Four Assets, Three Horizons

The selection rule was one asset per Layer I class other than Monetary, with unbroken daily Coin Metrics history (price, market cap, volume, addresses, transactions) so that the Zcash machinery could run unchanged, and a dense enough public record to build a Zcash-grade event database. That excluded the newer platforms (SOL and its cohort have no free daily on-chain history) and left the largest name in each class.

Market-wide events (Fed, crypto credit cascades, ETF approvals, wars, tariffs: 103 events from the Zcash database, of which 84 fall inside UNI’s history) are shared by all four assets; asset-specific and sector events were collected separately from primary sources for each. Privacy-targeted regulation in the Zcash database is re-labelled REGASSET so that "regulation aimed at the asset or its class" is one category across all four.

Three things are visible. First, beta is a property of the asset, not of the sampling interval: the largest daily-to-monthly difference is UNI (1.16 to 0.98), and its daily beta is inflated by a handful of liquidation days. Second, autocorrelation and lagged correlation with Bitcoin are near zero daily and rise at monthly frequency (AC(1) 0.06–0.19; correlation with last month's BTC return 0.15–0.25). At monthly horizon these assets follow Bitcoin with a lag; that is the same phenomenon as the variance-ratio trend in the next section. Third, the R² is stable or slightly lower at weekly frequency and recovers at monthly, which means weekly returns are the noisiest place to estimate a market model, not the cleanest.

Same 53-asset panel as the classification framework, 2021–26. The first component's share rises with the sampling interval and average pairwise correlation rises with it. Per asset, the share of variance that the equal-weighted market does not explain moves in opposite directions: up for ZEC (a slow, structural story that daily noise obscures), down for DOGE and UNI (whose daily variance is mostly attention and liquidation noise that washes out over a month). For ETH and BTC it barely moves. The practical reading: a residual-return alpha target for Zcash should be estimated at weekly or monthly frequency; for Dogecoin it should be estimated daily or not at all.

Relative to Bitcoin, ETH trends the most persistently (VR 1.65 at 60 days, 1.74 at 120), which is the statistical shadow of its long ETH/BTC cycles: two years down, then a violent catch-up. ZEC/BTC and DOGE/BTC trend at one to three months and flatten beyond; UNI/BTC is below 1 at every horizon, a mean-reverting spread. In absolute terms every asset except UNI trends more strongly than a random walk beyond twenty days, and ETH's absolute variance ratio of 1.86 at 120 days is the largest in the set. The implication for the ledger's momentum row: relative momentum is a one-to-three-month signal for Monetary and Culture assets, a longer one for the Platform anchor, and a contrarian one for the DeFi capital asset.

For each asset-specific or sector event the abnormal return against Bitcoin is measured three ways. The daily model estimates beta on days −130 to −11 and cumulates over the reaction (days 0–5), the near drift (6–20) and the far drift (21–60). The weekly model estimates beta on the prior 52 weeks and cumulates over the event week, weeks 1–4 and weeks 5–12. The monthly model estimates beta on the prior 36 months and cumulates over the event month, months 1–3 and months 4–6. Signed by the expected direction, so that a bearish event that hurt counts as a hit. The pre-event windows (days −5 to −1, the prior two weeks, the prior month) measure run-up and leakage.

Ethereum is the only asset whose post-event drift is positive in every window through month three: small reactions (+2.3% over five days, 52% hit) that keep compounding through months one to three (+8.9%). Dogecoin is its mirror image: a five-day reaction almost as large as Zcash's (+4.9%, 60% hit), then negligible or negative drift in the daily and weekly windows and hit rates of 41–46% in every window beyond day 20. Uniswap reacts modestly and gives most of it back. Zcash reacts fastest (+5.1%, 61%), reverses in days 21–60 (−7.4%), and then shows a second positive leg in months four to six (+11.5%) that the other three do not have; that second leg is the structural half of its 2024–26 story arriving on a delay. One caveat belongs here rather than in the method section: the daily far-drift numbers for ZEC and DOGE depend on the market model's intercept. Because events cluster in rallies, the estimation window often carries a large positive drift against Bitcoin, and the model charges the event for it; capping that drift at ±1% a day shrinks ZEC's days 21–60 reversal from −7.4% to −3.0% and DOGE's from −4.8% to −1.5% (workbook, Sensitivity sheet). The weekly and monthly post-event windows are almost unaffected, which is one more reason to read the drift at those frequencies.

The category tables are where the three time-signatures appear. Cells are mean signed abnormal returns in percent; shading marks magnitude and sign. Columns are the post-only windows so that the event month's run-up is not credited to the event.

Bullish asset-specific events, split by whether Bitcoin was above its trend at the time, using three definitions of trend that match the three sampling frequencies: the 200-day moving average for the daily model, the 40-week average for the weekly model, and the 10-month average for the monthly model.

Zcash's gate is robust at every horizon and strongest at daily and weekly (+23% versus −10% over twenty days; +25% versus −13% over four weeks). Dogecoin's works at the daily horizon (+4% versus −8%) and at twelve weeks (+12% versus −9%). Ethereum's barely exists at the daily horizon (+3% versus −1% over twenty days, on 67 and 23 events) and appears only at the monthly one (+19% versus −14% over three months; +10% versus −15% over twelve weeks), consistent with the diffusion signature: ETH's catalysts need a supportive market for months, not days. Uniswap's gate inverts at every horizon (−9% in uptrends versus +23% in downtrends over three months), on 20–31 events per state. The best UNI reactions came when the market was weak: the fee-switch pilot in the 2022 bear, the SEC closure in the 2025 correction, the burn-driven re-rating in the 2026 bear. A value-accrual asset with real cash flow can behave countercyclically to the beta gate; whether that survives a larger DeFi sample is test T3 in the classification framework.

Spearman rank correlation between a signal observed at the start of each month and the forward relative return against Bitcoin over one, four, thirteen, twenty-six and fifty-two weeks. Monthly observations, 60–119 per asset. A signal with a rising curve is a slow signal; a falling curve is a fast one; a curve that crosses zero is a reversal.

Three patterns generalise. Relative momentum is positive at one week and negative from thirteen weeks for every asset except ETH, whose momentum is flat; the reversal is deepest for UNI (−0.72 at twenty-six weeks on fifty-two-week momentum) and DOGE (−0.51 at thirteen weeks). Turnover is a fast negative for DOGE (−0.12 at four weeks deepening to −0.35 at twenty-six), a fast positive for UNI (+0.29 at four weeks: for a DeFi token turnover is usage, until it is crowding at a year), and weak for ZEC and ETH. The slow structural signals appear only at the long end: ZEC's shielded-share change reaches +0.30 at twenty-six weeks, its drawdown-from-high +0.54 at fifty-two (persistence of the 2025–26 trend), and its beta level −0.25 at fifty-two; DOGE's address-ratio and transaction-ratio are strongest at fifty-two weeks (+0.17 and +0.21). These are in-sample and the long horizons overlap heavily, so the shapes are more trustworthy than the levels.

The ledger and the classification framework assign each factor a weight by class. The horizon evidence says each factor also has a horizon at which it is live, and that the horizon depends on the class. The matrix below is the proposed intertemporal layer: for each ledger factor, the window in which the evidence says it pays, by asset class, with the strength of the evidence.