Thirty-Nine Seasons of Duck Days

Where you hunt: a factor-of-two spread

Refuge choice is real. Pooling all 39 seasons, Kern in the southern San Joaquin leads the serious-sample public areas at about 2.9 birds per hunter, with Gadwall and Little Dry Creek just behind at roughly 2.8, and Delevan and Colusa — Sacramento Valley stalwarts — close after. At the other end, a long tail of units averages closer to one bird per hunter. Top to bottom, the well-sampled refuges span roughly a factor of two to three in raw success.

The regions sort cleanly too. The Sacramento Valley tops the state at about 2.2 birds per hunter, the San Joaquin sits at 2.1, Southern California at 1.8, the Northeast at 1.6, and Suisun Marsh & Delta averages lowest at 1.4. But keep the caveat in mind: a refuge's raw average bundles together everything about it — its water, its pressure, its season timing, even which kinds of days its regulars choose to hunt. The raw leaderboard tells you where success has happened, not how much of it the marsh itself deserves credit for.

When you go: the season has one shape, everywhere

The arc through the season is the most dramatic pattern in the raw record, and it repeats almost everywhere. October opens hot: the first half of the month — early-zone openers, mostly in the northeast — runs about 3.0 birds per hunter, and the last half of October, when the main Central Valley openers land, holds 2.3. Then the bottom falls out. Early November averages 1.3 — less than half the opening pace — aka Slowvember. The easy local birds have been shot at and the big northern push hasn't arrived. From there success climbs steadily every half-month: 1.8 in early December, 2.1 in late December, 2.3 in early January, and 2.6 in the last two weeks of the season as fresh migrants pile in.

Chart 03 shows that arc as birds-per-hunter by half-month. Focus a region in the controls and a dashed line appears — the all-refuge average under the same conditions — so you can see whether your marsh beats the statewide pattern or just rides it. One honest note on the early-October bar: only about a thousand hunt-days land there, mostly in the northeast zone, so it's a real spike but a thin slice of the record.

The weather, as it looks raw — and why it looks wrong

Here is where the raw record gets genuinely interesting. Chart 04 breaks success down by eight weather patterns — sky, wind direction, daytime high, wind speed, cold-front swing, barometer trend, moon, and rain. These are plain marginal averages: every hunt-day sorted into a band, no adjustments.

Read them at face value and the results are surprising. Clear days (2.21) beat rainy ones (1.93). Dry days beat wet days. A cold-front day (2.00) is indistinguishable from a steady one (1.99), and mild warm-up days actually edge both. Cold days don't beat warm days by much of anything. Only two of the eight splits stand out cleanly: wind — genuinely windy days (13+ mph) stand clear of the pack at 2.28 — and the moon, where bright-moon periods drag noticeably below dark ones.

The catch is that these marginals mix everything together. Rainy days cluster in the heart of the season, cold-front days are also colder and windier, and the biggest single day of the year — the opener — falls on a warm, clear October weekend, quietly crediting "warm and clear" with a pile of birds that belong to the calendar. Untangling what each factor does on its own takes a statistical model, and that's the subject of the companion piece on the white paper — where several of these raw readings reverse direction.

What the raw record can — and can't — tell you

Taken on its own terms, the record supports a few plain statements:

• The calendar dominates. The swing from the opener (~2.3–3.0) to the early-November trough (1.3) and back to late January (2.6) dwarfs every weather split in the raw data. Nothing else in these charts moves birds-per-hunter by a factor of two.
• Refuge choice is a real, persistent edge. The same units sit at the top of the leaderboard decade after decade, and the well-sampled spread is roughly two- to three-fold.
• Wind is the one weather split that survives face-value reading. Genuinely windy days (13+ mph) out-produce everything else; a bright moon under-produces.
• Most other weather splits are flat — or backwards — in the marginals. Rain, cold fronts, and falling barometers all look indifferent or worse in the raw splits.

And here is what it can't tell you. The raw splits can't say whether clear days help or whether the opener just happens on them; whether Kern is a great marsh or a well-watered one in a desert; whether rainy days are bad or just mid-November in disguise. Every hunt-day carries its whole context with it — date, place, weather, and the self-selection of the hunters who chose that morning — and a marginal average smears all of that together.

There's also a whole layer of variables nobody records at all. Did the hunters on a given day simply shoot poorly? Did the better hunters draw the less desirable, less productive spots on the refuge that morning? Was a unit half-flooded, or freshly disked, or crowded at the parking lot? Over 39 seasons that kind of noise should mostly balance out — which is exactly why this record is worth studying at this scale — but on any single day it can swamp everything the weather is doing.

Sources & reproducibility