Strategies

LINK 4H Volume-Confirmed Breakout Long with Daily Regime Filter

Outcome: Abandoned

LinkFourHourVolumeConfirmedBreakoutLongDailyRegime

Outcome Summary

This strategy auto-replicated the paper-stage BNB 4H volume-confirmed breakout onto LINK, betting that LINK's clean daily trend edge would be fractal at 4H. A prior iteration produced no trades because a separate daily regime feed never populated; the iter-2 fix derived the 200-day regime from the 4H buffer, and the strategy finally traded — but disastrously. The 4H breakout signal had no edge (PF 0.54, 25% win rate, -$1,258/trade), and 95%-of-equity sizing with no volatility scaling compounded it into an account blowup: CAGR -100%, ~1,718% average position, 4,000% volatility, and physically impossible percentage swings as equity collapsed. The analyst abandoned it before optimization, judging there is nothing to tune in a total wipeout and that the BNB architecture simply does not carry an edge to LINK 4H regardless of how the sizing is fixed.

Strategy report

Hypothesis

A long-only single-instrument breakout strategy on LINKUSDT perpetual futures using 4-hour bars and OHLCV-only data, with a daily-bar regime filter. Architecturally MIRRORS the proven BNB 4H Volume-Confirmed Breakout (Sharpe 2.03, paper_stage), applied to LINK as the third sensitivity-validation test point of this construction class (BNB ✓, SOL ⏳ in pipeline, LINK proposed here). LINK already has empirical evidence of clean trend edges at the daily timeframe (LINK 1D Multi-Week Trend Continuation = Sharpe 2.55, promoting), which is a strong precondition for the 4H volume-confirmed breakout to succeed — momentum-followers tend to be fractal across timeframes, and assets with clean daily trend signatures typically also have clean 4H volume-confirmed signatures. Importantly, this proposal explicitly addresses the recently-documented 'auto-replication failure pattern' (BNB Golden Cross, ETH Spot Drawdown) by: (a) requiring sensitivity_passed=true with cliff_count=0 as a HARD GATE before optimization, (b) selecting LINK based on existing empirical evidence (proven daily edge) rather than a thesis-only argument about asset character. Existing LINK strategy in pipeline (LinkDailyMultiWeekTrendContinuationLong) uses daily bars and sustained momentum — this proposal uses 4H bars and intraday volume-confirmed breakouts: orthogonal in BOTH timeframe AND mechanism, providing genuine portfolio diversification with minimal signal correlation to existing LINK exposure. Single-dominant-filter design (200-day SMA daily regime + 20-bar 4H Donchian high + volume confirmation), explicitly NOT a multi-condition AND-gate. Calibrated for ~30-50 entries/year — well above the sparsity-failure threshold.

Iteration-2 minimal fix for the Layer-3 'No trades produced' failure. Root cause: the regime filter (self._regime_ok) read from self._daily_close, populated only by on_extra_bar from a separate DAILY extra-bar feed; that feed never populated in the sandbox, so _regime_on() was permanently False and should_enter (which requires breakout AND volume AND regime) never returned a side over all 13,855 bars. The breakout and volume halves already worked, so I changed ONLY the regime computation: _regime_on() now computes the identical 200-day trend filter from the existing 4H close buffer (regime_sma*bars_per_day = 1200-bar SMA), removing the cross-feed dependency. Config drops extra_instruments/extra_bar_types (single feed) and the buffer cap was raised to hold the 1200-bar regime window. Imports, calculate_signal's continuous breakout-margin signal, the volume confirmation, entry/exit/sizing logic, and the long-only BINANCE-futures routing (leverage 1.0) are unchanged, so Layers 1 and 2 stay green while the strategy now produces breakout entries whenever LINK is above its 4H-derived 200-day SMA with a volume-confirmed 20-bar Donchian breakout.

The strategy blew up the account: cagr -100%, max_drawdown 78-100%, avg_position_pct 1,718%, annualized_volatility 4,000%, return_kurtosis 408, with physically impossible daily returns (+5,118% on 2025-08-11, -3,197% on 2025-10-08, -1,471% on 2023-04-26) that reflect equity collapsing toward zero after which fixed-notional positions produce thousand-percent swings. Two compounding failures make it not worth optimizing: (1) the breakout signal itself has NO edge — win_rate 24.7%, profit_factor 0.54, expectancy -$1,258/trade, LINK 4H volume-confirmed breakouts are net-losing; (2) the 95%-of-equity sizing (position_pct 0.95, no volatility scaling) turns that losing signal into a death spiral that bankrupts the book (2025 -211%). The reported Sortino 9.83 / PSR 0.90 / Sharpe 0.61 are meaningless artifacts of the collapsed-equity distribution. Not optimize: a wipeout (cagr -100%, 100% drawdown) has nothing to tune. Not iterate: fixing the sizing alone leaves a net-losing signal (PF 0.54), so this is not a one-line fix that yields a viable strategy — the LINK 4H volume breakout has no edge AND the sizing is reckless. If the BNB-mirror breakout family is retried on LINK, position sizing must be drastically reduced and volatility-scaled (95%-equity with no scaling bankrupts on any losing streak), but the underlying signal showed no edge regardless. FAILURE PATTERN: a 95%-of-equity long-only breakout with no volatility-based sizing on a net-losing signal (PF 0.54, 25% win rate) produces a catastrophic account blowup (cagr -100%, avg position 1,718% of equity, 4,000% vol) as losses drive equity toward zero and percentage returns explode — aggressive fixed-fraction sizing on a no-edge signal is a bankruptcy machine, and the auto-replicated BNB breakout architecture does not transfer to LINK 4H (the signal has no edge here, unlike the proven daily LINK trend strategy).

Implementation

Long-only volume-confirmed Donchian breakout on LINKUSDT.BINANCE USD-M futures, 4H bars, with a 200-day trend-regime filter now derived from the SAME 4H buffer (200 days = 1200-bar SMA) instead of a separate daily feed. Enters LONG when flat and the 4H close breaks the prior 20-bar Donchian high AND volume exceeds 1.5x the trailing 20-bar average AND the latest close is above the 1200-bar (200-day) SMA. Exits on an 8% hard stop, a close below the prior 10-bar low, or regime-off. The signal is the continuous breakout margin (close/prior_high - 1); should_enter thresholds it with the cached volume and regime flags. Single feed, pure OHLCV. leverage 1.0.

Verification Results

Verification failed (Layer 3 — sandbox backtest): No trades produced Bar type used: LINKUSDT.BINANCE-4-HOUR-LAST-EXTERNAL, Bars processed: 13876 Diagnostics: should_enter() returned a side 0 times over 13855 evaluated bars -> your ENTRY CONDITION never triggered. Loosen the entry logic / thresholds. Ensure your strategy produces trades with the given data and parameters.

Backtest Review

Sharpe: 0.61
Total return: -64.27%
Max drawdown: 78.32%
Trades: 85
Win rate: 24.7%
Profit factor: 0.54

Implemented the breakout/volume/regime logic and single-feed ITER-2 fix; produced 85 entries (0 dropped)

Account blowup: cagr -100%, max_drawdown 78-100%, avg_position_pct 1,718%, annualized_volatility 4,000%, return_kurtosis 408 — physically impossible daily returns (+5,118%, -3,197%, -1,471%) reflect equity collapsing to near-zero

No edge in the signal: win_rate 24.7%, profit_factor 0.54, expectancy -$1,258/trade — LINK 4H volume-confirmed breakouts are net-losing

Reckless sizing: 95%-of-equity per position with no volatility scaling turns the losing signal into a death spiral (annual 2025 -211%)

Reported Sortino 9.83 / PSR 0.90 / Sharpe 0.61 are meaningless artifacts of the blown-up, near-zero-equity distribution

Outcome Summary

Aggressive fixed-fraction sizing (95% of equity, no volatility scaling) on a no-edge signal is a bankruptcy machine — once losses drive equity toward zero, fixed-notional positions produce thousand-percent swings — and the auto-replicated BNB breakout architecture does not transfer to LINK 4H, where the signal showed no edge despite LINK's proven daily trend strategy.

The analyst abandoned it at backtest review on two compounding failures — a net-losing signal (PF 0.54, 25% win rate) and reckless 95%-of-equity sizing with no volatility scaling that turned that losing signal into a death spiral (2025 -211%) — concluding a -100% CAGR wipeout has nothing to optimize and that fixing the sizing alone leaves an edgeless signal, so it is not a one-line fix.

A long-only volume-confirmed breakout on LINKUSDT 4H perpetual futures (pure OHLCV), mirroring the proven BNB 4H breakout architecture — entering long on a 20-bar Donchian high break confirmed by volume >1.5x average while above a 200-day uptrend regime (a 1200-bar 4H SMA after the iter-2 single-feed fix), exiting on an 8% hard stop, a trailing structure low, or regime-off, sized at 95% of equity.

It blew up the account: CAGR -100%, max drawdown 78-100% (CI to 100%), avg position 1,718% of equity, annualized volatility 4,000%, return kurtosis 408, with physically impossible daily returns (+5,118%, -3,197%, -1,471%) reflecting equity collapsing toward zero; the signal itself had no edge across 85 trades — win rate 24.7%, profit factor 0.54, expectancy -$1,258/trade — and the reported Sortino 9.83 / PSR 0.90 / Sharpe 0.61 are meaningless artifacts of the collapsed-equity distribution.

Backtest and paper results are hypothetical. Trading involves risk of loss.