Bridging academic rigor and quantitative trading.
An independent research lab publishing, in the open, what actually works in systematic trading and what doesn't, tested at scale. Built to contribute to both worlds at once: rigorous enough for academics, honest enough for practitioners.
Everything here is free, forever. No paywalls, no sign-ups, nothing to sell. Every paper, model, and dataset is open and reproducible.
Open source
lines of research code
across 17 public repositories on GitHub, plus the Research Review and Topology toolkit codebases (release pending), counted 2026-06-18
- Python87.9%· 91,756
- Rust10.8%· 11,259
- R0.8%· 882
- Shell0.4%· 454
Open by default, fork it, cite it, build on it. Support the research
By language
Lines per project
Every public repository featured on this site, the paper reproducibility packages, the article companion code, the single-project repos, and the walk-forward backtesting framework, plus the self-contained toolkit core of the in-flight Topology program. Counts are source lines only.
- Research-ReviewRelease pending28,753 lines
- topology-toolkitRelease pending19,210 lines
The self-contained core of the Topology of Strategy-Space program, the importable Python package (I/O, strategy-space, topology, evaluation, the baseline-model zoo, clustering, the CLI apps, the figure theme). Open-sources at program completion; the program's ~43k further lines of experiment scaffolding, figure scripts and tests aren't counted here.
- dex-tradeability-study18,908 lines
- Monte-Carlo-paper10,000 lines
Reproducibility package for the Monte-Carlo filter-evaluation paper, analysis scripts (Python / Rust / R), corrected aggregate tables, and a self-contained reproducer of the floating-point summation-order pitfall documented in the May 2026 revision.
- worldcup-attention-markets9,941 lines
- quant-research-framework4,697 lines
The walk-forward backtesting & robustness framework, the original Python build.
- quant-research-framework-rs8,114 lines
The Rust port of the backtesting framework, the production build.
- quant-surface-stability1,180 lines
Pre-registered study of parameter-surface stability, smooth ridges vs brittle spikes.
- strategy-overfitting662 lines
Empirical variance decomposition of in-sample → out-of-sample Sharpe degradation.
- strategy-manifold460 lines
Geometry of strategy space, dimensionality and manifold structure.
- strategy-regime427 lines
Latent regime classification on market data, then regime-conditional strategy selection.
- tail-evt403 lines
Extreme-value analysis of cross-asset tail dependence, peaks-over-threshold with GPD fits.
- strategy-robust-portfolio396 lines
Robust portfolio construction over a strategy universe.
- strategy-stats246 lines
Population-level statistics over a universe of algorithmic trading strategies.
- strategy-rmt208 lines
Random-matrix analysis of strategy correlation matrices, noise vs signal decomposition.
- signal-is-collective204 lines
Reproducible synthetic demos behind the article “The signal is collective”.
- hc-knockoffs184 lines
Higher Criticism under dependence + Model-X knockoffs for FDR-controlled strategy selection.
- strategy-tda181 lines
Persistent homology of strategy space, topological data analysis on strategy populations.
- strategy-corrcube177 lines
Cross-asset correlation cube, pairwise correlations between asset return series.
Source lines only, READMEs, methodology notes, data files and configs aren't counted. 56,388 lines sit in the 17 public repos. Two further codebases open-source at program completion: the López-de-Prado Research Review pipelines (28,753 lines) and the Topology toolkit core (19,210 lines). The Topology program's ~43,000 further lines of experiment scaffolding, figure scripts and tests, and its Rust accelerator crates, are not counted here.
Working papers
Papers
All researchPeer-facing research with full reproducibility packages. Three studies on the boundaries of statistical edge, tested at scale across asset classes.
SSRN 6955879 · cross-market study
The Reach of the World Cup Distraction Effect
The whole planet watches the World Cup, yet the global markets that carry most of the trading barely flinch. Across 11 instruments, 5 market structures, 26 countries and 7 World Cups, 158 statistical tests turn up nothing that survives correction, and the headline effects that seem to travel are artifacts of how they are measured.
Paper · SSRN 6636018
Does a popular statistical filter actually pick winners?
A simple question, tested at scale: does acting on a widely-used permutation test help you choose winning strategies ahead of time? The answer is no, and a famous result claiming it hurts turns out to be a rounding bug.
Paper · SSRN 6858778 · Published
Can you profit from the smallest coins after real costs?
Can you make money in coins that only trade on decentralized exchanges, once realistic trading costs are paid? We tested it across 27 chains, with demos you can poke at in the browser.
Reproducible essays
Long-form articles
All articlesInteractive, figure-driven write-ups of the central empirical claims, each reproducible from a public data table.
01 · Article · Flagship
Edge is in the Process
How we turned an unprofitable pile of half a million strategies into a portfolio that actually makes money, without touching the strategies themselves. The whole trick is which ones you choose to keep.
02 · Article · Pre-registered
The partition does the work
A smarter way to build a portfolio out of a pool of strategies, held to a strict statistical bar and pitted against the standard benchmarks. It beats all of them.
03 · Article · The corpus
What “1.6 million strategies” actually means
‘1.6 million strategies’ sounds like data-mining. Here is why it is not, shown with the actual numbers behind the corpus.
Open-source models
Projects
Open the labTwelve reference implementations of the lab's M-models, random matrix theory, Higher Criticism + knockoffs, topological data analysis, tail-EVT, each with an explainer, a live demo, and code.
Eigenspectrum of the strategy correlation matrix
Marchenko–Pastur and parallel-analysis eigenspectrum of strategy correlation matrices. Reference implementation of the firm's M/01 model.
Sparse signal detection with FDR control
Higher Criticism plus Model-X knockoffs for FDR-controlled strategy selection. Reference implementation of the firm's M/02 model.
Persistence barcodes on strategy structure
H0 persistence barcode under correlation-distance Vietoris–Rips on strategy populations. Reference implementation of the firm's M/03 model.
Peaks-over-threshold and pairwise tail-coupling
Peaks-over-threshold GPD fits and pairwise tail-coupling χ on cross-asset returns. Reference implementation of the firm's M/04 model.
Decomposing the IS-OOS Sharpe gap
Variance decomposition of the in-sample / out-of-sample Sharpe gap into selection bias, parameter-choice noise, and residual skill across 10 deep-WFO crypto assets.
Does in-sample smoothness predict out-of-sample skill?
Pre-registered empirical test of whether in-sample Sharpe-surface smoothness under a fixed five-perturbation suite predicts out-of-sample skill across SOL / DOGE / BTC walk-forward partitions.
PCA + UMAP geometry of the strategy population
PCA + UMAP embedding of large strategy populations from a 90-feature metric vector, with connectivity-based separation of robust vs fragile strategies.
Hidden-Markov regime segmentation
Gaussian HMM regime segmentation on (logret, volatility, trend) features with K selected by BIC; cross-asset 4-state preference across crypto majors.
Universe-saturation of minimum-variance portfolios
Universe-saturation analysis for minimum-variance portfolios drawn from large strategy pools, comparing Ledoit–Wolf shrinkage, Huber-style robust, and sample covariance estimators.
Public output to date
The author
Daniel Gatto
- Independent researcher
- Economic Sciences candidate, UNIP
- Quantitative Systems Consultant (NDA)
Daru Finance is the public output of one researcher. The work here is what gets done after-hours, in code, with receipts.
I write systematic-trading research with a strong reproducibility bias, Python for analysis, Rust for compute, R for verification. My current work focuses on the statistical structure of strategy populations: random-matrix bounds on correlation eigenspectra, Higher-Criticism + knockoffs for FDR-controlled selection, topological indicators for cluster stability, and extreme-value theory for tail co-movement. The throughline is that the unit of analysis is the population, not the individual strategy. From the research and the twelve open-source models down to this site itself, design, code, charts, every word, it’s all a one-person build.
Read the full bioBackground, in numbers
Currently in flight
Two roadmaps in motion
What I’m building next, with explicit phase status. Click through for the deep dive on either project.
Topological Data Analysis in Quantitative Trading
H4 portfolio construction PASS at corpus scale · all 6 H12–H17 extensions tested · Topology Deployment Stack (H18/H19) next
Quant Research Framework
Research-grade backtester · WFO + regime + robustness · 5 done, 1 in-flight, 3 planned
Get in touch
Replications, corrections, methodological questions.
Academic correspondence is always welcome, replications, corrections, methodology questions.