How the numbers are made

Everything on this site comes from auction results we've recorded and checked by hand. This page explains where the data comes from, what each number means, and what it can't tell you.

Where the data comes from

Every lot on Decoy Database is a bird that sold at a public auction. We record it from the auction house's published results: the catalog description, the estimate, the price it brought, and the photograph. Today that means American Sporting Auctions, GC Decoys, Guyette & Deeter and Zebs Auctions.

We don't record every lot in every sale. We record the makers we track, and for those makers we try to record every lot. A lot enters the site only after one of us has looked at the record against the house's page and marked it reviewed. That's what "every sale reviewed" means on the home page — a person, not a script, has checked it.

If you find a lot that's wrong — a misattributed bird, a bad price, a date that doesn't match the catalog — use the "Report an error" link at the bottom of any page. We fix it and note it under Corrections.

What a price means

The price shown for a lot is what the buyer paid, including the auction house's premium. Guyette & Deeter's premium is 20%, so a bird that hammered at $1,000 shows as $1,200. We do it this way because it's the number that actually changed hands, and because it's what the houses publish.

Where we know the hammer price separately, the lot page shows it as well. Estimates are the house's, as printed in the catalog. A lot is marked "over estimate" when the price including premium is above the high estimate, and "under estimate" when it's below the low.

How the typical price is worked out

A maker's line answers one question: are this maker's birds bringing more or less than they were? It doesn't tell you what any particular bird is worth — the lot pages, the comparable sales and the price lookup do that.

The rule

1. Take every sold lot for the maker in the 12 months ending on a given month.
2. Find the median price — the middle sale, not the average. One $570,000 wood duck shouldn't move the number for 176 other birds.
3. Do that for every month and draw the line. A month whose window holds fewer than 5 sales gets no point, so the line breaks rather than bridging a quiet stretch.

The range bars on the prices page come from the same window: the bar spans the 10th to the 90th percentile of the actual sales, and the tick is the median. Nothing is fitted to a curve — the one in ten birds above and below each bar are the exceptional ones and the rough ones, and they are left where they fell.

Why the median

Decoy markets are thin and prices are spread across four orders of magnitude for the same maker. An average would swing on every headline bird. The median is stubborn: it moves when the middle of the market moves.

What the line doesn't remove

It follows whatever sold. If a maker's miniatures dominate one year and full-size birds the next, the median moves for reasons that have little to do with demand. We show the individual sales as dots behind every line so you can see when that's happening, and every maker page can be switched to full-size birds only.

Nightly, we recompute everything from scratch. When a lot is added or corrected, history changes too. That's deliberate: the line should always reflect the best data we have, not the data we had when we first drew it.

When we don't publish a number

A number that looks precise but rests on three sales is worse than no number. So:

• A point on a maker's line is published only when its 12-month window holds at least 5 sales. Otherwise the line has a gap.
• A typical sale — the twelve-month median on a maker card, in the makers table or on a price bar — is published only above 10 sales in that window. Below it the card shows a dash and the maker gets no bar at all. Today 62 of 318 makers clear it.
• A percentage change is shown only when both ends of the comparison have at least 15 sales. Below that, the table shows a dash. Every percentage on the site is a change against the same twelve months a year earlier, with its sale count beside it.
• Makers with too few sales for a line still appear on the site, with their lots and their sales table, just no chart.

Wherever a number appears, the count of sales behind it is next to it or one tap away.

Data limitations

Every price index has a version of this section. The ones that don't print it tend to get found out, so here is ours: what we cover, how well we know when things sold, and what is missing.

Coverage

We hold 9,140 sold lots across 318 makers, from American Sporting Auctions, GC Decoys, Guyette & Deeter and Zebs Auctions. Records reach back to 2013, but coverage only becomes consistent from 2023 — before that we hold a scattering of notable sales, not a market.

Coverage is by maker, not by sale. We record the makers we track, and for those makers we try to record every lot that comes up. So a count of lots in a given sale is a count of the lots we track, never the whole catalog. A maker who isn't on the makers page has no data here — that is not the same as no sales.

Precision eras

Our older records know only the year a bird sold; most know the month; the newest know the day. We never display a date more precisely than we know it, and a maker's chart colours its dots by which of these it has.

Dates for online auctions are recorded from the sale listing and may be off by a few days.

The ten founding maker tables carry year-only dates, so those makers' charts are almost entirely year-precision: a Chesser chart shows what a year brought, not what a month did. Where more than a fifth of the sales behind a line are year-only, the chart says so. A further 35 lots have no sale date at all; they appear on the maker's list marked "date unknown" and are counted in nothing.

What isn't counted

• Lots that didn't sell. Passed and withdrawn lots are not in the data. These numbers reflect what changed hands, and they read a little stronger than a market that includes the birds nobody bought.
• Private sales, shows, dealers, eBay. Only public auction results. For some makers that's most of the market; for others it's a slice.
• Other auction houses. Decoys Unlimited, Frank & Frank, Grant Cole and regional sales aren't in yet. We'll say so here when they are.
• Condition, rarity and provenance. A typical sale is the middle of everything that sold, mint birds and rough ones together. Two sales at the same price can be very different birds.
• Makers we don't track. If a maker isn't on the makers page, we have no data on them — not zero sales, no data.

Recaps

An auction recap covers the lots we track from that sale, not the whole catalog. "Brought" is the sum of those lots including premium. "Over high estimate" is the share of tracked lots with an estimate that beat the high number. The written part of a recap is a person's opinion and is signed.

Corrections

When a number on the site changes for a reason other than new sales, it's logged here.

• 6 September 2026 — Consolidated 42 per-maker tables into one dataset. Dates standardized with explicit precision. No prices changed.
• 6 September 2026 — Merged two Grayson Chesser tables that had been kept separately; 111 lots, no duplicates.
• Pending — Five lots dated December 2022 are under review — likely a typo'd year.

Questions about the method: decoydatabase@gmail.com.