BnbSpotRegimeHoldLong

BNB Spot Long-Hold with 200-SMA Regime Filter (Daily, Pure Regime-Flat Exit)

A long-only single-instrument PURE regime-conditional buy-and-hold strategy on BNBUSDT spot (BINANCE_SPOT) using daily bars and OHLCV-only data. This applies the SIMPLEST POSSIBLE mechanism (single 200-SMA regime filter, binary regime-flat exit, one tunable parameter) to BNB as the fourth asset for this proven architecture class (after EthSpotRegimeHoldLong, SolSpotRegimeHoldLong, LinkSpotRegimeHoldLong — all pending or recently submitted). Critically, this is NOT the same as the failed BNB Spot Drawdown DCA (which had coding_stuck timeouts — a complex 30-day-drawdown + ATR-trail mechanism with multiple parameters). This is the OPPOSITE of complex: just one parameter (200-SMA period), one entry condition (regime change), one exit condition (regime change back). The simplicity is the structural defense against ALL the failure modes documented this session: (a) cannot have outlier-trade dependency (holds full regime), (b) cannot have parameter cliffs (only 1 parameter), (c) cannot have ATR-trail volatility-calibration issues (no trail), (d) cannot have walk-forward catastrophic OOS (binary mechanism produces signals across all market cycles by definition), (e) cannot have signal redundancy with any existing BNB strategy (BNB has only 4H Volume Breakout — which is intraday + volume-based — completely orthogonal to daily binary regime hold). BNB-SPECIFIC empirical justification: BNB has multiple working strategies confirming clean response to filters (BNB 4H Volume Breakout Sharpe 2.03 promoted). The 200-SMA filter has been validated across BTC/ETH/SOL/LINK as a regime classifier. NO bait-and-switch — strictly BNBUSDT.BINANCE_SPOT throughout. ONE tunable parameter only. Recent infrastructure failures (4 verification_loop failures with 'auditor heartbeat stale') were NOT strategy-quality issues — the simplest mechanism here should be MOST likely to pass verification once infrastructure stabilizes.

Implements the simplest proven regime-hold architecture (after the ETH/SOL/LINK spot regime-hold siblings) on BNB. calculate_signal returns the continuous normalized distance of price from the 200-SMA, (close - sma)/sma — positive in a bull regime, negative in bear — so the signal varies every bar and carries conviction, rather than being a flat binary; should_enter/should_exit reduce that to the regime test (>0 long, <=0 flat). The single-parameter design is the structural defense the hypothesis emphasizes: no outlier-trade dependency (it holds the whole regime), no parameter cliffs (one parameter), no ATR-trail calibration risk, and signals fire across every market cycle by construction. VENUE: long-only with no leverage and a pure price-momentum/OHLCV edge (not funding) — exactly the BINANCE_SPOT routing case — so it trades BNBUSDT.BINANCE_SPOT (CASH) as the hypothesis specifies, with leverage left at 1.0 and deliberately unreferenced in sizing (avoiding the leverage_set_but_unused rejection). Turnover is a handful of regime flips over years, so the ~0.20% spot round-trip fee is negligible against multi-month holds.

Failed deflated Sharpe: DSR=0.0443, expected-max=6.14 over 225 trials — the optimized Sharpe of 3.26 is below the best-of-N luck bar and statistically indistinguishable from selection noise (is_significant=false, sharpe_ci_low -12.99 straddles 0, PBO 0.8585 — the highest overfitting probability of the session). Overfit by over-selectivity: walk-forward is_overfitted=true with avg IS 5.88 collapsing to avg OOS 0.0 and ALL three OOS windows exactly [0.0, 0.0, 0.0] — the optimizer's chosen config (sma_period 200->297) produces ZERO out-of-sample trades in every window, so there is no out-of-sample activity to validate, and the holdout is a degenerate fail (holdout Sharpe 0.0 against a 0.0 WF-OOS reference, ratio 0). The 3306% optimized headline is a single-trade mirage: a 2022 trade drives a +1434% annual return (a +2036% single-day whole-trade attribution on 2022-01-08), with 74% of the headline open-position unrealized, win rate 16%, 8 consecutive losses, and negative 2018/2020/2023 years; annualized_volatility 6377% reflects the per-day-attribution distortion. This is overfit/single-trade concentration on a 1-parameter space, not a tunable-region case: the sweep already established its best config produces zero OOS trades, so a second pass hits the identical wall — there is no robust region to tune toward (sensitivity cliff_count=1 is the trivial spot cap, necessary but not sufficient). Not revise_hypothesis either: the pure-200-SMA-regime-hold mechanism has no promoted sibling to redirect and the failure is overfit/outlier dependence, not structural decay of a working mechanism — reframing onto another asset would just re-overfit. FAILURE PATTERN: single-instrument 200-SMA regime-hold on a high-vol crypto produces a headline carried by one bull-cycle outlier trade plus end-of-backtest open-position unrealized, passes sensitivity (1 trivial param) but fails DSR/PBO/holdout with all-zero walk-forward OOS as the optimizer thins to a long SMA period and the edge fails to generalize — third consecutive identical failure (SOL 200-SMA, SOL multi-week, BNB 200-SMA) this session.

Long-only, single-instrument regime-conditional buy-and-hold on BNBUSDT spot, DAILY bars. A single 200-day SMA classifies the regime: when the close is above the SMA (bull regime) the strategy holds a long; when the close falls below the SMA (bear regime) it goes flat. One tunable parameter (the SMA period) — no stops, no trailing, no second condition. Sized to 95% of equity, long-only with no leverage.

Strategy functions correctly and trades match the hypothesis: binary 200-SMA regime hold, long above / flat below. The dominant winner is a real 350-day BNB bull-regime hold (16.80->262.10), not a data artifact.

13% win rate is the expected structural signature of a single-SMA regime filter (many small whipsaw exits near the SMA, rare large regime wins); net profit factor 1.76 is positive.

Maximally simple — one tunable parameter (sma_period) — which is genuine structural defense against parameter cliffs and overfitting in optimization.

Clean single-instrument OHLCV-only spot routing, no bait-and-switch; same regime-filter mechanism being validated across the BTC/ETH/SOL/LINK family. 54 trades is ample sample.

Extreme single-trade concentration: the one +1460% 2020-21 trade drives essentially the entire 1759% return; strip it and the strategy is a net loser. Kurtosis 50, skew 7.0 confirm.

Most years are negative (2018/2020/2022/2023/2025/2026); only 2019/2021/2024 positive. Recent regime decaying (2025 -14.6%, 2026 -2.2%), so the holdout window lands in the negative zone — the recurring failure setup this session.

Severe path risk: max_drawdown 75%, 10 consecutive losses, 940-day max drawdown duration — punishing to hold between the rare winners.

Headline Sharpe (2.21) unreliable: sharpe_ci_low -11.5 straddles 0, annualized_volatility 2992% distorted by per-day whole-trade PnL attribution; impact_cost 6.1% of gross is non-trivial.

ample sample

1 of 54 trades dominates

watch

yes

must match

Clean implementation matching the hypothesis (binary 200-SMA regime hold, one tunable parameter); sensitivity has no genuine cliff (only the trivial risk_fraction>1.0 spot cap).

Dominant trades are real BNB bull-regime holds, not data artifacts.

Deflated Sharpe 0.0443 vs the 6.14 best-of-225 luck bar — optimized Sharpe 3.26 below the noise floor; is_significant=false, sharpe_ci_low -12.99 straddles 0.

PBO 0.8585 — the highest overfitting probability of the session.

Walk-forward is_overfitted=true: avg IS 5.88 collapses to avg OOS 0.0, all three OOS windows [0.0, 0.0, 0.0]; the chosen config (sma_period 297) produces zero out-of-sample trades.

Holdout degenerate-fail (Sharpe 0.0 vs 0.0 WF-OOS reference, ratio 0).

3306% headline is one 2022 trade (+1434% annual, +2036% single-day attribution) plus 74% open-position unrealized; win rate 16%, 8 consecutive losses, negative 2018/2020/2023; annualized_volatility 6377% confirms per-day distortion.

expected-max 6.14

BnbSpotRegimeHoldLong applied the maximally simple single-200-SMA regime-hold mechanism to BNB as the fourth member of the regime-hold family, pitched as the structural opposite of the timeout-plagued BNB Spot Drawdown DCA. It cleared the pre-optimization gate on a big but heavily concentrated base — one +1460% trade drove the entire +1759%, with most years negative and a 75% drawdown — and optimization confirmed the concern: deflated Sharpe near zero, the session's highest PBO (0.86), an in-sample 5.88 collapsing to zero OOS activity, and a degenerate holdout. The analyst judged it overfit/single-trade-dependent BNB beta with no robust tuning region and no promoted sibling to redirect — neither iterable nor reframable — and abandoned it. It reached optimization and analysis but was abandoned after a single iteration, never advancing to risk review.

Single-instrument 200-SMA regime-hold on a high-vol crypto produces a headline carried by one bull-cycle outlier trade plus end-of-backtest open-position unrealized PnL; it passes sensitivity (one trivial parameter) but fails DSR/PBO/holdout with all-zero walk-forward OOS as the optimizer thins to a long SMA — the third consecutive identical failure this session (SOL 200-SMA, SOL multi-week, BNB 200-SMA), confirming the family does not generalize.

After passing the backtest-review gate (verdict: optimize), the analyst abandoned it at the post-optimization stage: it failed the deflated Sharpe test (DSR 0.044 vs expected-max luck bar 6.14 over 225 trials, is_significant=false, PBO 0.86 — the session's highest), the walk-forward was overfit (avg IS 5.88 collapsing to all-zero OOS windows because the optimized config never traded out-of-sample), and the holdout was a degenerate fail (ratio 0).

The simplest possible strategy: a long-only single-parameter regime-conditional buy-and-hold on BNBUSDT.BINANCE_SPOT daily bars (OHLCV only) that held spot BNB while price was above its 200-day SMA and moved fully to cash when it fell below — a pure binary regime-flat exit applied as the fourth asset in the regime-hold family.

The base backtest over 2017-11-06 to 2026-05-18 showed a large but distorted headline across 54 trades — total return +1759%, CAGR 46.3%, Sharpe 2.21 (CI [-11.5, 4.10]), profit factor 1.76, a 74.8% max drawdown, 13% win rate — with essentially the entire return from one +1460% 2020-21 bull-regime hold; optimization (Sharpe 3.26, +3306%, sma_period 297) produced zero out-of-sample trades.