Short answer: a usable auction record points back to a primary source, states its basis, and matches the object. Without those three properties, a figure is anecdote, not evidence.
Six classes of public source are ranked below, followed by five frequent misreadings and a four-step method for turning comparable sales into a defensible estimate.
Key points (text version)
- 01 House archive
- 02 Printed catalogue
- 03 Price database
- 04 Association report
- 05 Press
- 06 Scholarly record
1. Six sources
| Source | What it proves | Limitation |
|---|---|---|
| Official house archive / catalogue | Lot, description, estimate, result, image | Older sales may be offline |
| Printed catalogue and price list | Full original description | Hard to obtain |
| Price databases | Cross-house, multi-year search | Paywall, mixed bases, transcription errors |
| Association and market reports | Turnover, sell-through, sector mix | Macro data, not object-level |
| Mainstream press | Major lots and context | High-price sample bias |
| Scholarly and exhibition records | Provenance and publication | Evidence of history, not price |
Exclude social-media screenshots, figures without a sale and lot number, and any unsolicited call using a 'record' to justify a valuation of your object. A 2026 municipal commerce notice specifically warns of fabricated auction records and buyer testimonials.
2. Five misreadings
First, post-sale figures usually include the buyer's premium while terminology standards define the hammer bid as the sale price. With Hong Kong's 28% first tier effective September 2026, mixing bases distorts comparisons. Second, estimates, reserves and opening bids are not results; a lively room can still end in a buy-in. Third, generic titles cover many different objects, so check dimensions, form, marks and condition. Fourth, record currency and date. Fifth, databases and press over-represent sold, expensive lots.
Key points (text version)
- Premium basis
- Estimate as result
- Same name, different object
- Currency and date
- Survivorship bias
3. Rather than searching 'top ten auction houses'
- Licences: verify the auction operating approval and, for cultural relics, the relics auction licence.
- Sale fit: has the house run comparable specialist sales in the last three seasons?
- Comparable results: list actual same-category results, sell-through and price range.
The useful question is not 'who is largest' but 'who has actually sold comparable objects recently, to whom, and at what price'. See our 2026 market data guide.
4. A comparable-sales template
| Field | Why | If missing |
|---|---|---|
| House / sale / date | Primary source | Unverifiable |
| Lot number | Unique identifier | Same-name confusion |
| Title, date, dimensions | Like-for-like | Large-lot price on a small object |
| Condition and provenance | Explains price spread | Restored vs intact mixed |
| Currency and date | Cross-market basis | FX and cycle errors |
| Hammer or inclusive | One basis | Systematic 20%+ overstatement |
| Sold or bought in | Sell-through rate | Survivorship bias |
5. Four steps to an estimate
- Filter three to five genuine comparables
- Adjust for condition, provenance, size and sale positioning
- Build a range around the adjusted median
- Back-test against a result you already know
Filter three to five genuine comparables; adjust for condition, provenance, size and sale positioning; build a range around the adjusted median; then back-test the method against a result you already know. The point is not precision but an estimate that can be questioned and corrected.
Key points (text version)
- Filter
- Adjust
- Range
- Back-test
6. FAQ
Can an online result price my object directly?
No. Confirm the basis, currency and date, then match size, condition and provenance across three to five comparables.
Why do two sites show different figures for one lot?
Usually hammer versus premium-inclusive, sometimes FX or transcription error. Defer to the house archive.
Does an estimate equal value?
No. It is a published reference range, not a result or an offer. Below reserve means no sale.
Are 'top ten' rankings useful?
Less than licence checks, recent specialist sales and same-tier sell-through data.
Are buy-ins recorded?
Inconsistently. That is why sold-only samples overstate the market; check sell-through rates instead.
Is an unsolicited 'identical piece sold for millions' claim credible?
High risk. Authorities warn of fabricated records used in scams. Demand house, sale, date and lot number, then verify independently.
What should I archive?
Official page or PDF, catalogue entry, price list, plus the seven fields in your comparison table.
Any alternative to paid databases?
House archives, association reports, specific press coverage and scholarly provenance records; always return to the primary source for object-level prices.
7. Sources and review
- 雅昌藝術市場監測中心《藝術品拍賣全攻略》與季濤〈不要混淆拍賣場上的各種價格〉 — 定義估價、保留價、起拍價、落槌價、成交價；並指出 2012 年《拍賣術語》行業標準與市場發布習慣存在差異。
- 佳士得官方「買方酬金費率」（2026 年 9 月 1 日起生效） — 香港：落槌價首 HK$15,000,000 為 28％、其後至 HK$60,000,000 為 22％、逾 HK$60,000,000 為 15％。分段累進，僅適用於落入該區間的部分。
- 中國拍賣行業協會《2026 年春季藝術品拍賣市場白皮書》（媒體引述） — 2026 年春拍全國藝術品拍賣總成交額 286 億元、同比上升 16.3％；四大核心品類貢獻 87.2％（二手引述，僅作趨勢參考）。
- 2026 年香港春拍公開報道（福布斯中國、中華收藏網） — 香港春季亞洲藝術拍賣成交總額約 HK$9.06 億、同比上升 54％；中國書畫專場合共逾 HK$7,050 萬（二手報道，僅作趨勢參考）。
- 鄂州市商務局〈風險提示〉（2026 年 5 月） — 歸納「畫大餅—演大戲—收費用」三步套路，提示長者為主要受害羣體。
- 《中華人民共和國文物保護法》（2024 年 11 月 8 日第二次修訂）第五章「民間收藏文物」 — 第六十六至七十五條：合法取得方式、禁止買賣範圍、文物拍賣許可、拍賣前審核、如實表述義務、備案與優先購買。