# Box-and-Papers Paired-Sales Pilot — preregistered protocol

**Status:** collection not started; no result or premium is claimed  
**Protocol date:** August 24, 2026

## Research question

Within one narrowly defined, sufficiently traded watch reference and a fixed observation window, is documentation status associated with a difference in comparable completed-sale prices after visible configuration, condition, channel, date, bracelet, and price basis are recorded?

The pilot is designed to test evidence quality, not to manufacture a universal “full set premium.” If the sample is too small or missingness is severe, the correct output is an inconclusive result.

## Preregistration decisions

Before collecting prices, a named researcher and independent reviewer must freeze:

- one or at most two exact references and case materials;
- observation start/end dates;
- eligible public completed-sale venues and their terms of use;
- permitted source-link and data-republication practice;
- currency conversion source and date convention;
- minimum visible-condition fields;
- document classes and accessory classes;
- duplicate/relist detection rules;
- exclusion reasons; and
- minimum sample size for descriptive analysis.

## Document classes

Record these separately: original named or numbered guarantee; undated/blank guarantee; later archive extract; service invoice; service card; retailer receipt; box only; manuals/tags only; claimed “full set” without visible support; and unknown. Never turn all of these into one binary field during collection.

## Inclusion checklist

- [ ] Exact reference/material can be supported.
- [ ] Sale is completed rather than active, withdrawn, passed, or merely estimated.
- [ ] Price basis is labeled hammer, all-in, or fixed transaction result.
- [ ] Sale date, currency, channel and source URL are recorded.
- [ ] Bracelet/strap and major accessories can be classified.
- [ ] Visible dial, case and component condition are recordable under the frozen rubric.
- [ ] Document status is visible or explicitly unknown; seller prose alone is not silently upgraded to proof.
- [ ] The record is not a duplicate, relist, or result copied from another venue.
- [ ] Collection and proposed sharing comply with the source's current terms and applicable rights.

## Analysis plan

Publish counts and missingness first. Show distributions by price basis and channel before combining anything. Any paired or matched comparison must document the matching variables and unmatched exclusions. Report nominal and, if responsibly possible, currency-normalized values; do not imply causation. A result may reflect condition, seller quality, venue, timing, provenance, photography, or unobserved differences.

## Stop rules

Stop and report the limitation if the permitted sample is too small, document classes cannot be verified consistently, price bases cannot be normalized, or the available records are concentrated in one biased venue. Do not relax the protocol after seeing a desirable result.

## Required release package

Protocol, frozen schema, data dictionary, source/rights note, raw permitted records, transformation script, exclusion log, reviewer reproduction, limitations, version number, and corrections address. No dataset has been collected in this repository yet.