CRUCIBLE · LANDSCAPE

$1.47 from $1 over ~5y (~7%/yr gross, max drawdown -7.0%) — a smooth-looking curve. But the honest autocorrelation-adjusted Sharpe is only 0.66, and the gate says DIES: sticky carry gives too few effective samples to confirm it beats luck — and this omits the basis/liquidation/exchange tail. Funding is negative right now, so the trade is off.

The carry is broad but uneven — some coins grind up, others bleed. Diversifying across them is what the basket does.

Rode BTC's bull run and ended slightly above buy & hold — looks like skill. But it's the BEST of 36 searched configs; once the gate deflates for those 36 tries the edge is within luck (DSR 0.9142). Not a distinguishable edge — the overfit control working on real data.

Both bleed; longshots bleed hard (favorite-longshot bias). The liquid prediction market is ~efficient — no free lunch.

Real, large premium (implied 61.2% vs realized 38.3%) — the first to clear the Sharpe gates, but DIES on the new tail gate (skew -4.34, kurt 30.9): one crash erases months. Real premium, uninvestable naked.

Long winners / short losers (14d). SURVIVES all four gates (honest Sharpe 0.98, OOS 0.395) — but transaction costs are NOT modeled, and daily rebalancing has brutal turnover. Fragile to fees.

Time BTC by stablecoin-supply growth. SURVIVES + beats buy & hold OOS (Sharpe 0.595 vs 0.91) — but the raw signal is near-zero (corr 0.033); mostly long beta with a -57% drawdown. Marginal.