Monolith Research
Risk And Reporting.
The discipline of position sizing, drawdown monitoring, and honest record-keeping.
Monolith Research conducts live capital testing, at small size, under a written reporting protocol. Risk rules are pre-declared in writing. Drawdown is tracked daily. Performance is recorded monthly. The exact sizing and signal logic are kept private; the framework and the trail are not. The company does not manage client capital, does not solicit investors, and does not offer investment products.
The first rule of the
research desk is that
size is decided before entry.
Position size is set by a volatility-targeted rule, calibrated so that a single test trade contributes a small, pre-declared share of total portfolio variance. Sizing is mechanical: it is computed from recent realised volatility and a fixed risk budget, not from the strength of a view.
Per-instrument exposure is capped. The same instrument cannot be entered twice in disguise — correlated tickers count against the same bucket. Overnight concentration in single-name names is not permitted under the current rules; the desk is biased toward broad-market ETFs, liquid FX crosses, and front-month commodities.
Exact sizing parameters — the risk budget, the volatility window, the per-bucket caps — are kept private. What is public is the principle: every test trade has a size that was written down before the trade was placed, and that size is reconcilable against the broker statement at end-of-day.
A drawdown is not a
vibe. It is a
number with a stop attached.
Drawdown is computed daily against the high-water mark of the test account and logged into the reporting journal alongside the broker statement. A daily alert is generated whenever drawdown deepens beyond a pre-declared threshold; a second alert is generated when the month-to-date figure crosses its own threshold.
Drawdown beyond X% — defined and recorded privately — triggers a research-pause review. The desk halts new test trades, sits in cash, and works only on post-mortem analysis until a written review is filed and accepted. The protocol is mechanical: it does not negotiate with the operator.
A monthly drawdown summary is written into the reporting record — peak-to-trough during the month, time-to-recovery if applicable, and the instrument families involved. The aim is forensic. A reader picking up the record later should be able to see exactly when the desk was sitting on its hands, and why.
Written down
before it is
remembered.
A monthly performance summary is produced at close-of-business on the final trading day of each month. It records month-end P&L on the test account, drawdown statistics, exposure breakdown by instrument family, and a written note on what was tried and what was paused. Each monthly file is committed to a versioned record alongside the underlying data snapshot.
Records are not published openly. Posting monthly P&L in public would risk being mistaken for marketing, and Monolith Research is not in the business of marketing returns. The record exists for forensic reasons: to be reconstructible, to be checkable, and to be shared on request with genuine principal, backing, or employment enquiries.
Where a figure cannot be reconstructed from the underlying notebook and data snapshot, it is withdrawn rather than re-cut. The reporting workflow treats absence as a more honest answer than a tidied-up number.
Four documents,
kept in plain language.
- I. Test trade journal
- Entry, exit, size, instrument, and the written rationale that was recorded before the trade was placed. Each entry is timestamped against the broker statement.
- II. Daily reconciliation
- The journal is reconciled against the broker statement every trading day. Discrepancies are noted before the day is closed; the reconciliation is filed alongside the statement.
- III. Monthly summary
- Month-end P&L, drawdown statistics, exposure breakdown, and a short written note on the month. Produced on the final trading day; versioned and archived.
- IV. Quarterly review
- A longer note covering what was tried, what was stopped, what changed in the rules. The quarterly review is the place where the protocol itself is revised — never mid-month, never mid-trade.
Three trails,
all kept in parallel,
each one able to check the others.
The record is built so that a serious third party — a principal, a backer, an auditor — could in principle reconstruct what happened without taking the operator’s word for it. Three trails are kept in parallel, and they cross-check each other.
- Broker statement The authoritative ledger. Daily statements are the primary record of what was actually executed. Every figure in the monthly summary ties back to a line on a statement.
- Research code & commits The Python research stack and the reporting notebooks are versioned in Git. Each published figure cites the commit hash and the data snapshot that produced it. The commit log is a dated, immutable trail of when each decision was made.
- Public news posts Method notes and short public posts on this site are themselves a versioning trail — a dated, externally-visible record of what was being thought about, in which week, in which order. Posts are not edited after publication.
None of this is offered as an audit in the formal sense. It is the working material an auditor would ask for first.
The monthly file is not posted in public, but it is sent to genuine enquiries.
bilal@monolithresearch.uk