Bitcoin Factor Analysis: Investigation Report

Objective

I set out to determine which factors most reliably predict Bitcoin's forward returns, and specifically how to handle situations where different signals conflict.

Background

I'm currently observing conflicting signals:

• The Power Law model indicates Bitcoin is 35% undervalued
• The halving cycle indicates we're in Distribution phase (historically bearish)

This raised a practical question: when signals disagree, which should take precedence?

Factor Selection Rationale

I selected three factors, each representing a distinct hypothesis about what drives Bitcoin returns.

Power Law Model

Hypothesis

Bitcoin follows a long-term logarithmic growth trajectory, and deviations from this trend tend to mean-revert.

Rationale

Several researchers (notably Giovanni Santostasi) have observed that Bitcoin's price, when plotted on log-log axes against time, follows a remarkably linear path. This suggests a power law relationship: price grows as a function of time raised to some exponent.

The intuition is that Bitcoin's value derives from network effects (Metcalfe's Law), and network growth follows predictable adoption curves. If true, significant deviations from the trend represent mispricing—overvaluation eventually corrects down, undervaluation corrects up.

I fitted a power law model with a damped sinusoidal component to capture cyclical behavior. The model achieves R² = 0.96, indicating strong fit to historical data.

Halving Cycle

Hypothesis

Bitcoin's fixed supply schedule creates predictable market cycles tied to halving events.

Rationale

Every ~4 years, the Bitcoin block reward halves, cutting the rate of new supply in half. This is a fundamental, programmatic feature of Bitcoin that cannot be changed.

The halving creates a supply shock: miners receive fewer BTC for the same work, reducing sell pressure. Historically, this has preceded bull markets. The cycle appears to follow a pattern:

• Pre-Halving (18 months before): Anticipation builds, accumulation begins
• Post-Halving Bull (18 months after): Supply shock plays out, euphoria peaks
• Distribution (18-36 months after): Early buyers distribute to late buyers, prices decline

Unlike most market cycles, the halving is scheduled years in advance. This makes it a unique, exogenous timing signal.

Fed/Macro Indicators

Hypothesis

Bitcoin, as a risk asset, responds to broader liquidity conditions in the financial system.

Rationale

Since 2020, Bitcoin has increasingly correlated with traditional risk assets. The proposed mechanism:

• When the Fed expands its balance sheet (QE) or M2 money supply grows, excess liquidity flows into risk assets including crypto
• When the Fed contracts (QT) or raises rates, liquidity drains from risk assets

I examined several macro indicators:

• Net Liquidity: Fed balance sheet minus Treasury General Account minus Reverse Repo
• M2 YoY: Year-over-year change in money supply
• WALCL YoY: Year-over-year change in Fed balance sheet
• 10Y Yield changes: Interest rate environment

The question was whether these macro factors add predictive value beyond Bitcoin-specific factors.

Methodology

Data

• Bitcoin price: Daily data, January 2015 – December 2025 (3,971 observations)
• Economic data: FRED (M2, Fed balance sheet, yields)
• Target variable: 90-day forward returns

Approach

1. Measure individual factor correlations with forward returns
2. Examine return distributions within factor subgroups
3. Test factor interactions when signals conflict
4. Validate findings statistically

Findings

1. Individual Factor Correlations

I first measured how each factor correlates with 90-day forward returns across the full dataset.

Factor correlations with 90-day forward returns. Power Law Deviation shows the strongest relationship.

2. Subgroup Analysis

I then examined return distributions within each factor's subgroups.

Power Law Quintiles

Q1 vs Q5 spread: +42.0% (t=26.71, p<0.001)

Halving Phases

Bull vs Distribution spread: +53.7% (t=30.97, p<0.001)

Return distributions by halving phase. Distribution phase shows consistently negative returns.

3. Factor Interaction Analysis

This was the critical test. I isolated observations where Power Law indicated "undervalued" (deviation < -20%) and examined how returns varied by halving phase.

The same "undervalued" signal produces dramatically different outcomes depending on halving phase.

I ran the reverse test to check whether undervaluation provides any benefit during Distribution:

Undervaluation is the least negative outcome, but still produces losses. The phase effect dominates.

4. Statistical Validation

The interaction effect is statistically robust.

Interpretation

Resolving the Correlation Paradox

The initial finding that Power Law has higher correlation (-0.449 vs -0.289) appeared to contradict the interaction analysis. I resolved this as follows:

The -0.449 correlation is computed across all 3,971 observations. This average blends:

• ~2,200 days in Pre-Halving or Bull phases, where Power Law performs well
• ~1,700 days in Distribution phases, where Power Law fails

The aggregate correlation masks regime-dependent performance. Power Law's predictive power is conditional on phase.

Regime Framework

This leads to a hierarchical interpretation:

Halving Phase functions as a regime filter that determines whether valuation-based signals are actionable.

Application to Current Position

Current Readings (December 4, 2025)

Historical Precedent

I filtered for observations matching the current setup: undervalued + Distribution phase.

Distribution of Outcomes

Even at the 90th percentile, the outcome was approximately breakeven.

Current Power Law deviation of -35.5% indicates undervaluation, but this signal is unreliable during Distribution phase.

Assessment

Based on the evidence, the current outlook is negative despite the undervaluation reading. The Power Law signal is not reliable during Distribution phase. Historical precedent for this configuration shows a 5.7% win rate with mean returns of -16.6%.

Conditions That Would Alter This Assessment

1. Phase transition: Distribution ends approximately 18 months post-halving. The next Pre-Halving phase begins ~18 months before the 2028 halving (~November 2027).
2. Extreme undervaluation: Deviation below -50% has historically marked cycle bottoms even during Distribution, though sample size is limited.
3. Monetary policy shift: Significant Fed easing could override cycle dynamics. No indication of this currently.

Conclusions

1. Power Law Deviation has the highest single-factor correlation with forward returns when measured across all data.
2. However, this correlation is regime-dependent. Power Law performs well during Pre-Halving and Bull phases but fails during Distribution.
3. Halving Phase functions as a regime filter. It determines whether valuation signals should be trusted.
4. During Distribution phase, undervaluation does not predict positive returns. Historical win rate for undervalued + Distribution is under 6%.
5. For the current position (undervalued + Distribution + contracting liquidity), historical precedent suggests negative expected returns over 90 days.

Limitations

1. Limited cycle history: Analysis covers only 3 complete halving cycles
2. Retrospective phase boundaries: Phase definitions are known with certainty only in hindsight
3. Sample size constraints: Undervalued + Distribution subset has 130-159 observations
4. Regime stability assumption: Future cycles may behave differently
5. External shocks: Model does not account for black swan events

Appendix

A.1 Power Law Model Specification

log(P) = a + b*log(t) + A*e^(-decay*t/T)*sin(2π*t/T + φ)

Parameters:
- t = days since genesis (Jan 3, 2009)
- T = halving cycle length (1,387 days)
- Fitted values: a=-38.19, b=5.71, A=1.75, φ=-0.71, decay=0.40
- R² = 0.9605

A.2 Halving Phase Definitions

Halving Dates:
- 2012-11-28
- 2016-07-09
- 2020-05-11
- 2024-04-19

Phase Boundaries (days from halving):
- Pre-Halving: -547 to 0
- Post-Halving Bull: 0 to 547
- Distribution: 547 to 1095

A.3 Correlation Matrix

A.4 Additional Visualizations

Comparison of deviation calculations across different power law model specifications.

Power Law deviation overlaid with halving cycle phases, showing how the two factors interact over time.

Combined dashboard showing Power Law, Halving Phase, and macro indicators together.