The Problem This Episode Solves

​

Most trading strategies do not fail because the original idea was flawed.

 

They fail because traders continuously modify the model while it is still being tested.

 

A new observation is made.

A rule is added.

Another observation appears.

Another rule is added.

 

This process continues until the original hypothesis becomes contaminated by a collection of untested ideas.

 

At that point, the model being tested is no longer the model that was originally defined.

 

This is the core issue.

 

When new rules are introduced without independent validation, you are not strengthening the model - you are weakening the validity of the data that supported it.

 

Good research does not add ideas to a model.

It tests them against it.

​

Where We Are in the Process

 

Within the quant-inspired development pipeline, we return to the hypothesis stage.

 

Although we previously moved into validation, this step requires a temporary regression.

 

The reason is simple:

 

New observations made during data collection must be tested independently before they are incorporated into the model.

 

Within the series-specific framework, we are now entering the stage of conditional dependence — conditioning our existing data based on newly observed behaviour.

 

At this point, you should already have:

 

• A defined behavioural model

• A structured dataset collected impartially

• Probability distributions for your original hypothesis

​

The purpose of this episode is not to rebuild the model.

 

It is to refine it without corrupting it.

​

The Objective of This Stage

 

During data collection, patterns begin to emerge.

 

You may notice that certain environments appear to influence outcomes.

 

For example, your anchor event may behave differently when it occurs within a specific context.

 

However, not every observation you make is meaningful.

 

Most are noise.

Some are variance.

Only a small number represent real behavioural influence.

 

The objective of this stage is to:

 

• Identify meaningful observations

• Determine which are worth testing

• Convert them into structured sub-hypotheses

• Preserve the integrity of the original model

 

The Retail Mistake

 

When traders notice something that appears to influence outcomes, they tend to act immediately.

 

They convert the observation directly into a rule.

 

This is where the model breaks.

 

By introducing a rule without testing it independently:

 

• The dataset becomes contaminated

• The hypothesis changes mid-test

• The original results become invalid

​

Instead of improving the model, this process destroys its reliability.

 

The Correct Approach — Sub-Hypotheses

​

Rather than modifying the original model, a new structure is introduced:

 

The sub-hypothesis.

 

A sub-hypothesis is a conditional version of the original hypothesis. It isolates a single new observation and tests it independently.

 

The process is simple:

 

1. Make an observation

2. Convert it into a sub-hypothesis

3. Test it separately

4. Validate or invalidate it

5. Only then consider including it in the model

 

This ensures that every new idea is tested on its own merit.

 

What This Is Not

​

This process is not used to introduce new features, variables, or market regimes.

 

Those changes alter the structure of the model itself and require a full rebuild.

 

If you want to:

 

• Add new features → return to Episode 02

• Redefine market regimes → return to Episode 03

• Introduce new variables → return to Episode 05

 

Sub-hypotheses do not rebuild the model.

 

They refine behaviour within it.

 

Identifying Which Observations Matter

​

During analysis, you may notice many potential influences on outcome distribution.

 

Most of them should be ignored.

 

Testing every observation leads to:

 

• Dataset fragmentation

• Reduced sample size

• Unnecessary complexity

​

To determine whether an observation is worth testing, it must satisfy four conditions:

 

First, it must appear repeatedly.

If an observation only occurs once or twice, it is likely random.

 

Second, it must plausibly change behaviour.

There should be a logical reason why the observation could influence outcomes.

 

Third, it must fit within the existing model.

It must relate directly to the defined anchor event.

 

Fourth, it must be objectively definable.

If it cannot be measured consistently, it cannot be tested.

 

Only observations that meet these criteria should be considered.

​

Constructing a Sub-Hypothesis

​

A sub-hypothesis follows the exact same structure as the original hypothesis.

 

It includes:

 

• The same anchor event

• The same variables and conditions

• The same binary outcome

​

The only addition is the new observation.

 

This is critical.

 

By keeping everything else identical, you can directly compare:

 

• Original hypothesis results

• Sub-hypothesis results

​

This allows you to isolate the impact of the new condition.

​

Example — Weak Displacement Within Confluence

​

In this model, the observation was:

 

When weak displacement occurs within an existing confluence, outcomes appear to change.

 

For example:

 

Price undergoes weak displacement through liquidity, but the move is initiated inside a bullish structure.

 

The hypothesis becomes:

 

If weak displacement occurs within an existing confluence, does the probability of a valid outcome change?

 

Before testing, this observation was evaluated against the criteria:

 

• It appeared frequently

• It had logical reasoning

• It aligned with the anchor event

• It could be defined objectively

​

Only after passing all four conditions was it tested.

 

Maintaining Structural Consistency

​

When constructing the sub-hypothesis, the wording remains consistent.

 

Even if you believe the outcome may be less likely, the hypothesis should still be framed identically.

 

This prevents confusion during data collection and ensures consistency across datasets.

 

The simplest way to achieve this is to copy the original hypothesis and insert the new condition.

 

Why the Original Model Must Not Be Changed

​

Modifying the original model prematurely introduces two major risks:

 

First, it fragments the dataset.

 

Second, it introduces rules that may have no real significance.

 

By testing sub-hypotheses separately, you create independent datasets that can be compared directly.

 

This allows you to determine whether the new observation truly influences behaviour.

​

The Challenge of Overlapping Data

​

Introducing sub-hypotheses creates overlapping datasets.

 

Some examples will belong to:

 

• The original hypothesis

• The sub-hypothesis

• Both simultaneously

​

This creates dependency within the data.

 

Maintaining consistency across these overlapping datasets becomes the next challenge.

 

Closing Thoughts

​

At this stage, the focus is not on expanding the model.

 

It is on protecting it.

 

By testing observations independently, you ensure that every rule within the model is:

 

• Measurable

• Validated

• Statistically justified

​

This is what separates structured research from discretionary decision-making.

 

What Comes Next

 

In the next episode, we will address the complexity introduced by sub-hypotheses.

 

Specifically:

 

• How to validate sub-hypotheses

• How to collect data under identical conditions

• How to manage dependency across datasets

​

This is where conditional modelling becomes fully structured.