Monolith Research
An institution of few  ·  Est. 2024

Systematic trading research, built and scored in the open.

Monolith Research is a systematic-research practice — an institution of few. What reaches the record is held to a standard hard enough that most of it does not: a live, third-party-scored model, a funding-grade research book, and an open-source gate.

The firm open-sources the gate it judges itself by, documents every retired strategy in public, and keeps only the live records it can actually defend. Signal logic still in use is kept private; the method, the evidence, and the failures are made legible. It does not manage client capital or provide investment advice.1

ETFs  ·  FX  ·  Commodities United Kingdom
Currently Running the firm as a validation discipline rather than a trading desk — maintaining the open-source promotion gate, building systematic securities research, and keeping a live, independently-scored record on Numerai. June 2026
§ 1   On Record What we’ve done. Verifiable  ·  Out-of-sample unless noted
i.
Live
Numerai model
third-party scored · positive record
ii.
Top 15
IMC Prosperity
of 2,200+ teams · global quant competition
iii.
DSR
Monolith’s gate
PSR · DSR · PBO · open source →
iv.
0
Strategies graded
through a blind out-of-sample gate
v.
3,000+
Securities screened
→ 39 graded → 5 to a fundable book
vi.
0
Open-source repositories
the gate · graveyard · microstructure →
vii.
0%
Figures reproducible
from a versioned commit + dated snapshot
viii.
0
Strategy families
documented across the research →
§ 2   The Book The edges that survived. Funding-grade research
~1.64
Combined out-of-sample Sharpe · five uncorrelated sleeves · equal-risk weighted

The gate kills most ideas. These are what it did not. Five research sleeves, each clearing the firm’s deflation gate on its own terms, and uncorrelated enough that combining them improves the whole.

Real, validated, and capital-constrained at a retail scale. This is what a desk unlocks.

Out-of-sample research results, post-cost — funding-grade, not a live track record. Real edges, but locked behind licensed data and un-tradeable on small capital. Signal logic is kept private; every figure is reconstructible from a versioned commit and a dated snapshot.

  1. 01 Insider-cluster buys Flagship Opportunistic cluster insider buying — two or more off-schedule open-market purchases on a name — as a long-only cross-sectional signal. The strongest single find: survives a skip-a-month test, near-zero beta, regime-robust. net IR 0.66 · DSR 0.94
  2. 02 Option-skew The steepness of a stock’s implied-volatility smirk as a cross-sectional signal — steep skew underperforms, shallow outperforms (Xing–Zhang–Zhao lineage). Forward-confirmed, market-neutral. net-1× Sharpe ~1.0
  3. 03 Analyst-revision Drift in sell-side consensus earnings estimates — long the largest upward revisions, short the largest downward. Weak alone; a −0.30-correlated diversifier inside the book. diversifier
  4. 04 Crude relative-value Brent–WTI mean-reversion to the structural logistics differential. Modest alone, additive in the book; 100% of its parameter-robust variants positive out-of-sample. 0.56 OOS · β −0.05
  5. 05 Residual-equity cross-section Momentum measured on the factor-stripped residual return rather than the total. Sign-stable across halves; correlation near zero with the crude sleeve. uncorrelated
§ 3   The Bar The standard a number must survive to get published here. Interactive · run the gate

A high Sharpe is the easy part. Surviving the gate is not. Deflate a headline backtest for the size of the search behind it, stress its costs, strip its market beta, and hold it out over years it never saw — most edges evaporate. Try it on the firm’s own retired flagship, Strategy 17, which headlined 1.72 and held at 0.108 over 17.6 years.

Variants searched 1
More search, more deflation — a high Sharpe found after thousands of tries is mostly luck. (Deflated Sharpe.)
Cost stress 0.0×
Spreads and slippage, inflated. A real edge survives its own trading costs.
Out-of-sample holdout 0.5 yr
How long the strategy must hold up, untouched. Short windows flatter; 17.6 years does not.
Strip market beta
Remove the part of the return that was just the market rising. What is left is the skill.
Σ Strategy 17 · deflating
1.72
Out-of-sample Sharpe
Passes the gate
0.01.02.0

The headline figure, before the gate asks a single question.

Strategy 17 — a macro-regime ensemble, retired and documented in the Strategy Graveyard. The deflation here is illustrative of the mechanism; the firm’s gate uses the full Deflated-Sharpe and PBO machinery. Every figure is reconstructible from a versioned commit and a dated snapshot.

§ 4   The Graveyard And here is the proof the bar is real. Every graded strategy · documented in full

Every strategy the gate retired, documented with the real numbers — no truncated curves, no favourable rounding. The kill-record is what makes “it passed” mean something.

OU stat-arb · deflationGold H4 · beta mirageStrategy 17 · 1.72 → 0.108CEF discount · microstructureCarry + trend · deflationDual-class spread · breadth-thinNAS/SPX intraday · search mirageMacro-sentiment · coin-flip AUC … and the rest
Read the kill-record
§ 5   The Method Three workflows kept in public view, one kept in private. Method · Evidence · Reporting

A research process held to institutional standards.

01 Research infrastructure

Versioned, reproducible, slow on purpose.

Monolith maintains a Python-based research stack for data ingestion, feature construction, hypothesis testing, and backtest evaluation across ETFs, FX, and commodities. Strategy candidates are grounded in published literature — Ornstein–Uhlenbeck mean-reversion, Engle–Granger cointegration testing, and walk-forward validation methodology. Every notebook and signal candidate is versioned; every published figure is traceable to a commit.2 The infrastructure is built to make negative results as legible as positive ones.

02 Risk & reporting

The numbers, written down before they are remembered.

A separate reporting workflow journals every test trade, tracks drawdown against pre-declared limits, and publishes a monthly statement of method and evidence. Reports are kept in plain language and archived alongside their underlying data. The aim is forensic, not promotional: anyone reading later should be able to reconstruct what happened, and why.

03 The gate

The test a strategy must survive to reach capital.

Before anything is deployed it has to clear an open-source promotion gate — Probabilistic and Deflated Sharpe, PBO by combinatorial cross-validation, a cost-and-beta strip, and a blind out-of-sample holdout. It is deliberately hard to pass, and most candidates never do. The live records the firm keeps are a third-party-scored Numerai model and a single intraday edge on a third-party-funded account.

§ 7   Open evidence The method is public. So is most of the failure.

The validation gate

The open-source promotion gate the whole firm is judged by — Probabilistic and Deflated Sharpe, PBO via combinatorial cross-validation — covered by known-answer tests. Runnable, not just described.

github.com/MonolithResearch/quant-validation →

The Strategy Graveyard

Thirty-nine strategies put through that gate, documented in full with the real numbers — no truncated curves, no favourable rounding. The kill-record is the product.

github.com/MonolithResearch/strategy-graveyard →

Microstructure study

A funding-grade order-flow / deep-LOB edge — the foresight is real; capturing it is constrained by latency and infrastructure at a retail scale. Real, and walled off.

github.com/MonolithResearch/deeplob-microstructure →
§ 8   Research note From the current month’s working papers. Note 09 / 2026
Abstract · Direction Note No. 09  ·  June 2026

A research practice, built to a hard bar.

By the Monolith desk · Research note, June 2026 · Approx. 3 minute read

Monolith Research runs as a systematic securities-research practice — an open-source validation gate, a documented research process, and a live, independently-scored model on Numerai. As the work matures, the firm intends to apply for FCA authorisation to expand.

The gate decides what is worth deploying — and most of what the firm builds does not clear it. That discipline is the product. On the method
Continue reading →
Filed under · Direction · Research · Authorisation Monolith Research  ·  UK
Also in the working file
All notes
§ 9   The Founder An institution of few. One discipline. The Team · UK

Bilal Malik

Founder · Monolith Research

London-based, reading Finance at Lancaster University. He founded Monolith Research to study liquid markets — ETFs, FX, and commodities — the way a journal studies a result: write the hypothesis down, test it against the record, and let the statistics, not the narrative, decide.

He set the method and built the core research stack end to end; the firm is led from the front and carried by a small, growing team. What survives that standard is published with the evidence behind it — the kind a careful reader can audit a year later, not the kind that lasts only while the screenshots are fresh.

The team carries institutional market exposure through Susquehanna International Group and Schroders — quantitative trading, market-making, and multi-asset portfolio construction on both sides of the market.

Meet the team
§ 10   Correspondence Correspondence welcome. Replies within three working days
Correspondence

Correspondence welcome — for research, technology, employment, and collaboration conversations.

bilal@monolithresearch.uk