Violin Auction Trends Analytics

Interactive results: https://www.ruederome.com/trend.php

This repository provides the estimator, the derived per-maker results, and a data template for measuring the long-run real price performance of historical violin makers. It is designed so that any reader can reproduce the analysis from publicly available auction records.

---

Data provenance and scope

The dataset of individual auction results used in the paper was compiled by the author from publicly available auction results (hammer prices, lot descriptions, sale dates), which auction houses disclose at the time of sale. The compiled transaction file is not redistributed in this repository. What is provided here is sufficient to reproduce the analysis from equivalent public sources:

• the estimator (analyze.py) and supporting tools;
• the derived results the paper cites (maker_cagr_master_list.csv) — per-maker aggregates only (computed CAGRs, sale counts, spans), containing no transaction-level prices;
• the biographical metadata (makers_meta.json);
• an empty data template (all_violin_sales_template.csv) showing the input schema.

To reproduce or extend the analysis, assemble your own transaction file from public auction records, format it to the template schema, and run analyze.py. The maker-ID system used here follows Tarisio's public numerical index (a numbering convention, not redistributed data); see below.

---

1. Design philosophy: why numerical maker IDs

Matching historical maker names across sources is error-prone — spelling variants, accents, and translation differences routinely split one maker's sales across several records. To make the pipeline robust, every transaction is anchored to a standardized integer maker_id rather than a name string.

The project uses Tarisio's public numerical indexing as that baseline. Storing rows by integer ID ensures name variants do not fragment a maker's record, and lets results be logged directly from public archives without a custom name-translation layer.

[ all_violin_sales.csv ]  <-- linked via maker_id -->  [ makers_meta.json ]
  maker_id (int)                                          maker_id (int)
  USD / EUR / GBP (price)                                 maker_name, country
  Sale Date (year)                                        birth, death, century
  Type (optional)                                         bio (optional)

The metadata in makers_meta.json extends the bare ID index: birth/death years cross-referenced against luthiery dictionaries and corrected where auction records disagreed, plus country and century classifications (which the bare index does not provide) to enable macro-level analysis.

---

2. Method (analyze.py)

The script estimates inflation-adjusted real growth for deceased makers with a settled sale record, via OLS log-regression of real price on sale year.

1. Input. Loads an auction CSV (searched in order: auction_data_enhanced.csv, all_violin_sales.csv, auction_data.csv) and makers_meta.json. Standardizes maker_id to integer; excludes bows if a Type column is present; parses currency values to numeric.
2. Active span. Start_Active = birth + 25 (default 1500 if birth unknown); End_Active = death, or birth + 85 if death unknown. Is_Dead = True if a death year exists or birth + 85 < 2026. Sales before Start_Active are dropped as misattributions.
3. Century. A manual century value in the JSON is respected and preserved; if missing, a fallback is derived as (birth + 35) // 100 + 1.
4. Inflation adjustment. Raw prices are deflated to constant 2026 currency using the selected index (CPI by default).
5. Growth model. Groups by maker; keeps makers meeting the liquidity filter (default ≥ 10 sales over ≥ 15 years); fits np.polyfit OLS to ln(real price) vs. year for the Real CAGR (a nominal rate is computed alongside).
6. Output. Text summaries, Markdown reports (violin_market_statistics.md, market_report.md), the master results CSV (maker_cagr_master_list.csv), and Matplotlib plots.

---

3. Data pipeline: raw entry to analysis

Raw collection and final analysis are deliberately separated to protect data integrity.

[record_sales.py] -> YYYY-MM-DD_violin_sale_RAW.csv  (hammer price + currency; USD column blank)
        |
        v
[manual spreadsheet step] -> add buyer's premium, convert to USD
        |
        v
   all_violin_sales.csv
        |
        v
   [analyze.py] -> real-CAGR OLS models + reports

Step 1 — record_sales.py. Enter auction catalogue results; choose the session currency (USD / EUR / GBP). Entries are written to Hammer Price and Currency, leaving the USD column blank. Output is versioned per day (_1.csv, _2.csv, …) so nothing is overwritten.

Step 2 — spreadsheet. Open the raw CSV and populate USD as USD = Hammer Price × (1 + buyer's premium) × exchange rate. Save as all_violin_sales.csv in the repository root.

---

4. File formats

all_violin_sales.csv (input — columns marked * are required by analyze.py)

Column	Format	Notes
maker_id *	integer	matches key in makers_meta.json
maker_name	string	reference helper
Sale Date Standard *	YYYY-MM-DD (or 4-digit year)	transaction timing
USD *	integer / blank	processed USD value incl. premium
EUR, GBP	integer / blank	optional
Type	string	used for bow exclusion
Auction House	string	ignored by analytics
Hammer Price	integer	original nominal value
Currency	string	unit selected at recording
Comment	string	lot notes

A minimal file needs only maker_id, USD (or EUR), and a date column.

maker_id,maker_name,Sale Date Standard,USD,EUR,GBP,Type,Auction House,Hammer Price,Currency,Comment
1042,Giuseppe Pedrazzini,2018-05-12,61200,,,Violin,Tarisio,45000,EUR,Sartory cert.
2081,Charles Jean Baptiste Collin-Mezin,1995-04-15,22350,,,Cello,Vichy,12500,GBP,Original pegs.

makers_meta.json (metadata — JSON array of objects)

maker_id (int), maker_name (str), country (str), birth / death (int or null), century (str or null — manual value takes precedence in analysis), bio (str or null).

[
  {
    "maker_id": 1042,
    "maker_name": "Giuseppe Pedrazzini",
    "country": "Italy",
    "century": "20",
    "birth": 1879,
    "death": 1957,
    "bio": "Worked primarily in Milan."
  }
]

maker_cagr_master_list.csv (output)

Per-maker aggregates: maker_id, maker_name, country, century, N_Sales, Span_Years, First_Sale, Last_Sale, Real_CAGR_%, Is_Dead. No transaction-level prices.

---

5. Metadata tool (up_meta.py)

A command-line utility to safely view, add, or edit records in makers_meta.json.

• Normalization. If the JSON is an object keyed by ID, it is converted to the list-of-objects form analyze.py expects, injecting maker_id.
• Lookup. Prompts for a numeric maker ID with safe type-casting.
• Editing. New IDs prompt for all fields; existing IDs show current values, and pressing Enter keeps a value unchanged.
• Century. Writes a computed century rather than a blank: "Unknown" if no birth year; "21" if born after 1950 (unless a death year before 2000 gives "20"); otherwise (birth + 35) // 100 + 1.
• Safe save. Sorts by maker_id and writes UTF-8 with 4-space indentation.

---

6. Requirements

Python 3.9+ with:

pandas
numpy
matplotlib

pip install -r requirements.txt
python analyze.py

---

7. Contributing

The database does not yet cover all maker IDs, and bio is often empty. Contributions are welcome:

1. Fork or branch the repository.
2. Update makers_meta.json (use up_meta.py to avoid formatting errors).
3. Open a pull request.

Corrections to dates, countries, or missing makers are especially useful.

---

License

Mozilla Public License 2.0. See LICENSE. © 2026 Frédéric Levi Mazloum.