February 2026 Case Study & Performance

This page analyzes the documented backtest results for the feb_2026_strategy during February 2026. The study evaluates the performance of the AI-driven sentiment engine in a high-volatility bear market characterized by significant macroeconomic shifts and geopolitical escalation.

Case Study Overview

The February 2026 period served as a stress test for the news-sentiment-ai-trader. During this month, Bitcoin (BTC) experienced a net move of −16.4%, falling from a monthly high of ~$79,424 to a low of ~$60,000. The strategy utilized the feb_2026_frame to define the temporal window and the ccxt-exchange adapter for market data.

Performance Metrics

Metric	Value
Total Trades	16
Net PNL	+16.99%
Win Rate	68.8% (11 / 16)
Profit Factor	2.25
Directional Bias	14 SHORT / 2 LONG
Max Drawdown (Trade-to-Trade)	-5.98% (Late month)

Data Flow & Implementation

The backtest execution follows a structured pipeline where news data is retrieved, processed by the LLM, and converted into actionable trading signals.

Signal Generation Logic

The feb_2026_strategy processes news every 24 hours (the NEWS_WINDOW) to prevent over-trading. It maps LLM sentiment labels to position directions:

• Bearish/Bullish: Opens SHORT/LONG positions.
• Wait/Neutral: Defers entry.

System Data Flow: News to Trade

The following diagram illustrates how the system transitions from "Natural Language Space" (News/Reasoning) to "Code Entity Space" (Signals/Positions).

Diagram: News-to-Execution Pipeline

Key News Drivers & Market Context

The strategy's success was largely due to its ability to identify and maintain a SHORT bias during several high-impact fundamental shifts:

1. The SaaSpocalypse & AI Re-pricing: Amazon's $200B AI capex fears and Nasdaq's 4.5% drop on Feb 7.
2. Fed Policy Shocks: The nomination of Kevin Warsh and his subsequent hawkish pivot, signaling sustained high interest rates.
3. Institutional Outflows: Record Bitcoin ETF outflows totaling $4B over five weeks and MicroStrategy (MSTR) falling below cost basis.
4. Geopolitical Escalation: US-Iran tensions and the deployment of aircraft carriers on Feb 20, which spiked the VIX above 20.
5. Trade Policy: The announcement of Trump's 15% global tariffs on Feb 24.

Allocation Model

The performance calculation uses a non-compounding model.

• Fixed Allocation: $100 per trade.
• PNL Calculation: Additive (Cumulative PNL% is the sum of individual trade PNL percentages).

Trade Execution Analysis

The strategy utilized three primary exit mechanisms to manage risk and lock in profits.

Exit Type	Count	Description
Trailing Take-Profit	9	Triggered when price retraced from peak after hitting profit thresholds.
Stop-Loss	4	Hard exit at 3.0% - 3.4% to protect capital.
Sentiment Flip	3	Immediate closure when the LLM produced a forecast opposing the current position.

Notable Trades

• Trade #3 (Short): The most successful trade (+14.28%). Opened Feb 4 at $75,740 following news of institutional deleveraging and a hawkish Fed pivot; closed at $64,657.
• Trade #11 (Long): A successful "Sentiment Flip." The strategy correctly identified a recovery bounce on Feb 19 based on US industrial output data and the Meta-Nvidia deal.

Execution Architecture

The backtest is executed via the Backtest.run() engine within the backtest-kit framework.

Diagram: Backtest Execution Components

Running the Case Study

To replicate these results, the system must be executed with the specific strategy and frame configuration:

npm start -- --backtest --symbol BTCUSDT \
  --strategy feb_2026_strategy \
  --exchange ccxt-exchange \
  --frame feb_2026_frame \
  ./content/feb_2026.strategy/feb_2026.strategy.ts
Copy