▮ THE TEST CELLS · PER-ASSET CHARACTER · LIVE READS

The dyno bays

Six machines on six stands. The pattern gauge reads return-autocorrelation — under 0.05 the tape is a random walk: you don't predict it, you follow flow. Verdicts are stamped by the edge lab, not by marketing.

BAY-01

MNTUSDT

Mantle

pattern gauge · |return acf|

0.031 · measured — efficient

.efficientweakpatterned → predict

funding carry ▸ +3.2%/yr live · thin (below risk-free)

EDGE RETIRED

The one edge that ever passed the dyno (OI-contrarian, +28.9% OOS) — until fresh full-history data showed the pattern gone. The lab pulled its license. Honest firms retire edges.

BAY-02

BTCUSDT

Bitcoin

pattern gauge · |return acf|

<0.05 · efficient

.efficientweakpatterned → predict

funding carry ▸ +0.6%/yr live · thin

FEE-EATEN

Seven research rounds, including an ML disproof. Every directional signal that survived the math died at the fee line. The most efficient tape we tested — we say so instead of forcing trades.

BAY-03

ETHUSDT

Ethereum

pattern gauge · |return acf|

<0.05 · efficient

.efficientweakpatterned → predict

funding carry ▸ +0.8%/yr live · thin

NO EDGE

Momentum loses out-of-sample. Carry below risk-free. The bay stays instrumented; the engine stays off.

BAY-04

HYPEUSDT

Hyperliquid

pattern gauge · |return acf|

<0.05 · efficient

.efficientweakpatterned → predict

funding carry ▸ +1.9%/yr live · backtest best: +7.2%/yr crash-robust

PROBATION

Best delta-neutral carry in the backtest (crash-robust). A flow signal also flashed +92% — one lucky fold drove it, so it sits in candidate probation, not the registry. Forward-confirm or scrap.

BAY-05

SUIUSDT

Sui

pattern gauge · |return acf|

<0.05 · efficient

.efficientweakpatterned → predict

funding carry ▸ +5.7%/yr live · backtest +5.1%/yr but lumpy

LUMPY CARRY

Carry beats risk-free on average but arrives in lumps with crash risk. Tracked, not trusted.

BAY-06

mETH

Mantle Staked ETH

pattern gauge · |return acf|

n/a · no perp tape

.efficientweakpatterned → predict

funding carry ▸ no perp venue — carry not executable

NO VENUE

The +96% convergence backtest lived here — and died on thin DEX liquidity (−73 bps a trade). See the crash test on the landing page. No executable venue, no bay time.

BAY-07

SOLUSDT

Solana

pattern gauge · |return acf|

efficient · momentum loses OOS

.efficientweakpatterned → predict

funding carry ▸ −0.6%/yr live · no harvestable carry

NO EDGE

Directional momentum loses out-of-sample and funding offers no carry. The bay stays instrumented; the engine stays off until a signal clears the cost line. Tracked on Bitget, traded by no one here without an edge.

BAY-08

XRPUSDT

XRP

pattern gauge · |return acf|

new · profiling — just listed

.efficientweakpatterned → predict

funding carry ▸ +1.4%/yr live · thin (below risk-free)

ENROLLED

The newest bay — one of three assets that fill on the REAL Bitget testnet (with BTC + ETH). It runs the exact same gate as everything else: no validated edge yet, profiled live before any conviction. New tape earns trust the hard way.

registry today: EMPTY — every bay idle, the firm holds N. exploration keeps hunting (paper budget, fresh data, FDR-gated) — the first hypothesis that survives the dyno gets the next license.