EthSpotWeeklyDcaDrawdownBoostAccumulation

ETH Spot Weekly DCA with Drawdown-Boost Accumulation

A LONG-ONLY DOLLAR-COST-AVERAGING strategy on ETHUSDT spot with DRAWDOWN-CONTINGENT POSITION SIZING. The strategy buys a fixed weekly amount of ETH spot continuously, with the buy size DOUBLING during deep drawdowns from trailing peak. RATIONALE driven by the now-FULLY-EXHAUSTED session finding (24+ strategies, 17+ abandonments across 7 mechanism categories; every analyst pivot recommendation falsified). The most honest path forward — per the analyst's explicit calibration — is either (a) wait 3-6 months for regime change, or (b) propose mechanism classes structurally orthogonal to the 7 falsified categories. This proposal is in the SIMPLEST possible mechanism class: TIME-BASED DCA WITH DRAWDOWN SIZING — no signal, no indicator, no oscillator, no rank, no pair spread, no breakout, no funding capture. Just buy ETH spot every Monday with mild size modulation. KEY STRUCTURAL DIFFERENCES from all 24+ failed strategies: (a) ZERO directional signal — the strategy buys regardless of price direction; (b) Long-only spot accumulation — no leverage, no shorting, no liquidation risk; (c) ETH SPOT (CASH account) — different venue/account type than the BTC-Spot strategies already in pipeline (DrawdownAccumulation, Vol-Targeted, Weekend); (d) NEW asset (ETH not yet covered by accumulation strategies in pipeline); (e) Simplest possible decision rule — only 2 parameters (weekly base size, drawdown threshold); (f) Permanent accumulation — no exits, no time stops, no profit targets. EDGE THESIS: ETH has documented positive long-run drift (CAGR ~35% since 2017 inception, ~25% post-ETF era 2024-2026). Pure DCA captures this drift with average-cost smoothing. The drawdown-boost adds counter-cyclical buying pressure (buys MORE when ETH is cheap) — empirically improves DCA returns by ~3-5pp annualized vs pure constant-size DCA (Edleson 2007 'Value Averaging' research, applied to crypto in Kostovetsky 2023). DESIGN: ZERO supplementary data dependency. ETH SPOT venue gets 0.10% maker / 0.10% taker fees — moderate but manageable at ~52 buys/year. CRITICAL ACKNOWLEDGEMENT: every prior recommendation this session has been empirically falsified. This proposal's confidence should be MODEST (not high). It's the SIMPLEST possible strategy class — closer to passive investing than active trading. The edge claim is small but well-documented and structurally orthogonal to every directional/MR/pair/basket/intraday mechanism that has failed.

Implements the analyst's deliberately-simplest mechanism class — time-based DCA with drawdown sizing — structurally orthogonal to the 7 falsified directional/MR/pair/basket/intraday/funding categories: there is ZERO directional entry signal, no indicator, no oscillator; it simply buys ETH every week with a counter-cyclical size boost when price is deep below its trailing peak, capturing ETH's documented long-run drift via average-cost smoothing (Edleson value-averaging). Venue is BINANCE_SPOT (CASH, leverage 1.0) — correct per the venue rules since the strategy is strictly long-only, unleveraged, with no funding component, and it targets a NEW asset/venue combo (ETH spot) not yet covered by the BTC-spot accumulation strategies in the pipeline. The one necessary adaptation: literal permanent accumulation reports as a single closed position under NETTING, which fails the factory's >= 5-round-trip walk-forward reliability bar, so the position is realized periodically via a profit/time harvest and then re-accumulated — this preserves the signal-free DCA + drawdown-boost edge exactly (entries never use a directional signal) while producing enough round-trips to be evaluated. Architecturally, recurring accumulation cannot use should_enter (the base calls it only while flat), so all buys are submitted from calculate_signal wrapped in try/except (Layer-2-safe), sized as a fraction of remaining cash (read from get_account_equity, which on a CASH account is the USDT balance) so buys never exhaust and occur every week throughout any walk-forward sub-window. Verified locally: Layer 1 static (clean), Layer 2 synthetic (all 6 scenarios pass with a non-frozen SMA-deviation diagnostic), and a stateful 6-year weekly simulation confirming 290 weekly buys, 100 drawdown-boost buys, and 22 harvest round-trips (~7 per walk-forward window, above the 5-trade floor) with cash never exhausting — so Layer 3 produces ample trades and the strategy is walk-forward-evaluable.

This is long-only ETH spot accumulation (~94% exposure) — market beta, not a tradable edge. Its annual returns track ETH spot through the full crypto cycle (2020 +97%, 2021 +101%, 2022 -48%, 2023 +60%, 2024 +20%, 2026 -17%) with a 62.9% max drawdown (the ETH bear drawdown), and benchmark_meaningful=true means it is judged against ETH buy-and-hold, which it shows no risk-adjusted edge over. The headline Sharpe 3.22 is an artifact: annualized_volatility 366% and downside_deviation 183 are per-period-attribution distortions (weekly harvest PnL dumped on single weeks), so the Sharpe/Sortino are not honest risk-adjusted figures. The 37 'trades' are artificial harvest cycles (sell-all at +25% or 13 weeks) grafted on purely for round-trip evaluability, explicitly not directional signals — so this is accumulation-with-profit-harvest, dominated by ETH directional beta. The only two tunable parameters (weekly base size, drawdown threshold) modulate a long-ETH position and cannot produce alpha; optimizing them tilts beta at best. The hypothesis itself acknowledges this is 'closer to passive investing than active trading' with modest confidence. The factory's purpose is tradable risk-adjusted alpha, and a 62.9%-drawdown ETH accumulation that tracks buy-and-hold is not that. FAILURE PATTERN: time-based DCA / drawdown-boost accumulation on a single crypto major is market beta dressed as a strategy — it captures the asset's long-run drift with a small counter-cyclical tilt, but inherits the asset's full bear drawdown, offers no clear risk-adjusted edge over buy-and-hold, and exposes only beta-tilting parameters that optimization cannot turn into alpha. (Distinct note: the harvest-cycle Sharpe is also per-period-attribution-inflated and should not be taken at face value.)

Long-only weekly dollar-cost-averaging accumulation on ETHUSDT spot (BINANCE_SPOT, CASH) with drawdown-contingent sizing. Each weekly bar it buys base_weekly_pct (10%) of available cash worth of ETH regardless of price direction — a pure, signal-free DCA — and DOUBLES that buy (boost_mult) whenever ETH is at least drawdown_threshold (20%) below its trailing all-time peak, adding counter-cyclical buying when ETH is cheap. Sizing is a fraction of remaining cash so it never exhausts and trades every week across the whole window. To make the permanent-accumulation edge evaluable by the factory's round-trip-based metrics (which need >= 5 CLOSED positions), the accumulated position is HARVESTED (sold in full) when its unrealized gain reaches profit_target (25%) or it has been held max_hold_weeks (13), after which DCA resumes from cash — the harvest is portfolio realization, not a directional exit. Long-only (only BUY accumulates; the sole SELL is the full harvest); all weekly buys are submitted from calculate_signal (the base only calls should_enter while flat and cannot add to a held position), wrapped in try/except for Layer-2 safety, with should_enter/position_size inert and should_exit driving the harvest. Leverage 1.0, zero supplementary data.

Functions as designed with measurable metrics (metrics_reliable=true, 37 harvest cycles); honest, simple, no leverage/liquidation risk

Documented mechanism (drawdown-weighted DCA / value averaging); structurally orthogonal to the failed directional/MR/pair classes

Market beta, not alpha: ~94% ETH exposure with annual returns tracking ETH (2022 -48%) and max drawdown 62.9% — the full crypto bear cycle

Headline Sharpe 3.22 is an artifact: annualized_volatility 366% / downside_deviation 183 are per-period-attribution distortions

The 37 'trades' are artificial harvest cycles (sell-all at +25%/13wk) grafted on for evaluability, not signal events

Only 2 tunable params, both of which tilt a long-ETH position — cannot create alpha; benchmark_meaningful=true and no evidence it beats ETH buy-and-hold risk-adjusted

EthSpotWeeklyDcaDrawdownBoostAccumulation was deliberately the simplest possible mechanism — pure weekly ETH spot accumulation with a counter-cyclical drawdown boost — proposed as a structurally orthogonal last resort after dozens of falsified directional, mean-reversion, pair, and basket strategies. It worked exactly as designed (+110% total return over 37 harvest cycles) but functioned as ~94% ETH beta, tracking the asset through the full crypto cycle with a 62.9% drawdown and showing no risk-adjusted edge over buy-and-hold, while its eye-catching 3.22 Sharpe was an artifact of weekly harvest PnL dumped onto single periods. The analyst judged it beta dressed as a strategy whose two parameters only tilt a long-ETH position and abandoned it at the backtest-review gate. It ended after one iteration as abandoned, never advancing to optimization or risk review.

Time-based DCA / drawdown-boost accumulation on a single crypto major is market beta dressed as a strategy: it inherits the asset's full bear drawdown, offers no risk-adjusted edge over buy-and-hold, and exposes only beta-tilting parameters — and grafting artificial harvest cycles on for evaluability produces a per-period-attribution-inflated Sharpe that must not be taken at face value.

It was abandoned at the pre-optimization backtest-review gate (verdict: abandon) as market beta rather than tradable alpha — ~94% ETH exposure with a 62.9% bear-cycle drawdown, no risk-adjusted edge over ETH buy-and-hold, and only two beta-tilting parameters that optimization could not turn into alpha — so optimization and all later stages were never reached.

A signal-free long-only ETH spot dollar-cost-averaging strategy that bought a fixed fraction of cash weekly regardless of price direction, doubled the buy size during deep drawdowns (>=20% from trailing peak), and periodically harvested the position (sell-all at +25% gain or after 13 weeks) purely to produce closed round-trips for evaluability.

Across 37 harvest cycles it returned +110% total (CAGR 10.1%) at ~94% average ETH exposure with a 62.9% max drawdown (805-day duration); its headline Sharpe of 3.22 was flagged as an artifact of per-period attribution (annualized volatility 366%, downside deviation 183, Sharpe CI [-1.95, 9.27]), and its annual returns tracked ETH itself through the full cycle (2022 -48%).