Explore broadly
Different market behaviours and relationships are examined without choosing an early winner.
METHODOLOGY
Built to be tested
Not just backtested
Curious Pro uses a structured research framework to develop, challenge and validate systematic models before they reach a dashboard
RESEARCH FRAMEWORKFrom broad research
to a validated model
Controlled processto a validated model
01RESEARCH
Explore a broad field
Different ideas and candidate paths are compared before a final model is selected
Multiple candidates
Evidence compared
Strong paths survive
02FREEZE
Lock the candidate
The selected model enters validation with its production rules fixed
Rules frozenNo further tuning
03VALIDATE
Challenge the result
The frozen model is tested for weaknesses before it can reach production
RobustnessDependencyReconciliation
BROAD BEFORE SELECTIVE
Start wide
Narrow only when the evidence allows it
The research process is designed to avoid choosing a favourite too early. Different ideas are first allowed to prove what they can and cannot contribute.
Map the evidence
Candidates are compared on more than return, including risk, activity, stability and robustness.
Select carefully
Only after the broad research is complete does the field become smaller.
Combine what survived
Strong independent building blocks can then be tested together.
FREEZE BEFORE VALIDATION
Validation starts when optimisation stops
Once a final candidate enters validation, its rules are frozen
The purpose of validation is to find weaknesses, not to improve the backtest
HOW THE MODEL IS CHALLENGED
Validation is designed to challenge the result
A strong historical result is only the starting point. The frozen model is then tested from several different directions.
Integrity
Is the timeline correct, is the data causal and can the historical record be trusted?
Reconciliation
Does the production implementation reproduce the researched model?
Robustness
Does the model still behave reasonably when nearby assumptions or execution conditions change?
Dependency
Is the result spread across time and trades, or dependent on a few exceptional events?
Time
Does the model remain viable across different starting points, rolling periods and market regimes?
THE INDEPENDENT ENGINE
An independent engine makes the research reproducible
Curious Pro uses an independent research and model engine so data, model logic and historical results can be reproduced and checked outside a single charting environment
- Controlled and confirmed data
- Independent model calculation
- Reconciliation with the production implementation
WHAT THE FRAMEWORK DELIBERATELY AVOIDS
Discipline also means
knowing what not to do
No early winner
Strong ideas are not discarded simply because another candidate initially looks better.
No single score decides
Return alone is never enough. Risk, stability, activity and robustness also matter.
No forced complexity
If an extra layer adds no credible value, it does not have to be added.
No tuning during validation
A failed final test is evidence, not an invitation to adjust the model.
EXPLORE FURTHER