TAM: ternary agreement model—a novel approach to lead-lag analysis in cryptocurrency markets

Cryptocurrency markets are characterized by extreme volatility, microstructure noise, and non-linear interdependencies that frequently compromise the reliability of lead–lag identification when using traditional magnitude-based linear estimators. This study introduces the Adapted Ternary Agreement Model (TAM), a noise-robust directional synchronization operator specifically engineered for high-frequency environments. The model discretizes 1-min logarithmic returns into ternary states {−1, 0, 1} by employing volatility-scaled adaptive thresholds ($\varepsilon_{i}$ = $\lambda\sigma_{i}$), effectively filtering out insignificant price fluctuations while preserving structural signals. Using high-frequency open, high, low, and close (OHLC) data from Binance, the framework identifies Solana-Tether (SOLUSDT) as a primary reference leader due to its dominant liquidity and ecosystem-wide correlation profiles. The model’s validity was rigorously tested through controlled ground-truth experiments with injected lags and benchmarked against a diverse suite of methodologies, including Granger Causality, Vector Autoregression (VAR) models, Wavelet Coherence, and an adapted Mean Absolute Directional Loss (MADL) metric. Empirical results demonstrate that TAM variants precisely recover lead–lag relationships, with the TAM-NZ (non-zero) variant achieving near-perfect agreement scores (0.9653–0.9889), significantly outperforming Pearson correlation and standard linear benchmarks. Furthermore, out-of-sample (OOS) validation underscores the framework’s real-time predictive utility, identifying robust cross-asset followers such as SOL–ADA with 91.88% directional agreement at a 4-min lag. By offering a stable, volatility-adaptive analytical pipeline, this study provides a significant contribution to high-frequency strategy design and market microstructure research, supported by a fully reproducible computational framework.