work, claim, method, validity
What does the research say, and does it still hold? Bathymark turns papers into bounded claim records, then keeps original evidence, method limits, replication state, conflicting results, and current validity separate.
A paper can contain several claims, and a claim can change as data, costs, venues, and market structure change. Bathymark keeps each link inspectable instead of compressing the chain into one confidence badge.
Momentum, attention, network activity, size, and factor research, kept separate from a claim that any result is tradable now.
Open current Bathymark evidenceLiu and Tsyvinski report a strong time-series momentum relationship and attention-based return predictability for Bitcoin, XRP, and Ethereum in their historical study.
The source study reports exposure to its cryptocurrency network-factor construction and no exposure to its production-factor construction.
Liu, Tsyvinski, and Wu report that cryptocurrency market, size, and momentum factors capture cross-sectional expected returns in their study universe.
Exchange segmentation, return predictability, and liquidity research with venue and sample boundaries left visible.
Open current Bathymark evidenceMakarov and Schoar document large recurrent cross-exchange price deviations, with larger gaps across countries and evidence consistent with constrained arbitrage capital.
Wei reports that return predictability and volatility were lower among more liquid cryptocurrencies in a 456-asset cross-section.
Current working-paper evidence on transmission, with peer-review state and identification assumptions stated plainly.
Open current Bathymark evidenceA revised 2026 BIS working paper estimates that a $3.5 billion stablecoin inflow lowers three-month Treasury-bill yields on impact and more over the following days.
A 2026 BIS working paper reports persistent stablecoin-versus-FX parity gaps and estimates spillovers from exogenous stablecoin demand into local exchange rates and dollar funding costs.
Halving and time-varying efficiency research treated as hypotheses with very small independent-event counts.
Open current Bathymark evidenceKhuntia and Pattanayak report that linear and nonlinear return dependence changes across rolling Bitcoin samples, consistent with time-varying market efficiency.
A 2025 preprint using synthetic control reports a positive 2024 estimate three months after the event, but no statistically robust causal estimate for the 2020 halving.
Backtest-overfitting controls and competing anomaly-replication results that apply before a paper becomes a live claim.
Open current Bathymark evidenceBailey and coauthors argue that ordinary holdouts can be unreliable after many strategy trials and propose combinatorially symmetric cross-validation to estimate overfitting probability.
Bailey and Lopez de Prado propose the Deflated Sharpe Ratio to adjust performance evidence for multiple testing and non-normal return distributions.
Hou, Xue, and Zhang report that most of 452 equity anomalies fail under their microcap, weighting, and multiple-testing controls, with smaller magnitudes among those that survive.
Jensen, Kelly, and Pedersen report that a majority of asset-pricing factors replicate under their Bayesian model and show external validity in international and post-publication data.
The publisher-defined graph exposes the same works, claims, source locators, reported metrics, relationships, status history, and coverage zeros that appear in HTML. It is not a raw paper feed, an external standard, or an action-agent capability.
Bathymark uses durable identifiers and public standards as design constraints. A completed future replication can be packaged with provenance and exact code identifiers, but the first release does not claim an RO-Crate export or archived software artifact that does not yet exist.
Research objects should be findable, accessible, interoperable, and reusable for people and machines.
Portable provenance relationships for entities, activities, agents, derivation, and revision.
JSON-LD packaging for research data, code, workflows, people, licenses, and provenance.
Content-derived identifiers for exact source files, trees, commits, and other software artifacts.
Bibliographic identity, DOI versions, update relations, and Retraction Watch links, excluding abstract and full-text ingestion.
Catalog bathymark-research/v0.1.0, reviewed 2026-07-13. Research evidence is information, not financial advice.