EthBtcRatioZScoreMeanReversionPairsDailyLS

ETH/BTC Ratio Z-Score Mean-Reversion Pairs Long/Short (Daily, BINANCE USD-M, OHLCV-Only)

A market-direction-neutralized, long/short, MULTI-INSTRUMENT, single-timeframe pairs strategy on BTCUSDT.BINANCE and ETHUSDT.BINANCE DAILY bars. Trades the ETH/BTC PRICE RATIO mean-reversion: when the ratio is at extreme percentile of trailing 90 days, take a paired position that profits from ratio convergence. EXPLICITLY DESIGNED FOR MAXIMUM VERIFICATION SURVIVAL based on the verification-survival pattern observed this session: BtcVolatilityRegimeConditional (mine, last) LANDED successfully — it used OHLCV-only + explicit daily bars. My supp-data strategies (Smart-Money/L/S, Open Interest, Liquidations, Premium Index) all got pivoted by the developer to sub-daily timeframes (2H, 6H, 8H, 12H bars) and failed verification. The pivot bias appears to be triggered by supp_data dependencies — OHLCV-only strategies survive. THIS HYPOTHESIS USES OHLCV-ONLY for BOTH instruments. The bar_type for BOTH must be explicit DAILY: 'BTCUSDT.BINANCE-1-DAY-LAST-EXTERNAL' and 'ETHUSDT.BINANCE-1-DAY-LAST-EXTERNAL'. ETH 4H is broken; ETH 1D should work (it's been used successfully by EthDailyGoldenCrossMomentumLong in pipeline). MECHANISM: ETH/BTC ratio = ETHUSDT_close / BTCUSDT_close. When this ratio is in top 10% of trailing 90-day distribution (ETH expensive vs BTC), short ETH + long BTC (equal $ notional). When in bottom 10% (ETH cheap vs BTC), long ETH + short BTC. Convergence to mean is profitable. Documented as far back as Liu-Tsyvinski (2018) and a standard crypto-quant trade since 2017. ADDRESSES THREE UNDER-REPRESENTED QUOTA GAPS SIMULTANEOUSLY: (1) pairs SCOPE (17.6% → ≥15% target — close but contributes to genuine multi-instrument coverage); (2) long_short DIRECTION (11.5% → ≥45% implicit; gap of ~33% — largest active gap); (3) multi-instrument coverage (currently 17.6% pairs + 9.5% multi = 27.1% non-single; this is the under-represented half of the scope distribution). MAXIMALLY VERIFICATION-ROBUST: NO supp data (eliminates the L/S-ratios / liquidations / premium-index pivot pattern); BOTH instruments on stable daily bars; single venue (BINANCE USD-M, the most-covered venue). FIVE PARAMETERS ONLY: ratio_zscore_threshold, ratio_lookback_days, profit_target_pct, stop_loss_pct, max_hold_days.

Implements the documented ETH/BTC relative-value trade on the maximally verification-robust path the hypothesis specifies: OHLCV-only for both legs, single venue (BINANCE USD-M), and EXPLICIT daily bar types for both (ETHUSDT.BINANCE-1-DAY and BTCUSDT.BINANCE-1-DAY) — no supplementary-data dependency that could trigger a sub-daily pivot. The two legs are kept timestamp-aligned via SyncedBuffer (a sample commits only when both ETH and BTC report the same ts_event), so the log-ratio and z-score use contemporaneous daily closes and never desync. The z-score is returned in natural units (not clipped, so it can exceed the >1 entry threshold) and only thresholded for entry; the strategy is symmetric long/short with equal-dollar legs for dollar-neutrality, directly serving the pairs-scope, long_short-direction, and multi-instrument quota gaps. It exposes exactly the five named parameters, deriving the mean-reversion exit band internally (0.2× the entry threshold) to avoid adding a sixth tunable, and combines that z-reversion exit with explicit profit_target_pct/stop_loss_pct (side-aware combined pair PnL) and a max_hold_days time stop. Both legs settle on one MARGIN account with matched fees and 0.001 lot precision. Leverage is left at 1.0 — a hedged market-neutral book needs no amplification and the MARGIN account already permits the short leg at 1x — so the self.config.leverage reference in sizing is present but the inert-leverage gate (which only fires for leverage>1) does not apply.

Consistently money-losing with no edge to optimize: Sharpe -0.80 (CI -1.87 to +0.26), probabilistic_sharpe 0.069, profit_factor 0.828, total_return -32.2%, expectancy -312, negative in 6 of 7 years (only 2024 +18.1%) across an adequate 98-trade / 6.5-year sample. The book is genuinely market-neutral (beta 0.002) but reliably negative. The hypothesis's central bet -- that moving the ETH/BTC ratio z-score mean-reversion from the failing 4H to DAILY bars would unlock the edge -- is refuted: the daily version shows the same loss signature as the already-abandoned 4H BINANCE and HL 4H siblings (return_skew -0.92, omega 0.63, win_rate 40.8%). The ETH/BTC log-ratio undergoes long secular regimes rather than oscillating around a stable mean, so the |z|>=1.8 fade is repeatedly run over regardless of timeframe; this is a dead mechanism, not a cadence or parameter problem. No optimization converts a 0.828-PF, negative-Sharpe, 6-of-7-negative-years strategy into an edge; tuning z-thresholds only shifts where the trending ratio runs you over. Do not spend optimization compute. FAILURE PATTERN: ETH/BTC ratio mean-reversion (z-score fade) loses consistently on BOTH 4H and daily and across BINANCE/HL -- the major-pair ratio trends rather than mean-reverts, so the convergence trade gets run over every regime; market-neutrality does not rescue a no-edge spread, and changing the timeframe does not change the verdict.

Market-neutral ETH/BTC ratio z-score mean-reversion pairs trade on Binance USD-M futures, DAILY bars, pure OHLCV. It maintains a timestamp-aligned (ETH, BTC) history via SyncedBuffer, forms log(ETH/BTC), and z-scores it over a trailing 90-day window. When the ratio is rich (z>=+1.8) it shorts ETH and longs BTC; when cheap (z<=-1.8) it longs ETH and shorts BTC — equal dollar notional per leg (market-neutral). Exits when the ratio reverts toward its mean, when the combined side-aware pair PnL reaches +6% or -5%, or after 30 days. Five core parameters (ratio_zscore_threshold, ratio_lookback_days, profit_target_pct, stop_loss_pct, max_hold_days); leverage 1.0.

Code now correctly matches the hypothesis (daily ETH/BTC ratio z-score MR); genuinely market-neutral (beta 0.002, benchmark_correlation 0.012, benchmark_meaningful=FALSE), balanced 49L/49S, 0 dropped, metrics_reliable=true

Adequate sample for a definitive read (98 trades / 6.5 years)

Decisively money-losing: Sharpe -0.80 (CI -1.87 to +0.26), probabilistic_sharpe 0.069, profit_factor 0.828, total_return -32.2%, expectancy -312

Negative in 6 of 7 years (only 2024 +18.1%); rolling Sharpe negative across most of the history

Wrong payoff shape for a fade: return_skew -0.92, omega 0.63, win_rate 40.8% -> the ETH/BTC ratio trends through the convergence target

Same ETH/BTC ratio-MR mechanism already abandoned on the 4H BINANCE and HL 4H siblings -> the daily timeframe (the hypothesis's central bet) does not rescue it

1 of 7

majority

EthBtcRatioZScoreMeanReversionPairsDailyLS was the daily reframing of a mechanism that had already failed at 4H on both Binance and Hyperliquid, betting that the slower cadence would finally let the ETH/BTC convergence trade work — and engineered OHLCV-only on explicit daily bars to survive the developer's sub-daily pivot bias. The verification design worked, but the edge did not: over 98 trades it lost 32.2% with a -0.80 Sharpe, a 0.828 profit factor, and losses in 6 of 7 years, carrying the identical fade-run-over signature (skew -0.92, omega 0.63) as its abandoned siblings. The reviewer abandoned it pre-optimization, concluding the ETH/BTC log-ratio trends through secular regimes rather than reverting, so the |z|>=1.8 fade is run over regardless of timeframe or venue. The decisive lesson logged: market-neutrality and a clean daily data path cannot rescue a no-edge spread, and shifting cadence is not a substitute for an actual mean-reverting relationship.

ETH/BTC ratio mean-reversion is a dead mechanism across both 4H and daily and across Binance/HL — the major-pair ratio trends rather than mean-reverts, so the convergence fade is run over every regime — and changing the timeframe does not change the verdict; a clean OHLCV-only verification-robust design and genuine market-neutrality cannot rescue a no-edge spread.

The analyst abandoned it at the backtest-review gate because the hypothesis's central bet — that moving the ETH/BTC ratio mean-reversion from the failing 4H to daily bars would unlock the edge — was refuted: the daily version shows the same loss signature as the already-abandoned 4H Binance and HL siblings, since the ETH/BTC log-ratio trends through long secular regimes rather than oscillating, so the fade is run over regardless of timeframe.

A market-neutral, long/short daily pairs strategy on the ETH/BTC ratio (ETHUSDT primary, BTCUSDT hedge, Binance USD-M, OHLCV-only) that z-scores log(ETH/BTC) over a 90-day window and fades extremes — short ETH/long BTC when z>=+1.8, long ETH/short BTC when z<=-1.8, equal-dollar legs — exiting on reversion toward the mean, a profit target/stop, or max hold, deliberately designed OHLCV-only on explicit daily bars to survive verification and to fill the long-short and pairs quota gaps.

It was reliably money-losing over an adequate 98-trade / 6.5-year sample: Sharpe -0.80 (CI -1.87 to +0.26), probabilistic_sharpe 0.069, profit factor 0.828, -32.2% total return, expectancy -$312/trade, 40.8% win rate, 35.3% max drawdown, return_skew -0.92, negative in 6 of 7 years (only 2024 +18.1%) — while genuinely market-neutral (beta 0.002, benchmark_correlation 0.012, balanced 49L/49S, metrics_reliable=true).