Skip to main content
Validraft

Independent validation desk

A descriptive research report. No black-box score. No investment advice.

You have a
tradingWe validate it.

A backtest that looks good isn’t one you can trust yet. Send us your hypothesis — we scope the data, assumptions and checks needed to study it, then deliver an independent report with the evidence and its limits. Research and final validation have distinct requirements. We confirm deliverables and timing before you commit.

Explore sample reportsCheck feasibility

Start with a research question

Explore research families

Backtesting is how a trading idea earns trust: replaying a strategy over decades of historical market data to measure how it would have behaved — before any capital is at risk. Validraft is built for rigorous, reproducible quantitative research. You don’t need a fully coded algo: discretionary, systematic, and hybrid hypotheses all start the same way — a brief in plain English. We first check whether the required data and execution model support the question. We then agree on research, candidate development or final validation. Protected out-of-sample testing applies to a frozen candidate in a supported validation scope.

We group strategies by the type of edge they claim: price behavior first, then order flow, news/NLP, relative value, fundamentals, insider disclosures, macro regimes, congressional disclosures, event catalysts, and alternative data. Pick an edge group to see concrete example backtests, or describe your own idea and we’ll scope it.

  • Price behavior
  • Order flow & microstructure
  • News sentiment & NLP
  • Relative value & stat-arb
  • Fundamental & factor
  • Insider trading
  • Macro & regime
  • Congressional trading
  • Event-driven catalysts
  • Alternative data
01Edge group

Price behavior

Includes Trend & momentum · Mean reversion · Breakout & ORB · Gap trading · Swing & multi-day

Trend, reversal, breakout, gap, and swing behavior.

Open 100 examples→
Browse all edge families

Don’t see your exact approach? Discretionary playbooks and partial rules are exactly what the brief is for — we scope what can be tested honestly before any compute runs.

How it works

Inside the validation engine

Every brief travels the same eight-stage architecture — five layers from immutable, point-in-time data to a signed-off report, held together by a lineage spine that tracks every asset and check. Step through the blueprint, or let it walk you through itself.

Validation engine

STAGE 01 / 08 · L0 · INTAKE

Brief & scope

STAGE 01 / 08L0 · INTAKE

Brief & scope

Your , captured and scoped before any compute runs.

You describe the asset, timeframe, rules, and what success looks like — in plain English. We translate it into a precise, testable specification and confirm data coverage and feasibility, so the engagement is scoped to what the data lake can genuinely support.

  • Hypothesis intake
  • Feasibility review
  • Tier scoping

Validation framework

Frozen, then tested. Never tuned to pass.

Your candidate is frozen before validation starts. It then runs through a fixed sequence of gates in six phases — from data integrity to a named reviewer’s sign-off. Each gate records what it observed against its threshold, and the report renders one verdict.

  1. 01

    Scope before compute

    Missing material data stops the run at preflight, not in the report.

  2. 02

    Freeze before out-of-sample

    Parameters, costs and data window are locked before protected evidence opens.

  3. 03

    Only what applies

    The strategy family decides which gates are required, diagnostic or not applicable.

  4. 04

    One decision source

    The report renders the pipeline verdict. It cannot recompute or soften it.

Explore the public control mapView sample reports

Validation checklist

Illustrative run · not client data

20-day false-breakout reversal — US equities · Validation · Frozen candidate

  1. Preflight & data integrity

    1. 01

      Data quality

      NaN 0.02% · duplicates 0 · gaps 0.1%

      PASS
    2. 02

      Corporate actions

      splits & dividends adjusted · 0 unadjusted gaps

      PASS
    3. 03

      Event leakage

      price-only rule · no event inputs

      Not applicable

  2. Feature & signal diagnostics

    1. 04

      Feature redundancy (PCA)

      single-signal rule

      Not applicable

  3. Performance, benchmark & deployability

    1. 05

      Performance thresholds

      Sharpe 1.31 ≥ 0.80 · max DD 8.4% ≤ 20%

      PASS
    2. 06

      Execution-cost reconciliation

      spread + commission applied per fill

      PASS
    3. 07

      Capacity

      participation 0.6% ≤ 2.0%

      PASS
    4. 08

      Benchmark alpha / beta

      beta 0.08 · excess Sharpe 0.74 ≥ 0.20

      PASS

  4. Statistical validation

    1. 09

      Multiple-testing-adjusted Sharpe

      12 trials · adjusted 0.92 ≥ 0.50

      PASS
    2. 10

      Walk-forward

      7 of 8 test folds positive ≥ 60%

      PASS
    3. 11

      PnL permutation

      p = 0.012 ≤ 0.05 · 2,000 null paths

      PASS
    4. 12

      Deflated Sharpe / PBO

      DSR 0.97 ≥ 0.95 · PBO 0.21 ≤ 0.30

      PASS

  5. Stress & regime robustness

    1. 13

      Regime stress

      high-vol regime Sharpe 0.71 ≥ 0.30

      PASS
    2. 14

      Historical crises

      7 windows tested · 2022 rate hike −6.1%

      WARN
    3. 15

      Monte Carlo path risk

      p95 max DD 13.2% ≤ 20%

      PASS

  6. Governance & approval

    1. 16

      Reviewer approval

      named sign-off recorded

      PASS
GO

Decision rendered by the pipeline

14 gates run · 1 advisory warning to read · 2 not applicable

The verdict

Three outcomes. All three are deliverables.

A GO authorises technical progression within the declared scope. It is not a profitability guarantee, a safety claim or approval for live trading.

  • GO

    Every blocking gate passed and a named reviewer signed off the result.

  • Conditional GO

    Gates passed, but a documented, ticketed threshold override is on record and must be read with the result.

  • NO GO

    At least one blocking gate failed. You still receive the report — which gate, which value, and why.

Post-validation monitoring

Keep the accepted baseline in view

A delivered strategy can drift. On the cadence you choose, the desk compares new observations with the baseline in your report and posts a descriptive snapshot to your workspace: health state, rolling Sharpe, drawdown against the stress envelope, and a short note on what changed.

Three health states
Healthy, Watch or Signal decay — always read against the validation baseline in your report.
Your cadence
Weekly, monthly or quarterly desk updates, depending on the monitoring plan.
Descriptive flags
Notes describe drift; they are never trading instructions. No paper or live trading.
Open workspaceMonitoring plans

Client workspace · Monitoring

Opening-range breakout — NQ

Breakout & opening range · Monitor Standard

ActiveHealthy

Cadence

Weekly

Started 2026-05-14

Next update

2026-06-04

Sharpe full

1.42

Sharpe 30d

1.31

Max DD

-8.4%

Updated

2026-05-28

Validraft desk

Desk context

Weekly descriptive tracking against the delivered validation baseline.

Equity curve

Rolling Sharpe

Current desk notes

  • Rolling Sharpe within expected band. · 2026-05-28

Snapshot history

2026-05-28

Healthy

Sharpe 30d 1.31 · Max DD -8.4%

The monitoring card of the client workspace, rendered with sample data. Your workspace shows your own delivered engagements.

Data

Provider-level data lake snapshot

Every brief is tested against the datasets your hypothesis actually needs — price, fundamentals, macro, sentiment and alternative signals, on the same immutable, point-in-time lake our research desk runs. Coverage includes retained historical snapshots. Several feeds are paused; freshness, date coverage and any required refresh are checked for each engagement before its scope is accepted.

1990200020102020Now
  • Databento
    2012
  • Massive
    2003
  • FMP
    1990
  • FRED / ECB
    2000
  • FMP News
    2015
  • Quiver
    2023
  • Databento
  • Massive
  • FMP
  • FRED / ECB
  • FMP News
  • Quiver
Databento
Price

Databento

since 2012

CME futures

CME futures data across continuous rolls: minute bars, ticks, L2/L3 order book, and contract metadata.

ESNQCLGCSINGRTYYMZBZNZTMESMNQ
  • OHLCV 1-minute bars · ES/NQ/CL complete; other symbols partial2012
  • Trades · tick data, v.0 and legacy c.0 rolls2025
  • MBP-10 / MBO · ES and NQ order book2026
  • Definitions, statistics, status · contract metadata snapshots2026
Open provider inventory
Massive
Price

Massive

since 2003

US equities

US equity cross-section with trades, minute aggregates, adjusted bars, reference, corporate actions, and processed daily bars.

  • Trades — daily cross-section · US equities2003
  • Minute aggregates · raw and adjusted2003
  • Corporate actions · splits, dividends, IPOs
  • Fundamentals · ratios, float, short interest, statements
Open provider inventory
Financial Modeling Prep
FXCryptoFundamentals

Financial Modeling Prep

since 1990

Fundamentals & multi-asset markets

The broadest static snapshot in the lake: fundamentals, events, ownership, index membership, ETF holdings, FX, crypto, and COT.

  • Company fundamentals · annual, quarterly, TTM statements2003
  • Analyst coverage · estimates, grades, price targets
  • Earnings · calendar and surprises
  • Ownership · 13F, institutional, insider trades
  • Index and ETF data · constituents, delisted, holdings, weightings
  • FX, crypto, macro, COT · multi-asset and derivatives context1990
Open provider inventory
FRED, ALFRED, and ECB
Macro

ALFRED / FRED / ECB

since 2000

Macroeconomic indicators

Vintage-aware macro data: roughly 231 FRED/ECB series with point-in-time layouts for regime and macro overlays.

  • Vintage macro series · ~231 FRED/ECB series2000
  • Treasury and policy rates · FRED / ALFRED
  • ECB macro series · European macro context
  • Derived macro regimes · processed features where materialized
Open provider inventory
Financial Modeling PrepFinBERT
Sentiment

FMP News

since 2015

News & sentiment

FMP news streams across equities, press releases, forex, macro, and crypto, then locally classified with FinBERT sentiment scoring.

  • Stock news2015
  • Press releases2019
  • Forex news2018
  • General news2020
Open provider inventory
Quiver Quantitative
Alternative

Quiver Quantitative

since 2023

Congressional trading

Congressional trading exports preserved as snapshots and normalized into event ledgers for alternative-data tests.

  • Congressional trades · normalized event ledger2023
  • Workbook snapshots · immutable XLSX source snapshots
Open provider inventory

Coming soon

Next provider integrations

These feeds are not part of the active lake yet. They are the next provider candidates we plan to evaluate, license where required, normalize, and document before they can be used in validation.

Benzinga
News

Benzinga

Coming soon

Market news & events

Planned market-moving news, analyst, calendar, and event feeds for sentiment and event-driven validation.

  • News streamplanned
  • Analyst and ratings eventsplanned
  • Earnings calendarplanned
ORATS
Options

ORATS

Coming soon

Options analytics

Planned options chains, implied-volatility history, greeks, and surface features for options-aware validation.

  • Options chainsplanned
  • IV historyplanned
  • Greeks and surfacesplanned
OANDA
FX

OANDA

Coming soon

FX & rates

Planned foreign-exchange rates and currency reference data for macro, carry, and cross-asset overlays.

  • Spot FXplanned
  • Historical ratesplanned
  • Currency metadataplanned
CoinAPICoinAPI
Crypto

CoinAPI

Coming soon

Crypto market data

Planned exchange-normalized crypto trades, candles, reference, and multi-venue market data.

  • Tradesplanned
  • OHLCVplanned
  • Exchange referenceplanned
PolymarketKalshi
Validraft dataset

Prediction Markets

Coming soon

Polymarket & Kalshi

Validraft-built prediction-market dataset planned from market metadata, odds, liquidity, volume, and resolution history.

  • Market metadataplanned
  • Odds historyplanned
  • Resolution historyplanned

Need data we don’t hold yet?

Tell us what the hypothesis needs. If the feed is available and licensable, we source it, ingest it into the research lake, and document the data assumptions before validation starts.

See full data terms

Provider names and marks are shown for identification only. Validraft is independent and not affiliated with or endorsed by the listed providers. Coverage shown reflects the current research lake; some feeds refresh on demand for scoped engagements.

Privacy & confidentiality

Your edge is not training data, marketing copy, or a shared benchmark.

A trading hypothesis can be the most sensitive thing you send us. Validraft separates two ideas clearly: personal data is handled under the privacy notice, and strategy materials are treated as confidential engagement materials by default.

Read privacy noticeLegal boundary

Data surfaces we design around

  • Brief intake and feasibility notes
  • Submitted files, parameters, and assumptions
  • Generated reports, run metadata, and messages
  • Workspace access, support, billing, and security logs

Strategy materials stay scoped

Briefs, parameters, files, reports, and monitoring notes stay tied to the engagement they came from.

No public reuse of your logic

Client hypotheses are not added to a public template library, marketing example, or cross-client benchmark.

Private report delivery

Reports and workspace updates are delivered through authenticated access or agreed private channels.

NDA path for sensitive work

Standard briefs stay lightweight; sensitive business, Tier C, or institutional scopes can move under NDA before details are shared.

Operational practices

  • TLS in transit for website, brief intake, workspace access, and report delivery
  • Least-privilege internal access to engagement materials
  • Per-engagement separation for briefs, runs, reports, and monitoring context
  • No sale of client personal data or submitted strategy materials
  • Published privacy notice and optional analytics controls

Positioning

Between a self-service tool and a blank-check consulting project.

Validraft is built for the moment when a backtest looks promising, but you need an independent desk to test whether the evidence survives proper validation.

Self-serviceDone-for-youBespoke

Validraft position

Human judgment where it matters, standardized validation where it must be repeatable.

You operate everything

DIY backtest tools

Flexible if you already own the pipeline, but every data bug, look-ahead issue, and overfit parameter choice is yours to catch.

  • You build and maintain the validation stack
  • Robustness checks depend on what you remember to run
  • Fast iteration, but no independent read on the result

Independent validation desk

Validraft

Human-scoped validation, institutional gates, and a descriptive report without turning the engagement into open-ended consulting.

  • Feasibility review before compute starts
  • 66+ scoped controls across data, stats, stress, risk, and lineage
  • Confidential by default; no public reuse of your logic
  • Research-only language, never trade recommendations

Custom research project

Generic quant consultants

Useful for bespoke build-outs, but often heavier, slower, and harder to compare because methodology changes engagement by engagement.

  • Long scoping cycles before evidence appears
  • Opaque or inconsistent gate coverage
  • Can drift into advisory language if boundaries are unclear
DimensionDIY toolsValidraftConsultants

Best fit

Builders who want full control

Founders, traders, and managers who need an external validation read

Bespoke research or platform build-outs

Client effort

High

Brief-driven

High-touch

Validation gates

Whatever you implement

Standardized control stack

Variable by engagement

Delivery model

Self-service tool

Scoped report

Open-ended project

Research notes · Latest field notes

The machinery behind a defensible backtest.

Field notes on evidence quality: from point-in-time data and execution realism to statistical gates, immutable reporting, and post-validation monitoring.

View all notes

Your hypothesis. Independently validated.

Start with a sample, then ask whether your question is feasible. We agree on the analysis, data, price and delivery before any work starts.

Check feasibilityWorkspace