Exploration is not validation
A fast research engine can reject weak ideas and compare broad parameter ranges, but it is not broker-terminal proof. Exploratory results are labelled as such and cannot become a customer claim by themselves.
JPTC Research Team
Trading strategy research and validation at JPTC
JPTC publishes selected methods and aggregate findings from strategy research. Exploratory simulations, broker-terminal tests and customer release decisions are labelled separately so one test cannot be mistaken for another.
Published by JPTC Research Team under the JPTC editorial policy.
Read the full methodologyBroker-terminal validation
A serious candidate is rerun in the relevant terminal with broker data, spread, commission, swap and order behaviour recorded. A result that changes materially at this stage is rejected or returned to research.
Out-of-sample and regime checks
Promising settings are checked outside the period used to select them, including difficult years and different volatility regimes. A strong recent window cannot compensate for structural failure elsewhere.
Trade frequency and stagnation
A result needs enough independent trades to be meaningful and enough activity to match the intended customer use. Very sparse files are reported as research candidates rather than presented as active customer defaults.
Failures remain part of the record
JPTC records why strategies fail, including trend exposure, spread sensitivity, over-filtering, recovery logic and weak out-of-sample behaviour. Failed ideas are not renamed and presented as new evidence.
Public evidence without exposing the edge
Research notes can publish test windows, assumptions, aggregate trades, rejection gates and limitations without disclosing proprietary entry code or customer setfiles.
Research and editorial standards
Simulator vs MT5 capability auditA public code-capability matrix explaining why fast screening and terminal validation can disagree.EA release rejection taxonomyA reusable classification of evidence failures that stop an EA or setfile from being released.Research methodologyTest levels, execution assumptions, release gates and reporting rules.Editorial policyHow JPTC assigns team ownership, checks sources and approves publication.Corrections policyHow material errors are reported, corrected and dated.EA HubCustomer software, current product scope and risk controls.
Common questions
Does a backtest prove that an EA will work?
No. A backtest is a research filter, not proof of future behaviour. Broker execution, unseen periods and current market conditions can change the result.
What does JPTC call a validated strategy?
A validated strategy has passed the documented broker-terminal and out-of-sample gates for its intended release. Exploratory simulations are never labelled as customer validation.
Does JPTC publish failed research?
Yes, when the failure teaches something useful without exposing proprietary strategy logic. Rejection evidence is often more useful than another selected winning curve.