← Back to command center

NILS TORNBERG. Skedar, 2 stycken, silver, Linköping 1796.

Nils Tornberg · Silver and metalware · RA Auktionsverket Norrköping

Open auction lot
47
research

Acquisition position

Current all in cost

1200.0 SEK
Current bid or estimate900.0 SEK
Modelled discount52.3%
Liquidity score84%

Potential market value

Specialist value interval

1002 to 3670 SEK
Value confidence71%
Comparable sales30

Evidence strength

38%
Identity confidence29%
Visual match0%
External references4

Market anomaly

60%
Information gap67%
Potential undervaluation52.3%

Risk profile

13%
Authenticity risk13%
Condition risk6%

Investigation rationale

Why this may be worth considering

  • maker or artist identity supplied
  • auction estimate is materially below the specialist model
  • Estimate materially below specialist value

Decision guardrails

What must be verified

  • category trap check: silver plate
  • missing hallmarks
  • missing fineness
  • missing weight
  • missing date_letter

Visual laboratory

Add and compare research imagery

Object record

Known facts and missing information

ArtistNils Tornberg
MakerMissing
MediumMissing
MaterialsMissing
Dimensions? × ? × ? cm
WeightMissing g
YearMissing
EditionMissing
MarksMissing
ConditionInskription.
ProvenanceMissing
Modelsilver_metalware_specialist 2.0.0

Längd ca 21,5 cm Vikt ca 116 gr.

Identity resolution

Artist and maker hypotheses

Ranked from independent evidence. Confirmation or rejection becomes controlled training data.

3 candidates
1 independent sources

Carl Albert von Lespilliez (German, 1723-1796) after François de Cuvilliés the elder (French, 1695-1768)

Visual 0% · Name 43% · Marks 25% · Material 40%

26% combined confidence
1 independent sources

Katsuma Antei Ryūsui, Kano Yoshinobu, Suzuki Rinsho, Komatsuken et.al 1711-1796

Visual 0% · Name 40% · Marks 25% · Material 40%

25% combined confidence
1 independent sources

James Gillray (English, 1756-1815) published by Hannah Humphrey (English, c. 1745-1818)

Visual 0% · Name 38% · Marks 25% · Material 40%

25% combined confidence

Closed deal feedback

Teach the valuation model

Enter the verified hammer price after the auction closes. Only verified outcomes become calibration data.

Learning guardrail

Truth before training

  • Unverified AI output is stored as a hypothesis only.
  • One-source matches are confidence capped.
  • Rejected matches improve source and feature calibration.
  • At least five verified closed sales are required per category.

Global research evidence

Reference material found elsewhere

4 references

Comparable logic

Specialist model explanation

Comparable IDMatchPriceReasons
1147953%1200.0 SEKidentity 57%, material 45%, size 62%, title 53%, hallmarks 45%, weight 62%
1948252%3300.0 SEKidentity 56%, material 45%, size 62%, title 46%, hallmarks 45%, weight 62%
1457852%11000.0 SEKidentity 54%, material 45%, size 62%, title 61%, hallmarks 45%, weight 62%
2206852%2400.0 SEKidentity 56%, material 45%, size 62%, title 46%, hallmarks 45%, weight 62%
1998452%1500.0 SEKidentity 53%, material 45%, size 62%, title 60%, hallmarks 45%, weight 62%
136252%5819.0 SEKidentity 55%, material 45%, size 62%, title 51%, hallmarks 45%, weight 62%
1003052%1200.0 SEKidentity 55%, material 45%, size 62%, title 47%, hallmarks 45%, weight 62%
49951%1378.0 SEKidentity 52%, material 45%, size 62%, title 58%, hallmarks 45%, weight 62%

Automatic signature module

Every artwork image is checked for signatures and marks

The local engine detects and isolates candidate regions, searches the trusted mark library and runs conservative signature forensics. Optional AI interprets unreadable regions and creates research queries without changing the forensic verdict.

Candidate regions0
Signature regions0
AI enrichmentLocal only
PipelineNot run

No automatic mark result yet

The module runs during normal analysis when a local image is available.

Marks laboratory

Isolate and identify signatures, hallmarks and maker marks

110 independent methods generate cleanup variants, stroke and symbol descriptors, library matches, trust controls and conflict warnings.

Open mark library

Match an unknown mark

Upload the tightest available crop. Include a little surrounding surface so the system can separate paint, engraving, glare and background.

Add a known reference

Signature forensics

Check consistency with known genuine variation

This does not declare a signature genuine. It measures fit against a multi-source reference set, confusing signers and natural variation.

No forensic check yet

Run mark matching first, then compare the questioned signature against a sufficiently broad trusted reference set.

Open command centerDashboard Open specialist model labModels Show extreme candidatesFilter Show priority reviewsFilter