From Experiment to Trading Strategy

Evidence, Not Signals

SentiLab experiments produce statistical evidence, not trading signals. This is a critical distinction. A trading signal says "buy now." Statistical evidence says "sentiment and price are correlated at this lag with this strength." The gap between evidence and action is where your judgment, experience, and risk management come in.

Converting Evidence to Strategy: A Five-Step Process

Here is a structured approach for turning SentiLab findings into actionable intelligence:

Step 1: Identify the Signal

Start with a Correlation Sweep. Suppose it shows: sentiment leads BTC by +8 hours with r = 0.35, FDR-confirmed, 200+ data points, Strong confidence.

You now have a quantified relationship: moderate correlation, reliable, with a specific time window.

Step 2: Identify the Source

Run Source Predictability for the same asset. Results show CoinDesk leads by +8h with r = 0.52. Now you know which source to watch, not just that sentiment in general is correlated.

Step 3: Set a Trigger

Define a concrete condition: "When CoinDesk sentiment for BTC shifts above +0.3 (bullish), I will consider this a potential leading signal and evaluate a position within 8 hours."

Note the language: "consider" and "evaluate" — not "execute immediately." The trigger initiates analysis, not a trade.

Step 4: Validate with Context

Before acting on the trigger, validate with additional tools:

• Run Narrative Clustering to check if the sentiment shift is part of a broader media convergence pattern (which may signal a contrarian opportunity)
• Check the sentiment vs price chart to see if the price has already moved (signal already priced in?)
• Review the Fear & Greed index for market extremes that might override the signal

Step 5: Mental Backtest

Review the historical chart. Look at the past 90 days: when CoinDesk sentiment spiked positively, did BTC price follow within 8 hours? Check 5–10 instances. If the pattern holds visually, you have additional (informal) confirmation.

What SentiSignal Does NOT Do

It is essential to be clear about the boundaries:

• SentiSignal does NOT execute trades
• It does NOT provide buy/sell recommendations
• It does NOT guarantee outcomes
• It provides intelligence to inform YOUR decisions

The Correlation Reality Check

A correlation of r = 0.35 means sentiment explains approximately 12% of price variance (r² = 0.12). The other 88% is driven by factors SentiSignal does not capture: order flow, whale movements, macro events, regulatory news, technical chart patterns, and random market noise.

This is not a weakness — it is reality. No single factor explains most of price movement. The value of sentiment analysis is that it adds an informational edge that most traders lack, even if that edge is modest.

Always Combine with Other Analysis

Sentiment evidence should be one input among several:

• Price action analysis: Support/resistance levels, trend direction, momentum indicators
• Volume: Is the price movement supported by volume, or is it a low-liquidity move?
• Macro context: Interest rate decisions, regulatory announcements, geopolitical events
• Risk management: Position sizing, stop-loss levels, portfolio exposure

Start with Paper Trading

Before risking capital based on sentiment signals, paper trade (track hypothetical positions) for at least 30 days. Record each decision, the sentiment evidence that informed it, and the outcome. After 30 days, review: did the sentiment edge manifest in your hypothetical results?

This is your personal out-of-sample test. If the strategy does not work on paper, it will not work with real money.

Why This Matters

The gap between "interesting research finding" and "actionable trading strategy" is where most people fail. By following a structured process — identify, source, trigger, validate, backtest — you maintain analytical rigor while translating evidence into decisions. For a complete end-to-end workflow combining all SentiSignal tools, continue to Module 12: Building a Trading Strategy.