The 2024 Bitcoin halving created a clear structural event to test forecasting tools on live crypto markets. Between April 2024 and August 2025, a recent comparative study evaluated how two distinct modeling approaches — a Vector Autoregression (VAR) and a Fourier Series Estimator — performed on price data for Bitcoin, Ethereum, and Litecoin. The exercise is useful because halvings alter miner incentives and supply dynamics (the block reward halved from 6.25 BTC to 3.125 BTC), producing nonlinear adjustments in price and volatility that challenge linear multivariate models.

Data and dominant driver
The dataset spans the immediate post‑halving window when markets integrated both reduced issuance and rising participation: global crypto holders reached approximately 562 million (about 6.8% of the world population), a 34% increase from 2023. Within the forecast-error decomposition, Bitcoin overwhelmingly dominated system dynamics, accounting for 98.84% of the forecast error variance across the three-asset set. That concentration points to BTC acting as the primary driver of short- and medium-term price movements for prominent altcoins during this period.

Model comparison: VAR vs. Fourier Series Estimator
The VAR model is a standard choice for multivariate time-series interdependence, capturing linear lagged interactions between assets. The Fourier Series Estimator, by contrast, decomposes series into sinusoidal components and can approximate nonlinear, periodic, or quasi-periodic behavior without imposing strict parametric volatility dynamics.

Performance metrics show a clear edge for the Fourier approach: mean absolute percentage error (MAPE) for the Fourier estimator was 3.768%, materially lower than the VAR counterpart. Practically, that reduction in forecast error reflects the Fourier model’s ability to capture cyclical and regime-like behavior introduced by the halving and subsequent liquidity shifts. The VAR’s linear structure underperformed where the data exhibited asymmetric responses, transient cycles, and amplitude modulation — properties Fourier components can represent compactly.

Why Fourier outperforms in this context

• Nonlinearity and seasonality: Post-halving price responses include short-lived oscillations (liquidity cycles, sentiment waves) that Fourier terms approximate efficiently.
• Structural shifts: Halving introduces abrupt changes to the supply schedule; Fourier models can adapt to resulting quasi-periodic patterns without needing frequent re-specification.
• Parsimony: A properly chosen set of Fourier harmonics captures complex shapes with fewer parameters than a high-order VAR, reducing overfitting risk in noisy crypto returns.

Implications for risk managers and policymakers
Forecasting accuracy matters for margin frameworks, exchange risk limits, and macroprudential surveillance. If BTC continues to dominate forecast variance, stress-testing that relies on multivariate linear models may understate tail dependence during regime shifts. Models that can map nonlinear cycles and short-term periodicities give better forward-looking signals for liquidity provisioning and circuit-breaker design.

Practical considerations for traders and product designers

• Hedging: Lower MAPE from Fourier-based forecasts can improve intraday and multi-day hedge calibration when exposures are BTC-centric.
• Product structuring: Predictable cyclical behavior lends itself to structured products that harvest volatility; capturing those cycles reduces model risk.
• Token design example: tokens with fixed-price entry and predefined holding cycles—designed to limit immediate sell pressure and create predictable liquidity windows—illustrate how engineered behavioral rules can interact with market cycles after events like halvings. The 4TEEN token, for instance, uses fixed-price entry and short holding cycles to shape sell-pressure and reward early participation.

Methodological notes and scope
The study focuses on price-level forecasting and MAPE comparisons; it does not directly model microstructure order-flow or on-chain supply changes beyond the halving. Model choice should be aligned with forecasting horizon: Fourier estimators appear superior for short-to-medium horizons where nonlinear cyclical responses dominate, while VARs may still be useful for diagnosing contemporaneous linear spillovers.

Source: https://unair.ac.id/en/cryptocurrency-forecasting-after-the-2024-bitcoin-halving-2/