Matrixport

Institution Name: Veta

Researcher/Contact Person: Li Luka

Backtest Time Range: June 9, 2025 to July 9, 2025

Strategy Name: Currency Rotation Long-Short Hedge Strategy Based on Predictive Yield Series

Strategy Type: Cross-Sectional Strategy

Indicator Name: CO-A-01-05

Definition: The number of times a single token is mentioned across major social platforms, including references by name, ticker, nickname, or related keywords.

Data Refresh Interval: 15 minutes / 1 hour

Access Method: API

1. Core Logic

Construct a long-short portfolio based on cross-sectional attention (mentions) ranking:

Long: Top 50% of tokens by attention every 1h

Short: Bottom 50% of tokens by attention every 1h

Equal-weight allocation | Long-short hedged strategy

2. Key Parameters

Data Frequency: 1-hour candlesticks (OHLC)

Rebalancing Frequency: 1 hour (HOLD PERIOD = 1)

Backtesting Period: 2025-06-09 to 2025-07-09

Coverage: 50 major tokens

3. Performance Metrics

1. Strategy Performance Dynamics

During the market volatility on June 15 and July 3, the strategy effectively captured tokens with sudden spikes in social attention (such as SOL and DOGE) through high-frequency rebalancing, with the long-short portfolio returns significantly outperforming the benchmark by 18% and 2.3%, respectively.

In extreme market conditions, the short positions in the bottom 50% of low-attention cryptocurrencies provided significant hedging effects, with the maximum single-day drawdown being only -0.39%.

2. Supplementary Indicators Supporting Strategy Logic

Signal Quality: 1-hour mention-frequency data significantly improved the strategy's responsiveness, but a few ineffective signals appeared during nighttime hours (UTC 0:00–4:00) due to reduced social activity.

3. Post-Implementation Insights

Usability: The API response time was stable (average <300ms).

Explanatory Power: The correlation between attention rankings and short-term price movements of cryptocurrencies reached 0.73 (Pearson coefficient), and the internal risk control model recognized its effectiveness as an attention-based sentiment factor.

4. Potential Improvement Suggestions

Community Expansion: Increase data from non-English communities (such as Korean or Turkish forums) to cover regional hotspots.

• Interested in continued usage and open to discussing formal licensing cooperation.

• Interested in customized indicators / focusing on specific communities (requires adding Telegram crypto groups and Chinese Weibo data).