Strategy Notes — Document Your Assumptions

The Value of Writing Things Down

Every good experiment starts with assumptions. Strategy Notes let you attach structured documentation to each saved experiment, capturing the why behind your choices — not just the what.

Each saved experiment can include a strategy_metadata field for your notes. This is freeform text, but structuring it consistently will make your research library far more useful over time.

What Good Strategy Notes Include

A well-documented experiment answers four questions:

1. Hypothesis

State what you expect to find before you see the results. This prevents post-hoc rationalization — the tendency to explain away unexpected results rather than learning from them.

Example: "I expect BTC sentiment to lead price by 6–12 hours based on the retail trading dynamics where social media buzz precedes order flow."

2. Parameter Rationale

Document why you chose each non-default setting. Future-you (or a collaborator) will want to know the reasoning.

Example: "Using minQuality 0.8 to exclude promotional content that adds noise. Extended lag range (±72h) because I believe BTC responds slower to institutional news than retail news."

3. Expected Outcome

Define what a "successful" result looks like in advance. This creates an objective benchmark for evaluating findings.

Example: "Looking for r > 0.3 at positive lags to confirm sentiment-price lead. If r < 0.15 at all lags, will conclude that this particular quality threshold is too aggressive and is filtering out signal along with noise."

4. Interpretation Plan

Decide ahead of time what you will do with different outcomes. This prevents paralysis when results are ambiguous.

Example: "If r < 0.1, will try extending lag range to ±72h for slower effects. If r > 0.3 but reliableBest = false, will extend time range to 180 days for more statistical power."

What Documentation Prevents

Without strategy notes, three problems emerge over time:

1. Duplicate experiments: You run the same analysis twice because you forgot you already tested it. Wasted time.
2. Forgotten rationale: You find a saved experiment with unusual parameters but cannot remember why you chose them. The result becomes uninterpretable.
3. Lost insights: You notice something interesting during analysis but do not record it. Weeks later, you vaguely remember a finding but cannot reproduce the conditions.

Treat Experiments Like a Trading Journal

Professional traders keep journals documenting every trade: the thesis, the entry, the exit, and the review. SentiLab strategy notes serve the same purpose for research trades.

The discipline of writing your assumptions before seeing results fundamentally changes how you analyze data:

• You become more honest about what you actually predicted vs. what happened
• You build a record of what works and what does not
• You can review past notes to refine your analytical approach over time
• You avoid the trap of "data mining" — running dozens of configurations until one looks good by chance

Quick Template for Strategy Notes

Copy and adapt this template for each experiment:

HYPOTHESIS: [What I expect to find and why]
PARAMETERS: [Why I chose these specific settings]
EXPECTED: [What a successful result looks like]
IF INCONCLUSIVE: [What I will try next]
NOTES: [Any observations during analysis]

Why This Matters

Data without context is just numbers. Strategy notes transform raw findings into documented research that compounds in value over time. When you publish experiments to the Strategy Marketplace, your notes help other researchers understand and build on your work.