Research
Background the monthly reports stand on.
| October Deep Dive — October 2, 2026 The full analysis behind the October report: every market print by state, the eight-variant nowcast ($271–308, median $277), weather and water, the grid, a verification table, and the review findings with their fixes. |
| How Sure Are We? — October 2, 2026 Which claims in the October report are solid, which are judgment, three separate checks on the ±$65 range, and what we did where confidence wasn't good enough. |
| Pre-Publication Review — October 2, 2026 An adversarial read of the first draft against the saved data, before publication. Published unedited, except private business details removed. |
| Grading Record — Rows #17–19 and #28 USDA August prices graded by script (Colorado alfalfa $235, our point exactly; Kansas $127, point off $28); the Sep 24 Colorado report graded by hand (a 25-ton Southeast farm sale printed when we said none would). Judgment beat the persistence available at freeze on Colorado and lost on Kansas. |
| Pre-Registered Predictions — November 2026 Releases Six predictions frozen Oct 2 for the Oct 22 – Nov 19 Colorado reports, the Nov 5 drought map and the Nov 30 USDA October prices. Two flagged low-information; a seventh withdrawn before release as padding. |
| SE Colorado Hay Price Primer What a ton of premium alfalfa costs, bale-weight math, and how quality grades map to price. |
| Irrigation Water in the Lower Arkansas Valley How the Fort Lyon Canal, Fry-Ark Project, and John Martin Reservoir actually deliver (or don't deliver) water. |
| Snowpack & Runoff, Water Year 2026 Why 2026 is the worst water year on record for the lower Ark, and what 2002 teaches about hay prices in a drought. |
| Water Rights Primer Prior appropriation, mutual ditch companies, and the compact call — the plumbing behind every price in these reports. |
| Pre-Registered Predictions — October 2026 Releases Twelve predictions frozen Sep 19 for the Sep 24 / Oct 8 / Oct 22 Colorado reports, the Oct 8 ENSO discussion, the Oct 9 hay re-forecast, and the Oct 30 USDA price print. Local rows are now categorical with the base rate stated (the target prints ~2.5% of the time); state-price rows use a separate, tighter family; weekly volume is dropped. |
| Correction After Independent Review — September 19, 2026 Same-day correction to the September report: a $170 Oklahoma lot left out of parity, a netback computed from an origin price, a mixed-product 'peak' claim, and holes in the calibration enforcement. What was wrong, what was fixed, the recomputed valuation. |
| Independent Review of the September 19 Report A second model's review of the September 19 report against the saved primary data. Verdict: request changes. Published unedited, except private business details removed. |
| Grading Record — September 2026 Releases Rows #20–27 graded by hand against the Sep 10 Colorado report and the Sep 10 CPC discussion, every outcome tied to a quoted line; scorecard from the ledger. |
| Pre-Registered Predictions — September 2026 Releases Eleven predictions frozen Aug 27 for the Sep 10 AMS/ENSO releases and the ~Sep 30 USDA price print — every range at our full calibrated width (±$65 on prices), and for the first time each numeric row records its mechanical anchor and the judgment adjustment separately, so the judgment layer itself can be graded. |
| Pre-Registered Predictions — Aug 12 USDA Print Written down BEFORE the Aug 12, 2026 USDA Crop Production report, so the grading can't be fudged. |
| The Lessons Ledger Every rule this project has paid for with a graded miss, newest first — binding on future reports. Bought most recently with the August 2026 production miss. |
| Grading Record — August 2026 Releases Every pre-registered row graded against the actual Aug 12–13 releases, source line cited. Interval coverage 2/10 (corrected same day from 3/10 — one row was graded against the wrong metric); the economically central cluster went 0-for-3. |
| Independent Forecast Audit — August 2026 An outside audit of this project: the facts check out, the precision doesn't. Where we overclaim, the auditor's own confidence table, and the fixes we're adopting. |