FOR ALLOCATORS · ENGAGED AND PAID ONLY BY YOU
Before you allocate to a systematic manager, an independent reviewer reproduces what the deck claims — from the manager's code where it can be shared, from the documents where it cannot — and reports what holds, what does not, and what is still unresolved. You engage us; you pay us, per decision; the manager under review pays nothing.
Founder-led · One named reviewer · Fixed fee per decision, no subscription · No fee from managers · NDA first · Reply within 24 hours
No sales pitch. No obligation. We say no when the date or the materials will not support a useful read.
Large single-family offices run diligence like institutions. Most family offices, OCIOs, and smaller funds of funds do not have a quant on staff — yet they are asked to approve allocations to managers whose case rests on a backtest, a model, or a track record they cannot reproduce. This page is for that desk.
Managers we read: systematic equity and futures, trend and CTA programs, machine-learning alpha, statistical arbitrage, options overlays, QIS providers, digital-asset and DeFi quant strategies, and the systematic sleeves of multi-strategy platforms — the technical layer only.
Are you the manager rather than the allocator? Get your own read before allocators ask →
Screening, pre-allocation, and monitoring need different depths. Each engagement is bought per decision — no subscription, no minimum number of managers — and is fixed in scope and fee before work begins. The fee does not depend on what we find.
| Engagement | Desk Read | Model Preflight | Independent Model Validation |
|---|---|---|---|
| Stage | Desk ReadScreening, or when the manager will not share code | Model PreflightPre-allocation, one model | Independent Model ValidationAllocation already carrying weight, or a committee that needs the full record |
| Question it answers | Do the documents hold together, and what should we ask before going further? | Is there a material problem worth a deeper look? | Can this model be trusted for the decision it supports? |
| What the manager provides | Documents only: deck, backtest report, tear sheet, returns, trade list if available, DDQ answers on research process | Code at a pinned state, one data snapshot, and a working environment with read-only access | Full code, data lineage, documentation, and environment |
| What we do | Six checks: recompute the headline statistics from the returns supplied; test statistical significance and multiple-testing exposure; check cost, capacity, and turnover plausibility; compare backtest with live where live exists; read the described methodology for look-ahead, survivorship, and selection red flags; check consistency across documents | Reproduction attempt of one claimed result in the manager's environment; timing and leakage checks; validation-design review; up to five principal findings | Full review against the public review standard: methodology, data lineage, validation design, costs and capacity, parameter stability, version control; complete issue register |
| Deliverable | Findings note (2–4 pp) with one-page summary, question list for the manager, and a 45-minute debrief | Findings note (3–6 pp) with one-page summary, evidence from the reproduction attempt, and a 60-minute debrief | Findings memo, issue register with severity and decision impact, reproducible evidence package, one-page summary, and debrief |
| Timeline | Five business days from complete materials and payment, plus up to three business days for the manager's factual check unless you waive it | Five to ten business days after environment acceptance and payment | Typically two to six weeks, agreed in writing |
| Fee | $2,900 fixed, paid before work starts · 50% credited toward a Preflight of the same manager within 60 days | $4,500 fixed, invoiced after environment acceptance · $2,250 credited toward a full validation within 60 days | From $15,000, scoped in writing · 50% on the signed scope, 50% on delivery |
| Limitation stated in the note | "Based on documents supplied by the manager; no code was executed and reproducibility was not independently verified." | Scoped to one model, one pinned state, one execution path | None beyond the agreed scope |
Monitoring and re-validation. Annual re-validation of a manager already reviewed, and a drift check when the manager ships a new model version, are scoped on request at a fraction of the initial fee.
Co-sourced on request. Your analyst can sit in the reproduction: the checks, the harness notes, and the findings stay with your team afterwards. Same fee, same timeline.
Operational due diligence checks the plumbing. Investment due diligence checks the story and the people. The model in between — the code that turned data into the track record — is usually accepted on the manager's word, because nobody on the buy side has the time or the seat to run it. Three reasons that is no longer comfortable:
A technical read of someone else's model needs their cooperation. Managers raising capital routinely agree under NDA — it is the same request a larger allocator's quant team would make. When they will not share code, we say so and read the documents instead, with the limitation printed on the first page.
Where we have done prior work for the manager under review, we disclose it before scoping and, where it was material, recuse and recommend an independent reviewer.
Six checks in every Desk Read · Six review areas in the public standard · Up to five principal findings, never padded · One named reviewer from scoping to handoff · Desk Read: Five business days from complete materials and payment, plus up to three business days for the manager's factual check unless you waive it
Each maps to a check we run. Ask them first; the answers are often the whole story.
If the answers are complete and consistent, a Desk Read will say so. If they are not, you already know where a Preflight should start.
Every engagement ends with the same first page: the decision the read supports, the principal findings ranked by decision impact, and what remains unresolved — written to be forwarded as is. Behind it sits the evidence: what was reproduced, what was not, which checks were run and how, and — for Preflight and Validation — the artifacts a second reviewer could rerun.
Read the Desk Read sample → SAMPLE-002: a documents-only review of synthetic materials. No manager strategy code was executed; this is not client work or a real track record.
Illustrative · synthetic data · not client work · not a real engagement · not a track record
Model & Backtest Audit — Findings Memo
A deliberately flawed synthetic strategy is rebuilt on point-in-time inputs so the decision-maker can see what survives.
| Metric | Reported | Reproduced |
|---|---|---|
| Sharpe (annualized) | 1.8 | 0.7 |
| Annual return | 14% | 6% |
| Max drawdown | −9% | −21% |
| Hit rate | 58% | 52% |
ConfirmedHigh
A fundamental feature was joined on report date rather than its as-available date.
We validate models we did not build. If StatGazer designed or materially modified a model, we say so and recommend an independent reviewer for its validation. When an allocator engages us, we take no fee from the manager under review. The fee is fixed and does not depend on what we find; findings are reported as found, and there is no "pass" product. Remediation after findings is separate, optional work — and any later re-validation of a model we helped remediate is disclosed as such in the findings note.
We do not select managers, build portfolios, run allocation mandates, or keep a library of manager reports to resell. You engage and pay us; the manager under review pays nothing. We deliver the note to you, and to your consultant if you wish. A Desk Read reviews supplied materials without executing the manager's code and states on its first page that reproducibility was not independently verified. A Preflight or full validation includes a scoped attempt to reproduce the agreed results in the manager's environment, with the agreed code and access.
For a Preflight or a full validation — yes: we need code at a pinned state and an environment with read-only access, and only the manager can provide them. Managers raising capital routinely agree under NDA; it is the same request a large allocator's quant team makes. If the manager declines, a Desk Read works from documents alone and states that limitation on its first page.
A technical read adds an evidence check to your existing diligence. A Desk Read checks headline statistics, costs, selection and capacity claims in the supplied documents; it does not execute the manager's code. A Preflight or full validation includes a scoped reproduction attempt in the manager's environment, with the agreed code and access. You engage and pay us; the manager under review pays nothing. We deliver the note to you, and to your consultant if you wish.
One reviewer. The founder scopes the engagement, runs the reproduction, writes the note, and takes the debrief — no juniors, no hand-offs. That limits capacity to one engagement at a time, which is why the date is confirmed before you commit.
No. Document-summarising tools tell you what the manager wrote; a technical read tests whether it holds, by reproducing the result. We use software to run checks and to keep the evidence reproducible; the judgement, the findings, and the signature are one person's.
We sign a two-way NDA with you and, where the manager requires it, the manager's own form. Findings are delivered to you only. We do not resell manager reviews and do not keep a report library.
The manager receives the factual findings, not our conclusions, for a check lasting three business days before the note is final. Corrections of fact are incorporated; disagreements are recorded as the manager's position. Conclusions remain ours, and the note is delivered to you. You may waive the factual check. For a Desk Read, the window starts when we send those findings and adds up to three business days after the five-business-day review. If no response arrives by the end of that Desk Read window, we record that and finalize the note using the available materials. We confirm the final delivery date before payment.
Yes, on request. In a co-sourced read your analyst joins the reproduction; the checks, the harness notes, and the findings stay with your team afterwards. Same fee, same timeline.
No. We report what the evidence shows about the model and the claim. We do not recommend whether to invest, size a position, or select a manager, and nothing we deliver is investment, legal, or tax advice.
No. ODD covers custody, valuation, service providers, compliance, and cyber. We cover the model and the claim: reproduction, data timing, validation design, costs and capacity. The two are complementary; most allocators run both.
Not on allocator-engaged work. If we have previously worked for the manager under review, we disclose it before scoping and, where it was material, recuse.
A Desk Read takes five business days from complete materials and payment, plus up to three business days for the manager's factual check unless you waive it. A Preflight takes five to ten business days after the manager's environment is accepted and payment is received. One reviewer works one engagement at a time. We confirm the final delivery date before payment and say no if it cannot be met.
Then you have the evidence that it was checked, and a written statement of what remained unresolved. The fee is fixed and does not depend on the outcome.
That is a finding. Cost, liquidity, and capacity assumptions are checked against the proposed allocation, not the manager's backtest size.
By Stripe invoice — card or bank transfer. A Desk Read is paid before work starts. A Preflight is invoiced after the manager's environment passes acceptance, and scheduling is confirmed on payment. A full validation is 50% on the signed scope and 50% on delivery. Institutions can be invoiced directly; W-9, certificate of insurance, and NDA are available on request.
Founder · Direct delivery
The person who scopes your review is the person who runs it.
StatGazer is my one-person firm by design. I validate models for investment teams and review research statistics for people who publish — one reviewer across econometrics, statistics, machine learning, and financial engineering, with twenty years of private teaching and consulting behind every engagement.
20 years teaching & consulting, since 2006 · 100+ students from five countries · NDA-first, founder-only access · Insured — E&O and cyber (Hiscox)
Credibility ledgerEvgenii Azarov · Founder, StatGazer LLC · hello@statgazer.com · LinkedIn (opens in a new tab)
Vendor onboarding in one email. W-9, certificate of insurance (professional liability and cyber, underwritten by Hiscox), NDA, and a one-page engagement letter — sent on request so your compliance desk can open the file the same day.
StatGazer provides independent technical review engaged by the allocator. Manager technical due diligence is not operational due diligence, not a financial-statement audit, not a regulatory attestation, not a certification, and not investment, legal, or tax advice. Findings inform your own decision process; we do not recommend whether to invest.
