BtcSpotVolatilityCompressionAccumulationLong

BTC Spot Volatility Compression Multi-Week Accumulation (Range-Bound Low-Volume Long-Only, BINANCE_SPOT Daily Bars, Macro Horizon)

Long-only macro-horizon accumulation strategy on BTCUSDT.BINANCE_SPOT that EXPLICITLY AVOIDS the now-saturated cross-venue basis arbitrage mechanism (analyst flagged: 'the portfolio already has the cross-venue basis class adequately represented... Adding a 4th instance compounds the same risk without adding breadth'). Instead, this hypothesis targets THREE different under-represented buckets: BINANCE_SPOT (8.0% vs 15% target), MACRO 1w+ horizon (0.5% — by FAR the most under-represented horizon bucket, only 6 experiments in the entire portfolio), and a fundamentally new mechanism class — volatility-compression accumulation. Core thesis from market microstructure / Wyckoff accumulation framework: when BTC enters a multi-week LOW-VOLATILITY + LOW-VOLUME consolidation, it signals patient accumulation by long-horizon holders quietly absorbing supply during retail apathy. These 'boredom phases' (compression) statistically precede expansion phases (Bollinger 1992; 'volatility breeds volatility' GARCH literature). The strategy enters on the COMPRESSION signature and holds through the eventual expansion phase. Distinct from BtcSpotDrawdownAccumulationLong (already in pipeline) — that strategy triggers on DRAWDOWN magnitude; this one triggers on RANGE CONTRACTION + VOLUME DRY-UP, which fires in different market regimes (it can trigger near highs or in mid-range, not just after sell-offs). Spot venue means no shorting required (CASH account compatible), no funding rate complications, no cross-venue execution realism issues (the failure mode that just sank 4 cross-venue basis strategies). Long-only is over-represented overall, but the SPECIFIC combination of BINANCE_SPOT + macro horizon + compression-mechanism is novel. Three parameters only (range_pct_threshold, volume_ratio_threshold, max_hold_weeks) — minimal overfitting surface after the recurring 'progressive optimization Sharpe explosion' failures.

Backtest-review iteration: the mechanism verified and backtested but produced only 19 trades in ~9 years, so the headline Sharpe 3.82 is a small-sample artifact (CI [-2.95, 10.85] straddles zero, edge concentrated in 2020). The analyst said macro horizon is legitimately sparse but the entry is over-choked, and gave explicit loosening directions, which I applied verbatim and minimally: range_pct_threshold 0.15→0.22 and volume_ratio_threshold 0.80→0.95 (both gates fire in more regimes), and max_hold_weeks 12→6 so a single open position no longer blocks re-entry for ~84 bars (now ~42). I also added the requested funnel instrumentation — counts of bars passing the compression gate, the volume gate, both, and both-but-blocked-by-an-open-position — logged on stop, so the next review can see exactly which constraint starves entries and whether the larger sample spreads across regimes. Nothing that already passed was disturbed: the continuous bar-varying proxy still keeps Layer 2 non-frozen (verified through the actual proxy: 435–436 unique signals across all six scenarios, no exceptions), imports stay clean, and the compression mechanism, exits, spot long-only sizing, and leverage 1.0 are unchanged. If the larger sample still shows the edge concentrated in 2020 and negative elsewhere, the analyst's guidance is to abandon at that point — but the strategy must first trade enough to make that judgment statistically valid.

Adequate sample now (75 trades after the iteration-2 densification), but the edge is marginal, beta-driven, and structurally flawed. Profit_factor is only 1.29 and, despite a 2.29 point Sharpe, sharpe_ci_low is -1.31 (CI straddles 0) -- the Sharpe is not distinguishable from zero and will fail deflated-Sharpe after 225-trial selection. Alpha is ~0 (+0.0034) with benchmark_correlation 0.56 and annualized_volatility 124.8%, so the returns are mostly long-BTC beta in compression windows at 50% equity, not excess edge. The max_drawdown of 44.3% (CI to 74%) comes from a structural flaw: 'compression precedes expansion' is DIRECTION-BLIND, but the strategy is long-only, so it eats every down-expansion (2018 -23%, 2022 -39%) -- the 'accumulation' framing provides no downside control. This is not fixable within the current setup: the obvious remedy (a directional/200-SMA trend filter so the expansion is more likely upward) would re-sparsify the strategy and undo the densification that was just needed to reach an adequate sample -- the mechanism is caught between sparse-with-a-filter and dense-but-direction-blind-with-44%-DD, a bind parameter optimization cannot resolve. Not iterate: the density iterate is already spent and the fix is structurally incompatible with keeping an adequate sample. FAILURE PATTERN: long-only volatility-compression accumulation on a single asset is direction-blind -- compression precedes expansion in EITHER direction, so a long-only book captures up-expansions but eats down-expansions (44% drawdown, -39% in 2022), leaving a PF ~1.3 / ~0-alpha / 0-straddling-Sharpe edge that is mostly beta and won't deflate; adding the directional filter that would fix it re-sparsifies the strategy below the validation floor.

Long-only BTCUSDT.BINANCE_SPOT daily volatility-compression accumulation (Wyckoff 'boredom phase' → expansion). It enters long when BOTH a range-contraction gate (20-bar high-low range / price <= 0.22) and a volume dry-up gate (20-bar avg volume <= 0.95x the 60-bar avg) fire, then holds through the expansion with a 30% take-profit, a 12% stop, or a 6-week (42-bar) macro time stop. Spot CASH account (no shorting, no funding), 50% of equity per trade, leverage 1.0. A funnel counter (compression / volume / both / both-blocked-by-position) is logged on stop.

Density iterate worked: 75 trades over 8.8 years is now an adequate sample; clean long-only spot routing, low kurtosis 2.87 (broad-based, not outlier-driven), positive in 6 of 9 years

Novel bucket coverage (BINANCE_SPOT + macro horizon + compression mechanism)

Marginal, statistically-indistinguishable edge: profit_factor 1.29, sharpe_ci_low -1.31 (CI straddles 0 despite the 2.29 point Sharpe) -> won't survive deflation

~Zero alpha (+0.0034), benchmark_correlation 0.56, annualized_volatility 124.8% -> mostly long-BTC beta at 50% equity, not excess edge

max_drawdown 44.3% (CI to 74%) from a structural flaw: compression is direction-blind but the book is long-only, so it eats down-expansions (2018 -23%, 2022 -39%)

Density iterate already spent (iteration 2); the obvious fix (a directional/trend filter) would re-sparsify and undo the densification

> 0 (meaningful)

~62 WF floor

marginal

Do NOT optimize — only 19 trades in ~9 years cannot support the 3-phase optimization (a 3-window walk-forward would have ~6 trades each), and the headline Sharpe 3.82 is a small-sample artifact (CI [-2.95, 10.85] straddles zero, annualized vol 130%, edge dominated by 2020 +52.5% with 2024/2025 negative). Macro horizon is legitimately sparse, but the entry is over-choked. Fixes to reach a meaningful sample (target 40+ trades): (1) loosen range_pct_threshold to ~0.20-0.25 and volume_ratio_threshold to ~0.9-1.0; (2) reduce max_hold_weeks so a single position doesn't block re-entry for ~84 bars; (3) instrument the funnel — count bars passing the compression gate, the volume gate, both, and those blocked by an open position — to find which constraint starves entries. Re-run and confirm dozens of trades spread across multiple regimes with a Sharpe CI clearing zero. If the larger sample still shows the edge concentrated in 2020 and negative elsewhere, abandon at that point.

BtcSpotVolatilityCompressionAccumulationLong deliberately steered away from the saturated cross-venue basis class to fill three genuinely under-represented buckets — BINANCE_SPOT, the macro 1w+ horizon, and a new compression mechanism — entering BTC spot on multi-week range contraction plus volume dry-up and holding through the expansion. The iteration-2 loosening successfully lifted the sample to 75 trades and the headline looked strong (+91.6%, Sharpe 2.29), but the substance was thin: the confidence interval straddled zero, alpha was ~0 with 0.56 benchmark correlation, and a 44% drawdown exposed the core flaw — a direction-blind compression signal on a long-only book absorbs every downward expansion. The reviewer abandoned it pre-optimization as caught in an unresolvable bind: sparse if filtered for direction, dense but direction-blind with a 44% drawdown if not. The lesson logged: filling novel quota buckets does not substitute for a directionally-controlled edge, and 'accumulation' framing provides no downside protection when expansion can break either way.

Long-only volatility-compression accumulation on a single asset is direction-blind — compression precedes expansion in either direction, so the book captures up-expansions but eats down-expansions for a 44% drawdown, leaving a ~1.3-PF, ~0-alpha, 0-straddling-Sharpe edge that is mostly beta and won't deflate; adding the directional filter that would fix it re-sparsifies below the walk-forward floor, so novel-bucket coverage cannot substitute for a directionally-controlled edge.

The analyst abandoned it at the backtest-review gate because the edge is statistically indistinguishable from zero (would fail deflated-Sharpe), mostly long-BTC beta rather than alpha, and structurally flawed: compression is direction-blind but the book is long-only, so it eats every down-expansion (2018 -23%, 2022 -39%) — and the fix (a directional trend filter) would re-sparsify the strategy below the validation floor, undoing the densification just needed to reach an adequate sample.

A long-only, OHLCV-only macro-horizon accumulation strategy on BTCUSDT.BINANCE_SPOT that enters on a Wyckoff-style volatility-compression signature — multi-week range contraction (range <= threshold) AND volume dry-up (short/long volume ratio <= threshold) — and holds through the eventual expansion via a take-profit, stop, or multi-week time stop, deliberately targeting three under-represented buckets (spot venue, macro 1w+ horizon, compression mechanism) with only 3 core parameters.

After an iteration-2 densification it reached an adequate 75 trades over 8.8 years with a strong-looking headline — +91.6% total return, point Sharpe 2.29, profit factor 1.29, positive in 6 of 9 years, low kurtosis 2.87 — but the edge was marginal and beta-driven: sharpe_ci_low -1.31 (CI straddles 0), alpha ~+0.0034, benchmark_correlation 0.56, annualized volatility 124.8%, and a 44.3% max drawdown (CI to 74%).