BtcDailyGoldenCrossMomentumLongOnly1D

BTC Daily Golden Cross Momentum Long-Only (BINANCE USD-M Futures, Daily, OHLCV-Only)

A long-only, single-instrument, single-venue, OHLCV-ONLY trend-following strategy on BTCUSDT.BINANCE 1D bars. Same proven golden-cross mechanism as the already-landed EthDailyGoldenCrossMomentumLong — applied to BTC futures. Enter long when 50-day SMA crosses above 200-day SMA (the classic 'golden cross' trend confirmation); exit when 50-day SMA crosses back below 200-day SMA ('death cross'). MECHANISM PIVOT AWAY FROM ALT SPOT DCA: per the factory's explicit error message on FTM spot, 'change instrument_id or bar_type to a major symbol with long history (BTCUSDT/ETHUSDT/SOLUSDT.BINANCE on a 1H/4H/1D timeframe is always safe)'. BTCUSDT.BINANCE 1D bars are the most-reliable data path in the entire factory. This strategy uses ONLY that path. Different mechanism class than my existing spot DCA book (trend-following crossover vs drawdown mean-reversion). Three parameters total.

Implements the hypothesis on the factory's most reliable data path (BTCUSDT.BINANCE 1D) using the same proven golden-cross mechanism as the landed Eth sibling. Cross detection compares current vs prior fast/slow SMAs so entries fire exactly on the crossover bar (no repeated entries while extended), and the exit mirrors it on the death cross — a clean, low-turnover trend filter distinct from the existing spot-DCA drawdown-mean-reversion book. The real stateful logic runs in on_bar (which needs the full 200-bar history and position state), while calculate_signal returns a continuous tanh(log-return) proxy purely to keep Layer-2 synthetic scenarios non-frozen and bounded (|signal| ≤ 1, never invalid). Sizing is capital-relative (25% notional via get_account_equity) and respects the min-notional floor. Leverage is left at 1.0 so sizing references no margin leverage (avoiding the inert-leverage gate) while the USD-M account cleanly executes the long-only book.

Structurally too few trades to validate, with no value-add — not optimizable or iterable within the stated mechanism. The classic 50/200 daily golden cross fires ~once a year, producing only 5 complete trades over ~6 years (avg hold 192 days) — statistically meaningless and far below any walk-forward floor; a 225-trial optimization on 5 events is pure noise-fitting and DSR cannot be computed. The result also shows no edge: Sharpe 0.17, alpha -0.0094 (NEGATIVE — underperforms its own market exposure), CAGR 1.47%, sharpe_ci_low -2.0, and the entire +18% total return is essentially a single trade (2024-08-10 +14.5%; kurtosis 774), with 2 of 5 trades negative. NOT optimize: 5 trades validate nothing and the optimizer would curve-fit to the one 2024 trade. NOT iterate: this is the key distinction from the thin-sample volume/breakout candidates I sent back for density (BNB, ETH volume-surge) — there, lowering a threshold densifies the sample; here the sparsity is INTRINSIC to the classic 50/200 crossover the hypothesis explicitly requires, and the only way to add trades is to drastically shorten the SMAs, which abandons the golden-cross thesis and creates a different strategy. With negative alpha and single-trade dependence, there is no edge to preserve even if the sample could be densified. FAILURE PATTERN: a 50/200 daily golden-cross on a single instrument generates ~5 trades over the full history — intrinsically too sparse to validate (DSR/walk-forward impossible), and on BTC it produces negative alpha with the entire return concentrated in one 2024 trade; long-horizon SMA-crossover crossovers are structurally untestable as standalone single-instrument strategies regardless of parameter tuning within the classic-period constraint.

Long-only, single-instrument golden-cross trend-following on BTCUSDT.BINANCE USD-M perpetual, daily bars, pure OHLCV. Enters long on a 50/200 SMA golden cross (50-day SMA crossing above the 200-day SMA) and exits flat on a death cross (50-day crossing back below 200-day); a death cross never shorts. Sizes 25% of equity notional per entry. Three tunables (sma_fast_period, sma_slow_period, per_leg_pct), leverage 1.0.

Clean execution on the safest data path (BTCUSDT 1D), no artifacts (metrics_reliable=true), low drawdown (13%), recognized trend-following mechanism, fee-light (turnover 0.7)

Only 5 trades over ~6 years — structurally far too sparse (50/200 daily golden cross fires ~annually); statistically meaningless, DSR/walk-forward impossible

No value-add: Sharpe 0.17, alpha -0.0094 (NEGATIVE), CAGR 1.47%, sharpe_ci_low -2.0 (deeply straddles 0) — underperforms its own beta exposure

Single-trade-dependent: kurtosis 774; the +18% total is essentially one 2024-08 +14.5% trade; 2 of 5 trades negative

Not fixable within the mechanism — the classic 50/200 crossover is intrinsically ~5 trades; densifying requires drastically shorter SMAs = a different strategy

62 (walk-forward floor)

BtcDailyGoldenCrossMomentumLongOnly1D ported a landed ETH golden-cross strategy to BTC on the factory's safest data path, betting the classic 50/200 daily crossover would carry across the major. Execution was clean, but the mechanism's defining property doomed it: the crossover fires only about once a year, so the full ~6-year history yielded just 5 trades with a 192-day average hold — statistically meaningless, with deflated-Sharpe and walk-forward validation impossible. The result also showed no edge (Sharpe 0.17, negative alpha -0.0094, the entire +18% return riding on one 2024 trade with kurtosis 774). The reviewer abandoned it as structurally untestable: optimization on 5 events is noise-fitting and the only way to densify — drastically shorter SMAs — would abandon the golden-cross thesis entirely, so long-horizon crossovers cannot stand alone as single-instrument strategies.

A 50/200 daily golden cross on a single instrument fires ~once a year, so it generates only ~5 trades over the full history — intrinsically untestable as a standalone strategy regardless of tuning, since the only way to add trades abandons the golden-cross thesis; unlike threshold-based thin-sample candidates, this sparsity cannot be densified, and here it also delivers negative alpha concentrated in one trade.

The analyst abandoned it at the backtest-review gate, before optimization, because the 50/200 daily crossover is structurally too sparse to validate — 5 trades is far below any walk-forward floor, making DSR/walk-forward impossible and optimization pure noise-fitting — and there is no edge to preserve anyway (negative alpha, single-trade dependence), with densifying impossible without drastically shortening the SMAs into a different strategy.

A long-only, single-instrument, OHLCV-only daily trend-following strategy on BTCUSDT.BINANCE USD-M futures applying the classic 50/200 golden-cross mechanism — enter long when the 50-day SMA crosses above the 200-day SMA, exit flat on the death cross — porting the landed Eth golden-cross sibling to BTC on the factory's most reliable data path, with three parameters.

It produced only 5 complete trades over ~6 years (avg hold 192 days) with no edge: +18.2% total return but CAGR just 1.47%, Sharpe 0.17, negative alpha -0.0094, sharpe_ci_low -2.0, 60% win rate, profit factor 1.50, 13% max drawdown — and the entire return was essentially a single 2024-08 trade (+14.5%, kurtosis 774) with 2 of 5 trades negative; execution was clean (metrics_reliable=true, fee-light).