BnbDailyMorningStarReversalLong200SmaRegime

BNB Daily Morning Star Reversal Long with 200-SMA Bull Regime Filter

A long-only single-instrument 3-BAR CANDLESTICK-PATTERN strategy on BNBUSDT perpetual futures using daily bars and OHLCV-only data, with a 200-SMA bull regime filter. The trigger is the classic 'Morning Star' pattern documented by Steve Nison (Japanese Candlestick Charting Techniques, 1991) and widely tested across asset classes for ~35 years. Morning Star is a SPECIFIC 3-BAR TEMPORAL SEQUENCE: Bar 1 is a long bearish candle (continuation of a prior down-move); Bar 2 is a small-body candle (doji/spinning top — indecision and exhaustion of the down-move); Bar 3 is a long bullish candle that closes above the midpoint of Bar 1's body (reversal initiation). This is mechanistically NOVEL within the portfolio: every existing candlestick strategy fires on either 1-bar (Hammer rejection, Inside Day compression) or 2-bar (Engulfing) configurations. NO existing strategy fires on a 3-bar temporal sequence. The 3-bar requirement means Morning Star captures a fundamentally different microstructure: it requires the market to traverse three distinct states — continuation, exhaustion-pause, reversal-initiation — across three full trading sessions. This makes the signal far stricter (and therefore sparser but higher-quality) than 1-bar or 2-bar patterns. The economic mechanism: Bar 1 represents committed sellers; Bar 2's small body shows seller commitment exhausted (no follow-through despite the bear setup); Bar 3 confirms buyer takeover with a strong bullish close above Bar 1's midpoint. BNB is chosen specifically because it is the most UNDERUTILIZED asset in the portfolio (only BnbFourHourVolumeBreakoutLong currently — 1 strategy vs 2-3 for every other tracked asset). BNB has well-documented daily-candlestick reversal dynamics driven by its Binance-ecosystem-specific flows: BNB-burn events, exchange-token-utility cycles, and broader Binance regulatory news produce sharp directional moves followed by clean reversal patterns. Daily timeframe is canonical for candlestick patterns (Nison's original tests were primarily daily). Only 3 explicit parameters (regime SMA period, body-size ratio for the indecision bar, stop-loss pct) — minimum-parameter-count for a 3-bar pattern strategy.

Implements the hypothesis precisely as a 3-bar temporal sequence (continuation -> exhaustion-pause -> reversal), which is structurally distinct from 1-bar and 2-bar candlestick patterns in the portfolio. Bar1/Bar2/Bar3 map to self._bars[-3]/[-2]/[-1] (the base class appends the current bar before calculate_signal), and the 'long candle' test normalizes each body against the average body over the 14 bars BEFORE the pattern (_avg_body uses [-(lookback+3):-3]) so the pattern bars don't contaminate their own benchmark. The three headline edge parameters requested by the hypothesis are exposed (sma_period, indecision_body_ratio, stop_loss_pct); supporting risk knobs are parameters too, with no hardcoded magic numbers. calculate_signal returns a CONTINUOUS reversal score (Bar-3 body/range + tanh penetration above Bar-1 midpoint) that varies every bar so Layer-2 frozen-signal detection passes, while the strict discrete Morning Star + regime gate is cached and read in should_enter — keeping the signal continuous rather than a gated constant. Venue routing: the strategy is long-only, uses no leverage, and has no funding edge, so by the venue-selection rule it belongs on BINANCE_SPOT (CASH) rather than futures; entries are BUY-only and exits SELL to close (never shorts), satisfying the CASH-account constraint. Sizing is capped at 95% of equity so every BUY remains fillable from available USDT on the CASH account, and leverage=1.0 (spot hard cap) is referenced in sizing. BNBUSDT has deep daily spot history, avoiding data_unavailable. The 10% TP / 5% SL multi-day swings dwarf the ~0.20% spot round-trip fee, and the strict 3-bar+regime filter keeps trades sparse but high-quality.

Not worth optimizing — insufficient sample plus no edge. The strict 3-bar Morning Star pattern gated by a 200-SMA bull filter fires only ~2x/year on BNB, producing just 19 trades over 8.5 years — far below any meaningful optimization floor (a 3-window walk-forward would have ~6 trades/window and overfit pure noise). What evidence exists is negative: profit_factor 0.86, expectancy -$138/trade, Sharpe -1.0, win_rate 37%, total return -2.5%, losing in most years. Tuning cannot create statistical significance from 19 events, and there is no positive base to build on. This is the same low-frequency-chart-pattern pathology seen on the abandoned SUI three-bar reversal. NOTE for research: a serious test of the Morning Star would need to POOL the pattern across many instruments to reach trade-count sufficiency — that is a different multi-instrument hypothesis, not a parameter iteration of this single-BNB version. FAILURE PATTERN: strict low-frequency candlestick chart patterns (~<10 signals/yr) on a single instrument do not provide enough trades to survive walk-forward optimization.

Long-only single-instrument 3-bar Morning Star candlestick reversal on BNBUSDT spot, daily bars, OHLCV-only, gated by a 200-day SMA bull regime filter. It fires when the last three completed daily candles form Nison's Morning Star: Bar 1 a long bearish candle, Bar 2 a small-body indecision candle (body <= indecision_body_ratio x the larger of Bar 1's body and the average body), and Bar 3 a long bullish candle that closes above the midpoint of Bar 1's body — and only when the current close is above the 200-SMA. It exits on a 10% take-profit, 5% stop-loss, a regime break (close falls below the SMA), or a 20-bar time stop. Risk-based sizing (1% equity vs the stop distance) capped at 95% of cash. Long-only spot, leverage 1.0.

Mechanism correctly implemented: strict 3-bar Morning Star + 200-SMA regime gate, clean diagnostics, low drawdown (11%).

Genuinely novel 3-bar pattern within the portfolio; minimal parameter count.

Only 19 trades over ~8.5 years (~2/yr) — far too few for a 3-window walk-forward (~6 trades/window); optimization would fit noise.

No edge: profit_factor 0.86, expectancy -$138/trade, Sharpe -1.0, win_rate 36.8%, total return -2.5% over 8.5 years.

Negative in most years (2021, 2023, 2024, 2025) — no working edge to tune.

Strict 3-bar pattern + bull-regime filter is structurally too rare on a single instrument to generate an optimizable sample; BNB history cannot be extended to fix this.

BnbDailyMorningStarReversalLong200SmaRegime introduced a genuinely novel 3-bar Morning Star reversal to the portfolio on the underutilized BNB, gated by a 200-SMA bull filter and built with minimal parameters. The pattern was correctly implemented but fired only 19 times in 8.5 years, and that thin evidence was negative — profit factor 0.86, Sharpe -1.0, -2.5% return, losing in most years. The analyst abandoned it at backtest review after one iteration, citing both an insufficient sample and no edge, explicitly echoing the same low-frequency-pattern pathology that sank the SUI three-bar reversal. The recorded failure pattern is that strict single-instrument candlestick patterns need cross-instrument pooling to become optimizable.

Strict low-frequency candlestick chart patterns (~<10 signals/yr) on a single instrument don't generate enough trades to survive walk-forward optimization; a serious Morning Star test would need to pool the pattern across many instruments, which is a different multi-instrument hypothesis rather than a parameter iteration.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate: the strict 3-bar pattern plus bull-regime filter is structurally too rare on a single instrument (~6 trades per walk-forward window), and the existing evidence was negative — tuning cannot manufacture significance from 19 events, and BNB's history can't be extended to fix it.

A long-only single-instrument candlestick-pattern strategy on BNBUSDT daily bars (routed to BINANCE_SPOT) that fired on Steve Nison's classic 3-bar Morning Star reversal — a long bearish bar, a small-body indecision bar, then a long bullish bar closing above bar 1's midpoint — taken only in a bull regime (close above the 200-day SMA), exiting on take-profit, stop, regime break, or time-stop.

Over ~8.5 years it fired only 19 times (~2/yr) with no edge: profit factor 0.86, expectancy -$138/trade, Sharpe -1.0, 36.8% win rate, and -2.5% total return, losing in most years (2021, 2023, 2024, 2025) despite a low 11% max drawdown.