Desktop research software · Pre-launch

MeasureCompareResolve

Research strategies.Test improvements.Trust the evidence.

Trading Research Workshop takes the trades a strategy actually executed, tests clearly defined changes against them under statistical control, and reports what the evidence supports — including when the answer is no.

Join early accessDemo recording in production
Any market you can supply
  • Forex
  • Crypto
  • Stocks
  • Futures
  • CFDs

The position

We don't promise profits.We report what survives.

Most trading products sell an outcome. Trading Research Workshop is built around a process that is permitted to disagree with its owner. Every research question it runs has three legitimate answers, and the software returns whichever one the data actually gives — not the one that would be easiest to market.

Where one measurement can land

  • SupportedThe evidence backs the change.
  • RejectedThe evidence contradicts it.
  • InconclusiveThere is not enough to say.

The method

How a research question is answered

One strategy, one bounded question, eight stages. Each stage below is marked with what the software does today — not with what it is intended to do eventually.
  1. 01

    Your strategy

    Available

    You supply what you actually traded: executed trades, the rules behind them, and matching market data.

    Generic importers, not a fixed broker or instrument.

  2. 02

    Understanding

    Available

    The Workshop reads the supplied source into a structured strategy blueprint, which you review and approve before anything runs.

    Nothing is researched against an unapproved blueprint.

  3. 03

    Research design

    Available

    The question is bounded before the search begins. The candidate universe is declared, frozen and hashed.

    A candidate that was not declared cannot be discovered.

  4. 04

    Discovery

    Available

    Candidates are measured against a paired baseline on the Discovery partition only, under permutation testing and false-discovery-rate control.

    Same trades, one changed decision — never a different sample.

  5. 05

    Freeze

    Available

    Finalists are sealed with a content hash before any validation data is touched. The search cannot be quietly widened afterwards.

    The frozen set becomes part of the evidence record.

  6. 06

    Validation

    Available

    Frozen finalists are re-tested on data the candidate search never observed, under the same locked protocol.

    Qualifying in Discovery earns a test, not a verdict.

  7. 07

    Robustness

    In development

    Cost stress, profit-concentration and parameter-neighbour checks run today. Deeper robustness analysis is still being built.

    Overfitting probability, deflated Sharpe and Monte Carlo are not implemented.

  8. 08

    Evidence verdict

    Available

    A client-facing research report states what was tested, what the evidence supports, what it does not, and what remains unknown.

    Limitations and recommended next research are required sections.

Capabilities

What the Workshop can and cannot do

Five research groups, 23 lines. Each one is marked with its real state in the software, including the ones that are not finished.
17
Available
2
In development
4
Planned

Strategy analysis

What the strategy actually did, measured on the trades it actually took — before any hypothesis is entertained.

  • MFE / MAE excursion analysisAvailable
  • Observed trade behaviour and filteringAvailable
  • Strategy intake and blueprint reviewAvailable
  • Research diagnostics and provenanceAvailable

Exit and management research

Would a clearly specified change to how a position is managed have improved the strategy? One bounded question at a time.

  • Fixed stop, target and time exitsAvailable
  • Break-even rulesAvailable
  • Trailing stopsAvailable
  • Partial exits and scale-outAvailable
  • Indicator-driven adaptive trailingIn development

Scientific validation

The machinery that separates a result worth acting on from a number that merely looks good.

  • Paired counterfactual comparisonAvailable
  • Cluster-aware permutation testingAvailable
  • False-discovery-rate controlAvailable
  • Discovery / Validation separationAvailable
  • Sealed final holdoutAvailable
  • Overfitting probability, deflated SharpePlanned

Execution evidence

What a price movement was worth, what it cost to capture, and how much of that the Workshop can actually prove.

  • Instrument and cost provenanceAvailable
  • Unknown cost blocks monetary claimsAvailable
  • Holdout censoring at the boundaryAvailable
  • Client-facing research reportAvailable
  • Cross-broker and cross-feed robustnessPlanned

Entry and signal research

The next research layer. Changing when a strategy enters requires reconstructing its signal path exactly, which is not yet proven.

  • Entry variant simulationIn development
  • Exact signal reconstructionPlanned
  • Session and filter researchPlanned

States are read from the Workshop's own capability records. Available means a working implementation exists and is reachable in the application today. Nothing on this page describes a capability that has only been designed.

Why trust it

Five constraints the software imposes on itself

There is no track record to show you and no customer to quote. What there is, is a set of rules the Workshop cannot be talked out of.

  1. No cherry-picking

    It cannot quietly drop what failed.

    Research history preserves what failed. A rejected candidate stays in the record next to the surviving one, and the report has a section for it.

  2. Sealed holdout

    It cannot look at the final partition early.

    The final partition is censored at its boundary. Candidate search cannot observe it, so a result there is a test rather than a rehearsal.

  3. Unknown is not zero

    It cannot treat an unproven cost as zero.

    A commission nobody proved is not free trading. Missing cost evidence blocks the monetary conclusion instead of quietly defaulting to zero.

  4. Discovery is not validation

    It cannot call a search result a finding.

    A candidate that qualifies during search has earned one thing: the right to be tested on data it has never seen. It is not a finding yet.

  5. Evidence before claims

    It cannot upgrade inconclusive into support.

    Support, rejection and inconclusive are all legitimate outcomes. The Workshop reports which one it reached, and never upgrades the third into the first.

  6. And the standing disclosure

    Trading Research Workshop is research software, not investment advice. Research can support a hypothesis, reject it, or remain inconclusive. Hypothetical and historical results do not guarantee future performance.

The software

A research instrument, not a dashboard

The Operations Deck is the Workshop's home screen. It shows live execution state and real persisted workspace state — and nothing else, ever.
Interface reproductionThe complete application window, rebuilt for the web from the Workshop's own design system and shown in its idle state — no project loaded, no run persisted, and no host telemetry, because a web page has no machine behind it. Not a screenshot, and not live. Two of the four capabilities the deck annotates, and five of its nine screens, are marked unavailable there because they are unavailable in the product.

The research report

Every report the Workshop produces carries the same sections. Limitations and Recommended Next Research are required, not optional.

  1. 01Report Identity
  2. 02Strategy Identity
  3. 03Source and Provenance
  4. 04Data Qualification
  5. 05Scientific Design
  6. 06Observed Baseline
  7. 07Research Performed
  8. 08Research History
  9. 09Findings
  10. 10Validation Status
  11. 11Robustness and Statistical Evidence
  12. 12Final Holdout Status
  13. 13Limitations
  14. 14Final Assessment
  15. 15Recommended Next Research
  16. 16Audit and Provenance Appendix

  • Local desktop application

    The Workshop runs on your machine. Your trades, your data and your research stay on it.

  • An immutable record

    Runs, candidate sets and reports are stored as durable artefacts with content hashes, not as a dashboard state you can quietly edit.

  • Provenance throughout

    Every result carries where its inputs came from and which assumptions were applied to reach it.

Early access

Be there when the evidence starts arriving

Trading Research Workshop is in active development. Early access is a short list of strategy owners who want to see it as it becomes real — development notes, capability releases, and first access when the software opens.

  • Product updates when there is something real to show.
  • No trading signals, no tips, no forwarding to anyone else.
  • Leave the list at any time.

Pre-launch listYour address is stored on the Early Access list and used only to contact you about Trading Research Workshop. Nothing is sold or shared — see the privacy notice.