be made without the other  _(ep 8)_
 - **Joint War Committee** — the London market body that publishes the list of waters treated as war-risk areas, whose listings oblige a vessel to buy separate war-risk cover at an additional premium  _(ep 23)_
 - **kilolitre** — one thousand litres, the volume unit Asian governments state biofuel mandates in, converted to tonnes using the fuel's density of about 0.88 t per cubic metre for biodiesel  _(ep 9)_
+- **landed cost** — the all-in delivered cost of a cargo at the buyer's port, the CFR value plus duties and port charges, and the only basis on which two origins can honestly be compared  _(ep 25)_
 - **Latency** — the elapsed time between the period a data series describes and the moment it is published, which sets how much of the market's reaction has already happened  _(ep 24)_
 - **laycan** — the window during which a vessel may present for loading  _(ep 1)_
 - **laytime** — the contractually allowed time to load or discharge before demurrage begins  _(ep 4)_
@@ -204,6 +211,7 @@ Units, conventions and desk expressions, accumulated as the show introduces them
 - **meal contract** — CBOT soybean meal, 100 short tons, quoted in dollars per short ton  _(ep 8)_
 - **melt loss** — the sugar lost between raws in and whites out, roughly six percent, which makes a refiner's break-even white premium a function of the raw price rather than a constant  _(ep 14)_
 - **metric tonne** — 2,204.6 lb, the grain trading weight unit outside the US  _(ep 1)_
+- **MFN duty** — the most-favoured-nation rate a country applies to imports from any WTO member with which it has no preferential agreement, which for soybeans entering China is 3 percent  _(ep 25)_
 - **minimum price contract** — a cash sale bundled with a bought call, giving the seller a price floor and retained upside in exchange for a fee  _(ep 19)_
 - **month codes** — F G H J K M N Q U V X Z for January through December, the Z is December  _(ep 1)_
 - **NASS** — USDA's National Agricultural Statistics Service, the body running the surveys behind the published numbers  _(ep 7)_
@@ -217,6 +225,7 @@ Units, conventions and desk expressions, accumulated as the show introduces them
 - **nomination** — formally naming the performing vessel under a cargo contract  _(ep 4)_
 - **non-convergence** — a futures contract that fails to fall to the cash value of its deliverable at expiry because holding the certificate costs less than the carry the market is paying  _(ep 18)_
 - **NOPA** — the National Oilseed Processors Association, whose monthly published crush figure makes US soybean crush a measured line rather than an inferred one  _(ep 8)_
+- **nothing behind it** — broker shorthand that a quote has no supporting interest at or near the level, meaning the print on the screen cannot be repeated in size  _(ep 25)_
 - **notice of readiness (NOR)** — the master's formal declaration that the vessel has arrived and is ready, it starts the laytime clock  _(ep 4)_
 - **offer** — the price a seller will accept  _(ep 1)_
 - **oil contract** — CBOT soybean oil, 60,000 pounds, quoted in cents per pound  _(ep 8)_
@@ -266,6 +275,7 @@ Units, conventions and desk expressions, accumulated as the show introduces them
 - **relative value** — a position expressing a view on the difference between two prices rather than on the direction of either  _(ep 16)_
 - **renewable diesel** — hydrotreated vegetable oil or HVO, a drop-in diesel chemically identical to fossil diesel and not limited by a blend wall, unlike FAME  _(ep 9)_
 - **replacement value** — what it would cost to buy back today what you have just sold, the test of whether a price was genuinely good  _(ep 11)_
+- **reserve rotation** — auctioning older stock out of a state reserve while buying newer stock in, which changes the age of the inventory rather than its size and frees tank space for imports  _(ep 25)_
 - **residual** — a figure obtained by subtraction, such as ending stocks, which absorbs any error in the larger numbers almost in full  _(ep 2)_
 - **retracement** — the partial give-back of a price move once the fear that produced it fails to be confirmed  _(ep 13)_
 - **reverse crush** — the opposite position, short beans and long products, used when a processor expects to idle capacity rather than run it  _(ep 8)_
@@ -292,16 +302,19 @@ Units, conventions and desk expressions, accumulated as the show introduces them
 - **soft red winter (SRW)** — the low-protein soft wheat the Chicago contract delivers, used for cakes biscuits and crackers  _(ep 5)_
 - **soluble solids** — the share of the coffee bean that dissolves in water, higher in robusta, which is why robusta dominates instant coffee  _(ep 12)_
 - **space time form** — the three transformations a merchant is paid for, geography, storage and processing  _(ep 2)_
+- **specific tariff** — a duty charged as a fixed amount per unit of quantity, invariant to price and therefore falling in relative terms during a rally  _(ep 25)_
 - **spread margin credit** — the reduction in initial margin an exchange grants a recognised spread, which lowers the cost of a position without lowering its risk per tonne  _(ep 16)_
 - **square** — desk shorthand for a net position of zero in a given month or location  _(ep 21)_
 - **squeeze** — a front month bid far above the cash value of its deliverable because open interest exceeds what can physically be delivered in the time available  _(ep 18)_
 - **standing bid** — demand that is present regardless of price because it is created by legal obligation rather than by choice  _(ep 9)_
+- **state buyer** — a government-owned purchaser such as a reserve agency whose bid is set by a stockholding or policy target rather than by a processing margin, and which is therefore insensitive to the crush  _(ep 25)_
 - **state reserve auction** — a government selling cotton or grain from its own stockpile into its domestic market, whose clearing rate is read as a signal of domestic tightness  _(ep 15)_
 - **statement of facts** — the port log of events both sides use to fight laytime claims  _(ep 4)_
 - **stocks-to-use** — ending stocks divided by total use, the market's tension gauge  _(ep 2)_
 - **storage tariff** — the published charge for commercial storage, quoted in cents per bushel per month or per day, or in dollars per tonne per month  _(ep 11)_
 - **stress test** — revaluing a book under one specific named scenario drawn from something that actually occurred, rather than from an estimated distribution  _(ep 22)_
 - **string** — a chain of back-to-back sales of the same parcel, each contract bilateral and at its own price, along which the cargo is nominated from the first seller to the final buyer  _(ep 23)_
+- **subject** — a qualifier making a quote or agreement conditional on something not yet done such as approval, allocation or a licence, and therefore not yet firm  _(ep 25)_
 - **substitution spread** — the price gap between two competing vegetable oils, which sets the point at which a refiner reformulates from one to the other  _(ep 9)_
 - **sugar mix** — the share of a mill's recoverable sugars turned into sugar rather than ethanol, bounded above by the plant's crystallisation capacity  _(ep 14)_
 - **Supramax** — a dry bulk vessel of roughly 50,000 to 60,000 dwt, normally carrying its own cranes, working minor bulks and shorter legs  _(ep 10)_

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@sdelsad/commodity-desk-daily",
-  "version": "1.0.72",
-  "description": "Soft Commodity Trading - Ep 24: Data Sources and Positioning",
+  "version": "1.0.74",
+  "description": "Soft Commodity Trading - Ep 25: China, and Reading Market Colour",
   "license": "CC-BY-4.0",
   "keywords": [
     "podcast",

package/ep24.md

@@ -1,217 +0,0 @@
-# Market pulse
-
-**Corn shipped nearly two million tonnes in a week, and still closed lower.**
-
-| Contract | Tuesday 22 Sep settle | Change |
-|---|---|---|
-| Dec corn | 536¾ c/bu | −6¼ c |
-| Nov soybeans | 1325½ c/bu | −2½ c |
-| Dec Chicago wheat | 717¼ c/bu | −9½ c |
-| Dec KC wheat | 781¼ c/bu | −13¼ c |
-| Dec Minneapolis wheat | 736¼ c/bu | −10 c |
-| Dec oats | 414½ c/bu | −1¾ c |
-| Nov canola | 824.30 C$/t | −10.90 |
-
-Monday's export inspections were strong. Total grain loadings in the week to 17 September came to 3.03 Mt, against 2.71 Mt the week before, with corn alone at 1.94 Mt and soybeans at 759,193 t. Wheat did 335,253 t.
-
-Corn fell anyway, and wheat fell hardest, with Kansas City giving up 13¼ cents. Two things sit behind that. The board has been bid for a fortnight on the expectation of Chinese demand news around the Washington meeting set for tomorrow, and an expectation is a position rather than a fact. And the loadings number, strong as it is, measures grain sold weeks ago. It is execution, not appetite.
-
-What to watch is the 30 September Grain Stocks and Small Grains pair, now a week away, and whatever comes out of tomorrow's meeting.
-
-```chart
-{"type":"bar","unit":"c/bu","title":"Tuesday's move across the board",
- "caption":"Wheat led the fall and corn followed it, on a day when the shipping data pointed the other way. Kansas City lost more than twice what corn did.",
- "source":"CBOT, KCBT and MGE December settlements, Tuesday 22 September 2026, GX94 closing commodity prices",
- "x":["Corn","Beans","Chi wheat","KC wheat","MGE wheat"],
- "series":[{"name":"22 Sep","values":[-6.25,-2.5,-9.5,-13.25,-10.0]}]}
-```
-
-## The geopolitical read
-
-The Black Sea story has moved out of the headlines and into the loading data, which is where it becomes a price. Both Russia and Ukraine are running export paces more than half below last year's. The sequence that got them there is now complete: war-risk cover repriced through August, freight absorbed it, and the volumes fell.
-
-That completion is itself the news. A market stops paying for a headline once the shortfall is in the baseline, which is why Chicago wheat can lose nearly ten cents on a day of grim-sounding reports from the region. The premium decays as the disruption becomes the expected state of the world. It would take a change in the level of disruption, in either direction, to move the price again — and a negotiated reopening would move it hard, because nothing in the current price assumes one.
-
-# Key takeaways
-
-- Every data series has a latency and a distance from the thing you actually want to know. The two usually trade off against each other, and the fastest series is normally the one furthest from demand.
-- Export inspections measure execution: grain leaving on a boat that was sold weeks or months ago. Export sales measure appetite. A question about demand is answered by the slower series.
-- Revision is the third axis. The first print of a weekly series is a draft, the market trades the draft, and conviction should be sized to the series rather than to the headline.
-- Positioning is a risk indicator, never a signal. A large net long has never obliged a market to fall, and crowded markets stay crowded for months.
-- What a crowded position does change is the shape of the reaction: violent to bad news, muted to good news, because everyone who wanted to buy already has.
-- The COT is a photograph of Tuesday's close published on Friday afternoon. Everyone gets the same three-day-old picture at the same moment, so there is no edge in the level — only in what you do with the asymmetry.
-- A fund liquidation is a futures event before it is a cash event. The board moves first and the physical lags, which lands on a basis book as a gain or a loss that nobody forecast.
-- The information edge is speed and interpretation, not secrets. The only genuinely private data a merchant holds is its own flow: what it was shown, what it bid on and lost, and who stopped asking.
-
-# Vocabulary
-
-| Term | What it means |
-|---|---|
-| **Commitments of Traders (COT)** | The weekly CFTC report that splits open interest in each futures market by category of holder, taken as of Tuesday's close and published the following Friday afternoon |
-| **Managed money** | The COT category covering hedge funds, commodity trading advisors and other discretionary or systematic speculators, whose net position is the standard measure of how crowded a market is |
-| **Commercial** | The COT category covering merchants, processors, elevators and end users who hold futures against physical grain, and who are therefore normally short when the speculative crowd is long |
-| **Net position** | Long contracts minus short contracts within one COT category, the single number used to describe how that group is placed in a market |
-| **Export inspections** | The weekly USDA count of grain physically inspected and loaded for export in the week ending Sunday, published Monday, measuring execution of business already sold |
-| **Weekly export sales** | The USDA report of new export commitments signed in a week, published the following Thursday with an eight-day lag, measuring fresh demand rather than shipment |
-| **Flash sale** | A same-day or next-day USDA announcement required when a single buyer commits to 100,000 t or more of one commodity, the only genuinely real-time demand data point in the calendar |
-| **Vessel line-up** | The daily published list of ships nominated to load at an export terminal, a statement of intention rather than of loading, and therefore subject to cancellation |
-| **Crop progress** | The weekly USDA survey of planting, condition and harvest by state, collected as enumerator judgment rather than measurement and routinely revised by the following week's print |
-| **Latency** | The elapsed time between the period a data series describes and the moment it is published, which sets how much of the market's reaction has already happened |
-| **Crowded trade** | A position held in size by most of one category of participant, whose risk is not that it is wrong but that its exit has no natural buyer on the other side |
-| **Unwind** | The forced or voluntary exit of a crowded position, typically compressed into very few sessions and producing a price move unconnected to any change in fundamentals |
-| **Open interest** | The total number of futures contracts outstanding and not yet offset or delivered, the denominator against which any one category's position is judged |
-
-# Quiz
-
-**Q1.** You are long 2,400,000 bushels of Gulf corn basis, hedged short 480 lots of December on the board. The COT published on Friday showed managed money net long 426,842 corn contracts as of the preceding Tuesday, at the top of its 52-week range. Over the following three sessions that long liquidates: December corn falls 30 cents, while Gulf cash values fall only 18 cents, because the exporters still have boats to load. Work out the P&L on those three sessions and say which leg produced it.
-
-**Q2.** Export inspections are published on Monday for the week that ended the previous Sunday. Weekly export sales are published on Thursday for a week that ended eight days earlier. A trader wants to know whether Chinese demand for US soybeans picked up last week. Which of the two series answers that question?
-
-**Q3.** A merchant's borrowing base advances 85 percent against hedged inventory and 0 percent against unhedged stock. One morning the desk lifts the futures hedge on a cargo it still owns, meaning to put it back on after a report. What happens to the credit line?
-
-**Q4.** A position sheet shows the book net flat on the total, but long 400,000 bushels in March and short 80 lots in May. Why is "net flat" the wrong description of that book?
-
-**Q5.** Conversion drill. A war-risk premium of $3.50 per tonne is quoted on a Supramax cargo of 55,000 t. What is the cheque?
-
-# SOLUTIONS (spoilers)
-
-**A1.** A gain of $288,000, and neither leg produced it on its own — the basis did.
-
-Take the legs separately. The hedge is short 480 lots of December corn, and a Chicago corn contract is 5,000 bushels, so 480 × 5,000 = 2,400,000 bu of futures against 2,400,000 bu of physical. The board is square, which is the point of the hedge.
-
-The board falls 30 cents. A short position gains when the price falls, so the futures leg makes 30 c/bu × 2,400,000 bu = $720,000.
-
-The physical is long, and Gulf cash fell 18 cents, so the physical leg loses 18 c/bu × 2,400,000 bu = $432,000.
-
-Net: $720,000 − $432,000 = **$288,000**, which is 12 c/bu on 2,400,000 bu. That 12 cents is exactly the change in the basis: cash minus futures went from wherever it was to twelve cents better, because the futures fell twelve cents further than the cash did.
-
-The trap is the phrase "which leg produced it". Neither did. A hedged book has no exposure to the board and no exposure to the cash price — it has exposure to the difference between them, and the difference moved because the two markets have different participants. A fund liquidation is transacted entirely on the screen. The exporter buying corn at the Gulf has a vessel arriving and cannot reprice his need as fast as a fund can reprice its opinion.
-
-That is the whole lesson of positioning on a physical desk. The COT did not tell you corn would fall — nothing tells you that. It told you that a very large paper long sat on one side of a market whose physical side could not move at the same speed. If that long ever left, the paper would move first and the cash would lag, and the gap between them is a basis position. You were already holding one.
-
-**A2.** Weekly export sales, the slower of the two.
-
-Inspections count grain that physically passed an inspector and went onto a boat in the week just ended. That grain was sold weeks or months earlier. The number is therefore a measure of execution: how fast the export programme already on the books is being worked off. A record week of inspections is entirely compatible with nobody having bought anything new for a fortnight.
-
-Export sales count commitments signed during the reporting week — new business, by destination. That is what "demand picked up" means. It arrives eight days late, and it is still the right series.
-
-The trap is treating speed as relevance. Latency and proximity to the question are separate properties of a data series, and on the export calendar they run in opposite directions: the fastest number is the furthest from demand, and the definitive one, customs data, arrives five weeks after the month it describes. The only genuinely fast demand series is the flash sale, which is announced within a day but only exists above 100,000 t to a single buyer.
-
-**A3.** It shrinks immediately, by 85 percent of that cargo's value, and the desk has to fund the cargo out of equity until the hedge goes back on.
-
-The borrowing base is not a fixed line. It is the sum of eligible assets multiplied by their advance rates, recalculated continuously. Hedged inventory carries 85 percent. Unhedged stock carries zero — not a lower rate, zero. Lifting the hedge does not change the grain in the silo at all, but it moves that grain from one bucket to the other, so the amount the bank will lend against it falls from 85 percent of its value to nothing.
-
-The point is the one that makes this worth knowing: the advance rate is a property of the paperwork, not of the commodity. A trader who lifts a hedge "just for the report" has, without noticing, made a treasury decision several times larger than the position decision. On a $15 million cargo that is roughly $12.75 million of bank money withdrawn on the morning he takes a view.
-
-**A4.** Because 80 lots is 400,000 bushels, so the book is not flat — it is long a March/May spread of 400,000 bushels that nobody decided to own.
-
-At 5,000 bushels a lot, the short of 80 lots in May is 400,000 bu, exactly matching the March long. The total nets to zero, which is why the summary line says flat. But the two positions sit in different futures months, and a futures month is a tradeable thing: March corn and May corn have their own prices, and the difference between them moves on storage economics, deliverable supply and who is short into the delivery period.
-
-So the book carries a calendar spread at full size. Nobody put it on as a view. It is the residue of a physical position hedged in one month and a sale hedged in another, and it will make or lose money on its own. That is the reason a position sheet is organised by futures month rather than by shipment month, and the reason the total line is the least informative row on the page.
-
-**A5.** $192,500.
-
-Three dollars fifty a tonne on 55,000 tonnes. The fast method: multiply the per-tonne figure by 5.5 and move the decimal four places. 3.50 × 5.5 = 19.25, so $192,500.
-
-Worth doing in the head, because a war-risk quote that sounds like a rounding error against a $250 per tonne cargo value is nearly two hundred thousand dollars of cheque, and it has to come out of a margin that was quoted before the premium existed.
-
-# The written edition
-
-## Everything arrives late, and the question is how late
-
-A trading desk does not suffer from a shortage of data. It suffers from an inability to rank it. Every morning brings loadings, sales, progress reports, satellite composites, line-ups, customs releases and price feeds, and almost none of them measure the same thing or arrive on the same clock.
-
-Two properties settle how much weight a series deserves.
-
-The first is **latency**: the gap between the period the number describes and the moment it is published. The second is **distance**: how far the thing measured sits from the thing you actually care about. The trap is that these two usually run in opposite directions. The fastest series in a category is often the least relevant one in it.
-
-The export calendar is the clean illustration.
-
-| Series | Latency | What it actually measures |
-|---|---|---|
-| Flash sale | Same or next day | A single commitment of 100,000 t or more |
-| Export inspections | 1 day | Grain loaded onto a boat, sold weeks or months ago |
-| Crop progress | 1 day | Enumerator judgment of condition and harvest |
-| COT | 3 days at release | Who held the open interest on Tuesday's close |
-| Weekly export sales | 8 days | New commitments signed, by destination |
-| Customs data | ~5 weeks | What actually crossed a border |
-
-Ask the series a question and the ranking inverts. If the question is "did Chinese demand improve last week", export inspections are the freshest number on the table and the wrong one: they count boats loading against business done in July. Export sales arrive eight days late and answer the question exactly. Customs data answers it beyond argument, five weeks after anybody can trade it.
-
-```chart
-{"type":"bar","unit":"days of latency","title":"How late each number arrives",
- "caption":"The two fastest series on this list say least about demand, and the definitive one arrives five weeks after it could be traded. Speed and relevance run in opposite directions.",
- "source":"USDA and CFTC publication calendars; worked example, episode 24",
- "x":["Inspections","Crop progress","COT","Export sales","Customs"],
- "series":[{"name":"Latency","values":[1,1,3,8,35]}]}
-```
-
-There is a third axis, less discussed and equally expensive: **revision**. Weekly export sales are revised. Crop progress is effectively rewritten by the following week's print. The first number is a draft, and the market trades the draft. A series can be fast, close to the question and still wrong on the morning it appears, which is an argument for sizing conviction to the series rather than to the headline it generated.
-
-The same discipline applies to the fashionable sources. Satellite imagery is continuous and cheap, and it measures greenness rather than yield, which is a serviceable proxy in a normal year and a poor one in the years that matter. A crop tour samples enormously and happens once, in August, producing a point estimate with no time series behind it. Vessel line-ups at Santos update daily and list nominations: a boat on the line-up is an intention, and intentions are cancelled. None of these are bad data. They are bad when used as something they are not.
-
-## Positioning: who is holding it, and what happens if they stop
-
-Every Friday at 3:30pm New York time the CFTC publishes the **Commitments of Traders**, which takes the open interest in each futures market and sorts it by category of holder. Two categories do the work.
-
-**Managed money** is the speculative crowd: hedge funds, CTAs, systematic programmes. **Commercials** are the merchants, processors and elevators who own physical grain and hold futures against it. The two are usually mirror images, and for a structural reason — the speculator is taking the price risk the merchant is trying to get rid of.
-
-The unit is contracts, and the conversion is where the number becomes real. A Chicago corn contract is 5,000 bushels. As of Tuesday 15 September, managed money was net long **426,842 corn contracts**. Multiply: 426,842 × 5,000 = 2,134,210,000 bu, or about 54.2 million tonnes of corn, held by participants who will never take delivery of a kernel. That is on the order of an eighth of a US corn crop existing purely as opinion.
-
-The same report showed soybeans net long 244,710 contracts, down 21,321 on the week, and Chicago wheat flipping back to net short at −4,706, a swing of nearly 9,000 contracts in seven days.
-
-```chart
-{"type":"bar","unit":"contracts, net","title":"Managed money, 15 September",
- "caption":"Corn sits at the top of its 52-week range while wheat has flipped back to net short. The two markets will react to identical news in completely different ways.",
- "source":"CFTC Commitments of Traders as of Tuesday 15 September 2026, released 18 September",
- "x":["Corn","Soybeans","Chi wheat"],
- "series":[{"name":"Net position","values":[426842,244710,-4706]}]}
-```
-
-### The convention that catches everyone
-
-The report is taken **as of Tuesday's close** and published **Friday afternoon**. On the day it appears it is already three days old. By the following Wednesday it is eight days old, and a fortnight of price action may have happened inside the gap.
-
-This is not a defect to be worked around. It is the nature of the product. Every desk receives the same three-day-old photograph at the same second, which means there is no edge whatsoever in knowing the level. The edge, such as it is, lies entirely in what you conclude from it.
-
-### What positioning is, and what it is not
-
-Positioning is **not a signal**. A large net long has never once obliged a market to fall. Crowded markets stay crowded for months, and the graveyard of traders who shorted corn because the funds were long is well populated.
-
-Positioning is a **risk indicator**. It tells you the shape of the reaction function, not the direction of the news. A heavily long market responds violently to bad news and barely at all to good news, because everyone inclined to buy has already bought and there is no marginal buyer left to impress. That asymmetry is a tradeable fact. The direction is not.
-
-Reading only the managed money line gets this half right. Every long has a short facing it, and in grains that short is usually a commercial holding physical. When commercials are heavily short and funds heavily long, the market has sorted itself into people who own grain and people who own an opinion — and under stress those two groups do entirely different things. The commercial delivers. The fund sells.
-
-### A crowded long, unwinding
-
-Soybean meal made a new two-year high in every session of the week beginning 14 September. On Friday 18 September it fell 3.8 percent in one session, while corn lost about half a percent. Nothing in the meal balance sheet changed that day.
-
-A CBOT meal contract is 100 short tons, so $14.10 off the price is $1,410 a lot. A fund carrying 2,000 lots lost $2.82 million in an afternoon on a market with no news in it. The buying that made the high was the same buying that had to leave, and it left at once. That is what a crowded position is: not a wrong view, but a position whose exit has no natural buyer on the other side of it.
-
-### Where that lands on a physical book
-
-The consequence for a merchant is the part that is rarely spelled out. Take a desk long 2,400,000 bu of Gulf corn basis, hedged short 480 lots of December.
-
-| Leg | Move | P&L |
-|---|---|---|
-| Short 480 lots Dec futures | −30 c/bu | +$720,000 |
-| Long 2,400,000 bu Gulf physical | −18 c/bu | −$432,000 |
-| **Net** | **basis +12 c/bu** | **+$288,000** |
-
-The board is square, so the flat price contributes nothing. What produced the $288,000 is the twelve cents by which the futures fell further than the cash did — and they fell further because a fund liquidation is transacted entirely on a screen, while the exporter at the Gulf has a vessel arriving and cannot reprice his need at the same speed.
-
-The COT never said corn would fall. It said a very large paper long sat on one side of a market whose physical side moves slowly. That is a statement about the gap between the two, and the gap is a basis position — one this desk was already carrying before it read the report.
-
-## The edge is not a secret
-
-None of this information is private. Inspections, sales, progress, the COT: every desk on earth receives them in the same second, and has done for decades. The idea that a trading house wins because it knows something nobody else knows is mostly a story told by people who have not worked on a desk.
-
-The edge is two things, and only one of them is interesting. The first is **speed**, which is now measured in milliseconds and is bought rather than earned. The second is **interpretation** — knowing which series answers which question, what each one is a proxy for, how badly it revises, and what the market has already priced of it.
-
-There is one genuinely private dataset, and every merchant has it: **its own flow**. What it was shown this morning. What it bid on and did not win, and by how much. Which buyer stopped asking. A tender lost by a dollar a tonne says more about where the market truly is than any published series, and it exists in nobody's database but yours. Firms that are good at this treat their own flow as data — logged, aggregated, reviewed — rather than as gossip that evaporates at the end of the day.
-
-## The thing to carry away
-
-Rank every number by two questions: how late is it, and how far is it from what I want to know. Then accept that the answer to the second question is usually "further than it looks".
-
-And treat positioning as a description of the room, not a forecast of the weather. It tells you who is standing where, and therefore how hard the floor will shake when somebody moves. It has never told anybody which way.

package/ep24.script.txt

@@ -1,84 +0,0 @@
-Fifty-four million tonnes of corn. That is what the speculative crowd was long in Chicago, on paper, last Tuesday. ||| 0.5
-Nobody will ever deliver it. Nobody wants it. And it is one of the most useful numbers on a grain desk this morning. ||| 0.7
-This is Soft Commodity Trading, episode twenty four. Today: where a desk gets its information, and what positioning actually tells you. ||| 0.8
-First, the tape. Tuesday, the twenty second of September. ||| 0.4
-December corn settled five thirty six and three quarters, down six and a quarter. ||| 0.3
-November beans, thirteen twenty five and a half, down two and a half. ||| 0.3
-December Chicago wheat, seven seventeen and a quarter, down nine and a half. Kansas City down thirteen and a quarter. Minneapolis down ten. ||| 0.5
-Wheat led the fall, on a day when the headlines sounded bullish. ||| 0.6
-Export inspections came out Monday. Three point zero three million tonnes shipped in the week to the seventeenth, up from two point seven one the week before. Corn did one point nine four million of that. ||| 0.5
-Corn shipping nearly two million tonnes in a week is a strong number. Corn fell anyway. ||| 0.7
-The Black Sea story has moved from the headlines to the loading data. Both Russia and Ukraine are running export paces more than half below last year. ||| 0.5
-That is the part with a price attached. Insurance repriced in August, freight followed, and now the volumes simply are not there. ||| 0.5
-The market has stopped paying for the headline, because the shortfall is already in the price. A war premium decays when it becomes the baseline. ||| 0.6
-And Washington and Beijing meet tomorrow. Corn and beans have been bid on that meeting, which is a position, not a fact. ||| 0.7
-Which is exactly where today's subject starts. ||| 0.6
-A desk drowns in data. The skill is not collecting it. The skill is knowing what each number actually measures, and how old it is. ||| 0.6
-Take three series that all look like demand. ||| 0.4
-Export inspections. Published Monday, for the week that ended Sunday. One day old. It counts grain that physically went past an inspector onto a boat. ||| 0.5
-Weekly export sales. Published Thursday, for a week that ended the Thursday before. Eight days old. It counts contracts signed. ||| 0.5
-Customs data. Published five weeks after the month ends. Definitive, and useless for trading. ||| 0.6
-Now the trap. Inspections is the fastest of the three. It is also the one that tells you least about demand. ||| 0.6
-A boat loading this week was sold in July. Inspections measure execution. Sales measure appetite. ||| 0.6
-So if you want to know whether China came back last week, the eight-day-old number is the right one and the one-day-old number is the wrong one. ||| 0.7
-Speed is not relevance. That sentence is most of the job. ||| 0.8
-Every series sits on two axes. How late it is. And how far it is from the thing you actually care about. ||| 0.6
-Crop progress is one day old, and it is a judgment, collected from people walking fields. ||| 0.4
-Satellite imagery is continuous, and it measures greenness, not yield. ||| 0.4
-A crop tour has an enormous sample and happens once, in August. ||| 0.5
-Vessel line-ups at Santos update every day. They are nominations. A boat on the line-up is an intention, and intentions get cancelled. ||| 0.6
-None of those are bad numbers. They are only bad when you use them as something they are not. ||| 0.7
-And there is a third axis nobody mentions: revision. ||| 0.4
-Export sales get revised. Crop progress gets rewritten by the following week's print. The first number is a draft, and the market trades the draft. ||| 0.6
-So a series can be fast, close to what you care about, and still wrong on the morning it prints. Size your conviction to the series, not to the headline. ||| 0.8
-Now the other half of the job. Positioning. ||| 0.5
-Every Friday afternoon the C F T C publishes the Commitments of Traders. It takes the open interest and sorts it by who is holding it. ||| 0.6
-Two categories carry the weight. Managed money — funds, C T As, the speculative crowd. And commercials — merchants, processors and elevators, the people who own actual grain. ||| 0.7
-The unit is contracts, and this is where people get lost. A Chicago corn contract is five thousand bushels. ||| 0.4
-Last Tuesday, managed money was net long four hundred and twenty six thousand, eight hundred and forty two corn contracts. ||| 0.4
-Multiply it out. Two point one billion bushels. Fifty four million tonnes. Roughly one bushel in eight of the entire American crop, held by people who will never see a kernel of it. ||| 0.8
-Two more numbers from the same report. Beans, net long two hundred and forty four thousand, down twenty one thousand on the week. ||| 0.4
-And Chicago wheat flipped back to net short, minus four thousand seven hundred, a swing of nearly nine thousand contracts in seven days. ||| 0.6
-Then there is the convention that catches everyone. The report is as of Tuesday's close. It is published Friday at half past three, New York time. ||| 0.6
-So the freshest positioning figure available to you, on the day it comes out, is already three days old. By Wednesday it is eight days old. ||| 0.6
-That gap is not a flaw in the report. It is the report. Everyone is looking at the same photograph, at the same moment, of the same three-day-old crowd. ||| 0.8
-So when a broker rings about positioning, nobody asks what the funds are doing — they ask what happens if the funds stop. ||| 0.5
-TRADER: Who is left to buy corn? ||| 0.25
-BROKER: Nobody I can see. Managed money is at the top of its range. ||| 0.25
-TRADER: And the commercials? ||| 0.25
-BROKER: Short. They have been selling the rally into the funds all month. ||| 0.25
-TRADER: So it is a one-way book. ||| 0.25
-BROKER: It is a one-way book. ||| 0.7
-Neither of them said whether corn was going up or down. ||| 0.5
-They said who is on each side, and what would have to happen next. That is what positioning is for. ||| 0.8
-Read both sides of that report, always. Managed money is only half of it. Every long has a short facing it, and in grains that short is usually a commercial who owns the physical and has hedged it. ||| 0.6
-When commercials are heavily short and funds heavily long, the market has sorted itself into people who own grain and people who own an opinion. ||| 0.6
-Under stress those two groups behave completely differently. The commercial delivers. The fund sells. ||| 0.8
-So positioning is not a signal. A large net long has never once meant a market must fall. Crowded markets stay crowded for months. ||| 0.6
-Positioning is a risk indicator. It tells you the shape of the reaction, not the direction of the news. ||| 0.6
-A heavily long market reacts violently to bad news and barely at all to good news. Everybody who wanted to buy already has. ||| 0.6
-That asymmetry is tradeable. The direction is not. ||| 0.8
-Watch what it looks like when it breaks. Soybean meal set a new two-year high in every session of the week before last. ||| 0.5
-Then, on Friday the eighteenth, meal fell three point eight percent in a single session. Corn that day lost half a percent. ||| 0.6
-Nothing in the meal balance sheet changed on that Friday. The buying that made the high was the same buying that had to leave, and it left all at once. ||| 0.7
-A meal contract is one hundred short tons. Fourteen dollars ten off the price is fourteen hundred and ten dollars a lot. ||| 0.4
-A fund carrying two thousand lots lost two point eight million dollars in an afternoon, on a market with no news in it. ||| 0.8
-Now put yourself on the physical side of that. ||| 0.5
-Say you are long two point four million bushels of Gulf corn basis, hedged short four hundred and eighty lots on the board. ||| 0.5
-The funds liquidate. The board drops thirty cents over three sessions. Gulf cash only drops eighteen, because the exporter still has boats to load. ||| 0.6
-Your short futures make thirty cents. Your physical loses eighteen. You are up twelve cents on two point four million bushels. Two hundred and eighty eight thousand dollars. ||| 0.7
-You made money on a liquidation you never predicted. A fund exit is a futures event, and the cash market cannot follow it that fast. ||| 0.7
-That is what the C O T was telling you. Not that corn would fall. That if it fell, the paper would move first and the physical would lag. ||| 0.6
-Which is a basis position. And you already had one. ||| 0.8
-One last thing, and it is the uncomfortable one. ||| 0.5
-None of this is secret. The inspections number, the sales number, the C O T — every desk on earth gets them in the same second. ||| 0.6
-The edge is two things. Speed, which is now measured in milliseconds and is mostly bought rather than earned. And interpretation, which is not. ||| 0.7
-The genuinely private information a merchant holds is its own flow. What it was shown. What it bid on and did not win. Which buyer went quiet last week. ||| 0.7
-A tender you lost by a dollar tells you more about where the market really is than any published series, and it sits in nobody's database but yours. ||| 0.8
-So. Three things to carry out of this. ||| 0.5
-First: every number has a latency, and a distance from the thing you actually care about. The fast one is usually the far one. ||| 0.6
-Second: positioning sizes the reaction, not the direction. Crowded is a statement about what happens next, never about what is true. ||| 0.6
-Third: the report you read on Friday is a photograph of Tuesday. Trade the asymmetry, never the level. ||| 0.8
-Next time: China. One buyer large enough to move a world balance sheet, and how to read what a broker is actually telling you. ||| 0.6
-Four questions in the notes, as always. The first is a full worked problem on exactly the position we just built. ||| 0.4
-The solutions are there too. Do the work first. ||| 0.6