Cryptoracle Data Analysis Team

This study aims to validate how exogenous shocks (E) propagate through social communities (G) to influence user behavior (U), ultimately driving price volatility of cryptocurrencies (C). Specifically, social platforms serve as the initial stage for event diffusion and act as amplifiers of user decision-making, particularly among KOL-led groups and speculative investors. User reactions—such as copy trading, panic selling, FOMO-induced buying, or FUD-driven selling—typically manifest first in social discourse. These behaviors are then translated into on-chain actions (e.g., accumulation or liquidation) and order book dynamics, eventually impacting token prices. This causal chain is especially relevant for high-frequency trading, event-driven strategies (e.g., exchange listings, hacks, regulatory shifts), and sentiment-driven market movements.

This forms a closed-loop feedback system comprising two interconnected causal chains:

1. Upward Propagation Chain (Information Diffusion → Trading Activity) (E → G → U → C)

Social media fermentation (topic/emotion) → Users form consensus (especially speculative types and KOLs) → Concentrated trading certain crypto → Leads to price fluctuations

2. Downward Feedback Chain (Market Volatility → Emotional Reaction) (C → G → U)

Ctypto surge/plunge → Feedback to the community (FOMO/FUD) → Emotion secondary fermentation → Users further follow (add position/cut losses)

1. Token Popularity Growth Rate（E→G）

1.1 Short-Term Event Surge Detection →  Use Logarithmic Growth Rate or Z-score

More smooth, more suitable for time series modeling

μ and σ represent the historical mean and standard deviation of token mention volume over a given time window. They are used to quantify the deviation of current popularity from its historical baseline, making this approach highly effective for event detection.

1.2 Smoothed Trend Analysis via EMA Rate of Change

Here, EMA refers to the Exponentially Weighted Moving Average (EWMA)：

α is the smoothing factor (typically between 0.2 and 0.5), which helps filter out random noise and capture underlying trend growth.

1.3 Routine Change Tracking → Use Relative Change Rate

Emphasize the proportion of changes relative to the current value, which better reflects the intensity of change for data with explosive increases.

2 ．Sentiment Momentum Index（G→U）

The number of active users, KOL posting frequency, and sentiment score reflect three distinct dimensions of community dynamics—volume, influence, and emotional orientation, respectively. Together, these variables illustrate how communities amplify the impact of external events and ultimately drive user behavior.

Active User Count → Reflects the overall level of user engagement within the community.

KOL Frequency → Indicates the mobilization power and influence intensity of key opinion leaders.

Sentiment Score → Measures whether collective sentiment is strong and directional enough to trigger user action.

Multiplication = the amplifying effect of combined actions.

→ Only when all three variables are at high levels does the combined effect become the strongest. If any one of them is 0 (or close to 0), the overall driving force approaches 0.

→ If all three values are high, it indicates that:

"Massive user engagement + clear KOL guidance + strong emotional consensus"

→ Translated into user behavior:

Increase position, copy trade, dump, chase the pump, stop loss

Positive values → Likely to form buy orders driven by FOMO

Negative values → Likely to form sell-offs driven by FUD

2.1 Community Sentiment score

Definition:The average sentiment score of all messages within a given time window, computed using a sentiment analysis model. The score ranges from -1 to 1, where -1 indicates strongly negative sentiment and 1 indicates strongly positive sentiment.

2.2   KOL Dominance

√  Definition: Measure whether the discussion in the community is dominated by KOLs.

2.3  Growth in active users within the community（Active User Growth）

√  Definition: Measure the trend of community activity changes

• Posti,t: The number of message posts related to an event (including posts, replies, comments) made by user i at time t

• |⋅| (Cardinality), indicating "the number of elements in the set," which is essentially the count after removing duplicates.

3.  Whale Net Position Change Rate（U→C）

Under the influence of community-driven sentiment (e.g., FOMO or FUD), users (G → U) may take action through on-chain transactions—such as increasing or reducing their positions—and order book behaviors, such as aggressive bidding or order cancellation. These actions directly affect market supply and demand, thereby contributing to changes in token volatility.

(Current Total Holdings of Whales−Previous Total Holdings of Whales)/ Previous Total Holdings of Whales

• Ht：Total Token Holdings of Top 5% (or 10%) Addresses at Current Time

• Ht-1：Total Holdings of Whale Addresses in the Previous Time Window

Interpretation:

Positive value → Whales are increasing their positions → Buying pressure is increasing

Negative value → Whales are reducing their positions → Selling pressure is increasing → Market volatility may rise

4.  Tokens price fluctuation

•   Realized Volatility

5. Baron & Kenny Mediation Effect Test

There are three types of variables:

• Independent Variable (X): Mention Growth Rate (mention_growth)

• Mediator Variable (M1/M2): Community Sentiment, On-Chain Behavior Changes

• Dependent variable (Y): Volatility

Mention volume growth (X) → Community sentiment (M1) → On-chain behavior change (M2) → Volatility (Y)

Step 1: Does X affect M (via a path)?

• Test： M1 = α0 + α1 * X + ε

• Result: α₁ is statistically significant (p<0.05)

Interpretation: An increase in mention volume must be capable of driving changes in community sentiment.

Step 2: Does M affect Y (via the b path)?

Test 1: M₂ = β₀ + β₁ · M₁ + ε

Test 2: Y = γ₀ + γ₁ · M₂ + ε

Result: β₁ is significant, γ₁ is significant (p < 0.05)

Interpretation:

• Community sentiment must be able to drive on-chain user behavior.

• On-chain behavior must influence token volatility.

Step 3: Does X directly affect Y (via the c path)?

Test: Y = θ₀ + θ₁ · X + ε

Result: θ₁ is significant (p < 0.05)

Interpretation:

Does mention volume directly influence volatility? This tests whether a direct effect exists from X to Y.

Step 4: Does the effect of X on Y weaken after adding mediators? (via the c′ path)

Test: Y = δ + δ₁ · X + δ₂ · M₁ + δ₃ · M₂+ ε

Observation:

• Is δ₁ < θ₁?

• Has the significance of δ₁ decreased?

Interpretation:

• If δ₁ is no longer significant, this suggests full mediation.

• If δ₁ remains significant but the coefficient shrinks, this suggests partial mediation.

Model Mathematical Formula Summary:

Total Effect: Y = c * X + ε (c = Total Effect)

Mediating Model:

M1 = a * X + ε1

M2 = b1 * M1 + ε2

Y = b2 * M2 + c' * X + ε3

- Total Effect Decomposition:

Total Effect = Direct Effect (c‘) + Indirect Effect (a* b1* b2)

Illustration:

Mention Volume Growth (X) - a - Community Sentiment (M1) - b1 - On-chain Behavior (M2) - b2 - Volatility (Y)

If c' decreases even slightly, it indicates that the chain is effective.

Summary: If paths a, b1, b2 are significant, and c' significantly decreases (or is not significant), then the link "mention volume → emotion → on-chain behavior → volatility" is effectively existing.