SolSpot25DayDrawdownAccumulationRecoveryLong

SOL Spot 25-Day Drawdown Accumulation with Recovery Exit (Daily)

A long-only single-instrument accumulation strategy on SOLUSDT spot (BINANCE_SPOT) using daily bars and OHLCV-only data. Architecturally mirrors the highly successful BTC Spot 30-Day Drawdown Accumulation (Sharpe 3.33, paper_stage) and the same architecture applied to ETH spot already in the pipeline — but applied to SOL, which is fundamentally different in volatility character. SOL exhibits ~1.5× the realized volatility of BTC/ETH, which means: (a) drawdown signals fire MORE often (higher trade frequency, addressing the sparse-signal failure mode that doomed AVAX Calm-Regime and DOT Pullback), (b) recoveries are larger when they occur (5-30% per signal vs BTC's 5-15%), (c) the 200-day SMA regime filter is even more critical because SOL's bear regimes are deeper and longer (e.g., 2022 -95% drawdown). Rather than copying BTC's 30-day / 12% threshold, this strategy uses a 25-day / 18% threshold — calibrated to SOL's higher vol so trade frequency stays in the proven 5-10/year zone (the explicit failure-analysis target), NOT the 2-3/year sparsity zone that killed AVAX Calm-Regime. Single-dominant-filter design (200-SMA + drawdown threshold), NO multi-condition AND-gate (which has been the recurring failure mode for low-frequency strategies). Uses ONLY OHLCV — no funding rates, no liquidations, no proxies — eliminating the bait-and-switch failure mode and the 'trigger disappeared' failure mode (BTC negative funding). SOL spot has not been used in any current pipeline strategy; existing SolDailyEmaTrendContinuationLong is perp futures with a different mechanism (EMA crossover continuation, not drawdown DCA), so this adds genuine portfolio diversification across both venue (spot vs perp) and mechanism (drawdown DCA vs trend continuation).

Implements the hypothesis by porting the proven BTC Spot 30-Day Drawdown Accumulation architecture (Sharpe 3.33) to SOL spot, recalibrated for SOL's ~1.5x realized volatility. The 25-day lookback + 18% drawdown threshold (vs BTC's 30/12%) is chosen so signals fire often enough to land in the proven 5-10 trades/year band, avoiding the 2-3/year sparsity that killed AVAX Calm-Regime and DOT Pullback. A single dominant filter (200-SMA regime floor + drawdown threshold) avoids the multi-condition AND-gate failure mode. The wider 20% stop matches SOL's larger swings. Uses ONLY OHLCV — no funding/liquidations/proxies — eliminating bait-and-switch and 'trigger disappeared' failure modes. Routed to BINANCE_SPOT (CASH, leverage=1) because the strategy is long-only, unleveraged, and not funding-based, per the venue-selection rule. Provides genuine diversification: SOL spot is unused in the current pipeline and the mechanism (drawdown DCA) differs from the existing SolDailyEmaTrendContinuationLong perp (EMA-crossover continuation). Per-trade sizing is 20% of equity (>$10 min notional), capital-relative via get_account_equity(). The continuous tanh(log-return) signal satisfies the Layer-2 verifier while the event-driven DCA mechanism runs in on_bar.

Failed deflated Sharpe: DSR=0.0008, expected-max=11.42 over 225 trials — the optimized Sharpe of 4.49 is far below the best-of-N luck bar and statistically indistinguishable from selection noise (is_significant=false, PBO 0.6587 > 0.5). Compounded by failures across every other gate: sensitivity FAILED with 4 cliff parameters (max_hold, regime_sma_period, drawdown_threshold, regime_breakdown_pct — e.g. drawdown_threshold 0.144 yields Sharpe -0.39 vs 0.18 yields 3.22); walk-forward is_overfitted=true with avg IS 10.9 collapsing to avg OOS 0.0 and ALL three OOS windows exactly [0.0, 0.0, 0.0] (the optimized params produce no qualifying trades out-of-sample); and the holdout failed catastrophically with a NEGATIVE Sharpe of -5.43 (actively losing) against a 0.0 WF-OOS reference. The optimizer relaxed drawdown_threshold from 0.18 to 0.09 to manufacture 54 in-sample trades, curve-fitting on a base of only 19 trades and defeating the hypothesis's own single-dominant-filter premise. This is overfit-driven failure on a thin sample, not a tunable-region case: the 225-trial sweep already established that its best config produces zero OOS trades and a losing holdout over a cliff-ridden space, so a second pass would hit the identical wall — there is no robust region to tune toward. The drawdown-accumulation mechanism is validated and promoting on its BTC spot sibling (Sharpe 3.33, paper_stage), so SOL is a dead target for it rather than a premise needing reframing. FAILURE PATTERN: single-asset daily drawdown-DCA ported to a higher-vol alt (SOL) overfits in optimization — high IS Sharpe, but sensitivity cliffs on the drawdown/regime thresholds, zero walk-forward OOS trades, and a deeply negative holdout; thin baseline trade count (~19) cannot support fitting 7 parameters.

Long-only SOLUSDT spot (BINANCE_SPOT) daily drawdown-accumulation strategy. Buys when SOL has fallen >=18% from its rolling 25-day high while still within 10% of its 200-day SMA (regime filter). Exits on recovery to the entry-time rolling high, after 60 bars, on a 20% stop-loss, or on a deep regime collapse below the 200-SMA. OHLCV-only, single dominant filter, no leverage, no shorting.

Healthy baseline economics: profit factor 1.54, max drawdown 14.8%, recovery factor 1.87, favorable win/loss asymmetry (avg_win $6.8k vs avg_loss $3.2k) at 42% win rate

Clean, non-outlier-dependent distribution: kurtosis 1.80, skew 0.26 — return is broadly sourced, not one lucky trade

Reasonably consistent across regimes: positive in 2021/2023/2024, only modestly negative in 2022/2025

Proven architecture (mirrors paper-stage BTC Spot drawdown DCA, Sharpe 3.33); trades match the drawdown-accumulation hypothesis; low fee drag (1.6% of gross)

Trade density 19 over 5.5 years (~3.5/yr) is below the hypothesis's 5-10/yr target and thin for optimizing ~7 params across 3 walk-forward windows (~6/window) — overfitting risk

Very wide Sharpe CI [-4.51, 10.3] — the headline Sharpe 3.22 is not yet statistically trustworthy on this sample size

Modest absolute return (19% over 5.5 years, CAGR ~3.8%) — the edge is real but small

Clean implementation; trades match the long-only drawdown-accumulation hypothesis

Baseline distribution was healthy (kurtosis 1.8, not outlier-dependent) — the candidate was worth testing

Deflated Sharpe 0.0008 (<<0.95); optimized Sharpe 4.49 is far below the expected-max-by-luck of 11.42 over 225 trials — best-of-N selection noise (is_significant=false)

Sensitivity FAILED: 4 cliff parameters (max_hold, regime_sma_period, drawdown_threshold, regime_breakdown_pct) — e.g. drawdown_threshold 0.144 → Sharpe -0.39 vs 0.18 → 3.22

Walk-forward is_overfitted=true: avg IS 10.9 collapses to avg OOS 0.0 with all three windows [0.0, 0.0, 0.0] — optimized params produce no qualifying trades out-of-sample

Holdout failed catastrophically: holdout Sharpe -5.43 (actively losing), WF-OOS reference 0.0, ratio 0

PBO 0.6587 (>0.5); optimizer relaxed drawdown_threshold from 0.18 to 0.09 (a cliff param) to curve-fit the in-sample, defeating the hypothesis's single-dominant-filter premise

Only 19 baseline trades — too thin to fit 7 parameters without overfitting, as the OOS collapse confirms

SolSpot25DayDrawdownAccumulationRecoveryLong applied the paper-stage BTC spot drawdown-DCA mechanism to higher-volatility SOL, with a 25-day/18% calibration intended to keep trade frequency in a healthy band and avoid prior sparsity failures. Its base backtest looked promising and non-outlier-dependent (Sharpe 3.22, PF 1.54, 14.8% drawdown), earning an 'optimize' verdict, but the thin 19-trade sample doomed the 7-parameter optimization: deflated Sharpe near zero, four sensitivity cliffs, an in-sample Sharpe of ~10.9 collapsing to zero out-of-sample, and a deeply negative holdout. The analyst judged it overfit-driven with no robust region to tune toward and abandoned it as a dead target — noting the mechanism remains validated and promoting on its BTC spot sibling. It reached optimization and analysis but was abandoned after a single iteration, never advancing to risk review.

Single-asset daily drawdown-DCA ported to a higher-vol alt overfits in optimization: a thin ~19-trade baseline cannot support fitting 7 parameters, the optimizer relaxed the drawdown threshold from 0.18 to 0.09 (defeating the single-dominant-filter premise) to manufacture in-sample trades, and the result showed threshold cliffs, zero OOS trades, and a losing holdout.

After passing the backtest-review gate (verdict: optimize), the analyst abandoned it at the post-optimization stage: it failed deflated Sharpe (DSR 0.0008 vs expected-max luck bar 11.42 over 225 trials, is_significant=false, PBO 0.66), failed sensitivity with 4 cliff parameters, was flagged walk-forward overfit (avg IS 10.9 collapsing to all-zero OOS windows), and failed the holdout catastrophically (Sharpe -5.43).

A long-only drawdown-accumulation strategy on SOLUSDT.BINANCE_SPOT daily bars (OHLCV only) that bought when price had drawn down ≥18% from its rolling 25-day high while not catastrophically below the 200-day SMA, exiting on recovery to the entry-time rolling high, a 60-bar hold, a 20% stop, or a regime collapse — porting the paper-stage BTC 30-day drawdown DCA mechanism to higher-volatility SOL with vol-calibrated thresholds.

The base backtest over 2020-08-11 to 2026-05-18 showed healthy economics across 19 trades — total return +19.2%, CAGR 3.83%, Sharpe 3.22 (CI [-4.51, 10.3]), profit factor 1.54, 42% win rate, 14.8% max drawdown, low outlier dependence (kurtosis 1.8) — and optimization produced a higher headline (Sharpe 4.49, +59%, 54 trades) that did not generalize.