Drill: No decoration: a landing page using ONLY typography, spacing, and alignment. No gradients, no illustrations, no decorative SVG, no glassmorphism. Target: looks designed, not empty.
Interpretation: Agricultural extension variety-trial bulletin (Bulletin B-447) — the document family historically typeset with zero decoration by necessity. 24 winter wheat entries × 4 harvested sites with LSD/CV separation and letter groups, worked comparisons recomputed on load, an inverted disclosures band for a hail-lost site and a dropped replication. Newsreader/Inter/JetBrains Mono on warm paper, one wayfinding green enforced as a runtime accent budget, and the no-decoration contract asserted by on-load audits.
Twenty-four entries, five sites, two public checks — and every number on this page derived from the tables below, including the ones that disappoint us.
We harvested four of five sites, tested 24 entries against two public checks, and paid for a fifth site the June hail took back. 7 entries are statistically tied at the top — the distance between them is smaller than the error of the experiment.
The tied count, the trial mean, and every worked comparison on this page are computed from Table 1 on load. Nothing that could drift is typed by hand.
Entries ordered by four-site trial mean. Yields are bushels per acre at 13.5% moisture; TWT is test weight in lb/bu; Prot is grain protein at 12% moisture. Site columns are means of four replicates (three at Lodi — see §04).
Table 1 · 2026 harvested sites · scroll horizontally on small screens
| Variety | Hat Cr. | Lodi | Mill Bd. | Cedar Fl. | Mean | TWT | Prot | Grp |
|---|---|---|---|---|---|---|---|---|
| LSD (0.05) | 4.8 | 5.2 | 4.4 | 5.6 | 4.2 | 0.8 | 0.5 | — |
| CV % | 7.1 | 8.9 | 6.2 | 8.3 | 5.4 | 1.1 | 2.7 | — |
† Public check varieties — entered every year since 2011 by the program, paid for by no seed company. Grp: entries sharing a letter do not differ from their group’s leading entry by more than the trial LSD (0.05) = 4.2 bu/ac; grouping rule stated in §05.
An LSD is the smallest difference that is probably not luck. Smaller gaps — even between rank 1 and rank 7 — are ties in every sense the experiment can defend. This year the trial LSD is 4.2 bu/ac, and the top of the table is crowded: the #1 entry leads the #7 entry by less than that.
So a catalog can say “#1 in the 2026 Plains Ridge trial.” It is quoting a real table. It cannot honestly say “beat” anyone ranked 2 through 7 — the data does not carry that sentence. The comparisons at right are recomputed from Table 1 on load; the ad-copy rule follows from them.
Protein runs the other way. The highest-protein entry this year — the yield-for-protein trade this region expects, visible in a single table.
Practical read: any group-a entry grew as well as any other this year. The table cannot separate ranks 1–7, so seed cost, standability, and where you buy it can break the tie honestly — this table cannot.
Four sites carried to harvest. Rainfall is September–July precipitation at the trial edge.
Table 2 · Locations, dates, and the conditions behind the numbers
| Site | County | Soil series | Planted | Harvested | Reps | Rain in. | Previous crop | Tillage |
|---|---|---|---|---|---|---|---|---|
| Hat Creek | Banner | Duroc loam | 2025-09-28 | 2026-07-08 | 4 | 14.8 | Proso millet | No-till |
| Lodi | Garfield | Holdrege silt loam | 2025-10-06 | 2026-07-12 | 3 | 16.2 | Field pea | Minimum |
| Mill Bend | Sherman | Keith silt loam | 2025-10-02 | 2026-07-01 | 4 | 11.9 | Fallow | No-till |
| Cedar Flats | Logan | Kuma silty clay loam | 2025-10-11 | 2026-07-15 | 4 | 17.4 | Sunflower | Conventional |
| Bromley | Keith | Rosewood silt loam | 2025-10-04 | — | — | — | Wheat (failed) | No-till |
The Bromley row is retained deliberately: a results table that quietly drops the missing site invites you to forget it existed. See §04.
Hail on 14 June 2026 shredded the canopy three weeks before heading. Stand uniformity was gone; any harvest number would have measured the storm, not the varieties. 96 plots — 24 entries × 4 reps — were planted, sprayed, and never harvested. The loss is weather, but the promise of five sites was ours; the year is one site shorter than we said it would be.
The fourth replication sat in the field’s ponded south corner — a drainage problem we knew about and planted anyway. It was dropped before analysis, not after a bad number appeared, and Lodi is reported as a 3-rep site throughout. Table 1 and Table 2 both say 3. This paragraph is why.
Randomized complete block, four replications (three at Lodi); plots 5 ft × 24 ft, seven rows at 7.5 in.; seeded at 1.35 M seeds/ac; 46-0-0 pre-plant. Harvested with a small-plot combine; moisture and test weight measured per plot.
ANOVA by site and across sites (mixed model, site as a random effect). Mean separation by Fisher’s protected LSD at α = 0.05. Letter groups in Table 1 are simplified: an entry receives the letter of the highest-mean ungrouped entry it does not differ from by more than the LSD. The simplification can only make groups look more separated than Tukey grouping would — never less.
No entry paid to be in this trial. Entries are set by a grower ballot each July; sponsors may not buy placement. Two public checks — Rampart and Cedar 480 — have run since 2011 so this year’s numbers can be read against a decade of the same yardsticks.