BtcDailyDonchianBreakoutLong

BTC Daily Donchian Channel Breakout Long-Only (BINANCE USD-M Futures, Daily, OHLCV-Only)

A long-only, single-instrument, single-venue, OHLCV-ONLY breakout strategy on BTCUSDT.BINANCE 1D bars. Classic Turtle-trader Donchian channel mechanism: enter long when close breaks above the 20-day Donchian channel high (new 20-day high); exit when close drops below the 10-day Donchian channel low (10-day trailing exit). Distinct from existing trend (golden cross 50/200) and mean-reversion (RSI < 30, drawdown) strategies — captures breakout momentum from consolidation ranges. Per factory directive: BTCUSDT.BINANCE 1D bars are always safe. Three parameters total (entry_window=20, exit_window=10, stop_pct).

Implements the classic Turtle Donchian breakout exactly as specified: entry_window=20 high-channel breakout entry, exit_window=10 low-channel trailing exit, plus a stop_pct hard stop. The breakout test uses the prior N bars' high (bars[-N-1:-1], excluding the current bar) so a close above it is a genuine new high, not a self-referential comparison. The signal is continuous (distance past the channel as a fraction) rather than a gated constant, so the verifier sees a varying decision variable. VENUE OVERRIDE: the hypothesis suggested BINANCE USD-M futures, but this strategy is long-only, uses no leverage, and has no funding-rate edge — by the venue-selection rule that places it on BINANCE_SPOT (CASH account, leverage hard-capped at 1). Spot also guarantees we never submit a SELL-to-open; exits only SELL the held long. Daily timeframe matches the multi-week breakout horizon, and breakouts captured this way routinely move far more than the ~0.20% spot round-trip fee, leaving ample per-trade edge. It is mechanistically distinct from trend (golden-cross MA) and mean-reversion (RSI/drawdown) strategies — it trades range breakouts via price-extreme channels.

Failed deflated Sharpe: DSR=0.0408, expected-max=9.08 over 225 trials — the optimized Sharpe of 5.99 is below the best-of-N luck bar and statistically indistinguishable from selection noise (is_significant=false, PBO 0.6362 > 0.5). Compounded by a textbook overfit signature: walk-forward is_overfitted=true with avg IS 8.48 collapsing to avg OOS -1.586 and OOS windows [-2.25, -2.50, 0.0] (two deeply negative, none positive), plus a failed sensitivity pass with 2 cliff parameters (exit_window, risk_fraction; leverage >1.0 collapses Sharpe to 0). The holdout 'passed=true' is a DEGENERATE artifact, not a validation: its WF-OOS reference is negative (-1.586) and the ratio is 0, so a positive holdout Sharpe (1.81) measured against a negative reference trips a meaningless pass — it must not override the failed DSR/PBO/OOS gates. PSR 0.9996 does not rescue it because PSR ignores the 225-trial selection that DSR corrects for. This confirms the plausibility concern flagged at pre-optimization: the baseline Sharpe (3.57) and headline rested on a handful of 2021 bull-run outliers (the +345% 2021-03 day), and once optimization diluted that outlier (kurtosis 31.6 -> 5.6) the out-of-sample edge went negative — the breakout 'edge' was outlier/regime artifact, not a generalizable signal. This is overfit-driven failure, not a tunable-region case: the 225-trial sweep already established its best config loses across every out-of-sample window over a cliff-ridden space, so a second pass would hit the identical wall — no robust region exists to tune toward. The Turtle Donchian mechanism fit BTC's noise rather than a structural edge, so this is a dead instance rather than a premise to reframe. FAILURE PATTERN: long-only daily Donchian/Turtle breakout on BTC shows a high in-sample optimized Sharpe carried by 2021 bull-run outliers but fails DSR/PBO with deeply negative walk-forward OOS and only a degenerate (negative-reference) holdout pass.

Long-only Turtle-style Donchian channel breakout on BTCUSDT 1D bars (Binance spot). Enters long when the daily close breaks above the highest high of the prior 20 bars (new 20-day high), capturing breakout momentum out of consolidation ranges. Exits when the close falls below the lowest low of the prior 10 bars (10-day Donchian trailing stop) or hits a hard 8% stop-loss from entry. calculate_signal() returns the continuous breakout fraction (close - 20-day-high)/20-day-high so the signal varies every bar; should_enter buys when it turns positive. Positions are sized to 95% of equity (spot, 1x, no leverage).

Faithful implementation of the classic Turtle Donchian-breakout mechanism, long-only, OHLCV-only, 3 parameters (low overfitting risk)

Strong positive expectancy: profit_factor 1.70, omega 3.67, positive in 7 of 10 years across a full 2017-2026 multi-regime sample

Correctly routed to BINANCE_SPOT (long-only/no-leverage belongs on spot) despite the title saying futures — venue is right

46 trades over 9 years with ~28-day holds — framework-evaluable for walk-forward and holdout

Headline metrics are outlier-inflated and implausible as steady-state: annualized_volatility 848%, Sortino 51.9, Calmar 66 — not to be taken at face value

Single trades show +345% (2021-03) and +91.9% (2019-06), implausibly large for an unlevered long-only spot position — must verify it isn't a sizing/compounding artifact

Extreme outlier dependence (skew 5.17, kurtosis 31.6, tail_ratio 4.88); 2021 alone (+377%) carries the record

max_drawdown 50% (CI to 76%) — heavy; recent regime soft (2025 fading, 2026 -16.7%)

Standard, cleanly-implemented Turtle Donchian mechanism; optimization did reduce outlier dependence (kurtosis 31.6 -> 5.6, max DD 50% -> 29%)

Failed deflated Sharpe: DSR 0.0408 (<<0.95), is_significant=false; optimized Sharpe 5.99 below the best-of-225 luck bar of 9.08

PBO 0.6362 (>0.5) — parameter selection more likely than not overfit

Walk-forward overfit: is_overfitted=true, avg IS 8.48 collapsing to avg OOS -1.586, OOS windows [-2.25, -2.50, 0.0] (two deeply negative, none positive)

Holdout 'passed' is degenerate: WF-OOS reference is negative (-1.586), ratio 0 — a false pass that cannot validate the strategy

Sensitivity failed with 2 cliffs (exit_window, risk_fraction); leverage >1.0 collapses Sharpe to 0

Confirms the pre-opt plausibility concern: the baseline Sharpe rested on a few 2021 bull-run outliers (the +345% day), and once optimization diluted them the out-of-sample edge went negative

BtcDailyDonchianBreakoutLong implemented the classic Turtle 20/10-day Donchian breakout on BTC daily, cleanly coded with just three parameters and correctly routed to spot. It cleared the pre-optimization gate on strong expectancy (+2419%, PF 1.70), though reviewers flagged its outlier-inflated headline (kurtosis 31.6, 848% vol) as resting on 2021's bull run. Optimization confirmed the concern: deflated Sharpe near zero, PBO 0.64, a walk-forward whose in-sample 8.48 collapsed to a negative OOS average across all windows, two sensitivity cliffs, and only a meaningless negative-reference holdout pass. The analyst judged the edge an outlier/regime artifact with no robust tuning region — a dead instance rather than a reframe — and abandoned it. It reached optimization and analysis but was abandoned after a single iteration, never advancing to risk review.

Long-only daily Donchian/Turtle breakout on BTC shows a high in-sample Sharpe carried by a handful of 2021 bull-run outliers (a single +345% day), but once optimization dilutes those outliers the out-of-sample edge goes negative — the breakout fit BTC's noise and bull-cycle beta rather than a generalizable structural signal, and a degenerate holdout pass must never override failed DSR/PBO/OOS gates.

After passing the backtest-review gate (verdict: optimize), the analyst abandoned it at the post-optimization stage: it failed the deflated Sharpe test (DSR 0.041 vs expected-max luck bar 9.08 over 225 trials, is_significant=false, PBO 0.64), the walk-forward was flagged overfit (avg IS 8.48 collapsing to avg OOS -1.59 with two deeply negative windows), sensitivity failed with 2 cliffs, and the holdout 'pass' was a degenerate artifact of a negative OOS reference.

A long-only single-instrument Turtle-style Donchian channel breakout on BTCUSDT daily bars (OHLCV only, routed to BINANCE_SPOT) that entered long when close broke above the prior 20-day high and exited when close fell below the prior 10-day low or hit an 8% stop — a 3-parameter breakout-momentum mechanism distinct from the pipeline's golden-cross trend and RSI/drawdown mean-reversion strategies.

The base backtest over 2017-08-17 to 2026-05-18 showed strong positive expectancy across 46 trades — total return +2419%, CAGR 45.8%, profit factor 1.70, positive in 7 of 10 years — but headline metrics were outlier-inflated (Sharpe 3.57 with CI [0.75, 6.32], skew 5.17, kurtosis 31.6, annualized vol 848%, 50% max drawdown), and optimization (Sharpe 5.99, +454%) reduced outlier dependence but collapsed out-of-sample.