Repository Public Research preview
1# CLAUDE.md — UQO Research WP46: The Great GARCH Benchmark23**Project ID:** `wp46_uqo` · **Folder:** `Recherche-UQO/finance/wp46_garch_benchmark` · **Repo:** `gitsrv:srv/git/wp46_uqo.git`4**Working title:** *Four Decades of Conditional Volatility Modeling: A Comprehensive Empirical Benchmark Across Assets,5Frequencies, and Market Regimes* · **PI:** Simon-Pierre Boucher (UQO) · **Language:** English.6The owner's original brief is kept verbatim in `docs/BRIEF.md` and is the authoritative text; this file condenses it and7records the decisions taken while building (see `docs/methodology_decisions.md`, `progress.md`).89## Core question10Across asset classes, horizons (1, 5, 22 days), regimes and information sets, when — if ever — do more complex conditional11volatility models reliably beat GARCH(1,1), and is any gain worth the numerical and computational cost?1213## Hard rules (non-negotiable)141. Never fabricate data, estimates, references, DOIs, p-values, run outcomes or novelty. Planned ≠ completed: `progress.md`15 and `docs/PHASE_N_REPORT.md` separate the two; every number in the paper is generated by code from `results/`.162. Pre-registration: `docs/preregistered_protocol.md` is frozen (git tag `prereg-frozen`) BEFORE any out-of-sample test17 loss is inspected. The test period is never used for selecting orders, optimizer settings, regime definitions or losses.183. Data: FinData OS (`https://findata.spboucher.ai/api/v1`, FirstRate Data private licence) is the only primary source.19 No raw bars in git, logs, figures, manuscript or release; aggregate forecasts/losses only. Keys only from20 `~/Desktop/Cluster/secrets/findata/api-key-claude-code.txt` or `$FINDATA_API_KEY`; never logged.214. Units are explicit everywhere: daily log returns in **percent**; variances in percent²; forecasts are conditional22 variances of the h-day cumulative return matched to a same-horizon positive proxy.235. Failed fits are recorded as failures (never silently replaced by a baseline); unsupported model × distribution pairs24 are recorded as unsupported, not as bad performance.256. Every accepted implementation has: unit tests (recursion, likelihood, admissibility, multi-step), a likelihood26 cross-check against `arch` 8.0.0 and/or `rugarch` 1.5-6 on a shared fixture, seeded simulation recovery, and a model card.277. No external publication, push to a new remote, SSRN deposit or paid compute without explicit authorization.28 Pushing to the private `gitsrv` mirror is routine.298. Remote-first: computation runs on the MacLustr nodes (`~/wp46_uqo` clones), the laptop is the LaTeX/mirror host.3031## Where things are32| Item | Location |33|---|---|34| Brief / decisions / progress | `docs/BRIEF.md`, `docs/methodology_decisions.md`, `progress.md` |35| Configs (frozen by the protocol) | `configs/{study,instruments,model_registry,forecast_protocol}.yaml` |36| Code | `src/wp46/{findata_client,calendars,preprocessing,realized_measures,models,estimation,forecasts,scoring,inference,diagnostics,reporting}` |37| Scripts | `scripts/{audit_findata,run_pilot,run_benchmark,build_paper,…}.py` |38| Tests | `tests/{unit,integration,simulation,replication}` (`pytest`; live tests need `WP46_LIVE=1`; R tests need rugarch) |39| Docs | `docs/{findata_audit,literature_matrix,preregistered_protocol,feasibility_report,methodology_decisions,data_dictionary,replication_guide}.md` |40| Model cards | `model_cards/<slug>.md` |41| Results | `results/{manifests,metrics,figures,tables,logs}` (forecast parquet files under `results/forecasts/` are gitignored) |42| Paper | `paper/main.tex`, `sections/`, `appendices/`, `figures/`, `tables/`, `references.bib` (Crossref-verified) |4344## Compute45Data node M3U96a (hosts FinData; `FINDATA_BASE_URL=http://127.0.0.1:8600/api/v1` there, `http://192.168.2.80:8600/api/v1`46from the LAN). Benchmark shards on M3U96a/M3U96b/M2U64/M5M36/M4M64a/M4M64c (`scripts/node_run.sh`). Cache47`~/wp46-cache` (outside git). Laptop: Python 3.14 + arch 8.0.0, R 4.6.1 + rugarch 1.5-6 for cross-checks.48