LtcVolNormalizedDonchianBreakoutLS

LTC Volatility-Normalized Donchian Channel Breakout, Trend-Filtered Long-Short (LTCUSDT.BINANCE USD-M, 12H Bars, Enter on a New N-Bar Channel Extreme Aligned With a Slow Absolute-Trend Filter, ATR Trailing Exit, 3-Parameter)

A LONG-SHORT, single-instrument, pure-OHLCV TREND-FOLLOWING BREAKOUT on LTCUSDT.BINANCE USD-M perpetual (12H bars). Trend-following is the ONLY mechanism family that has survived to paper stage in this factory (BTC and ADA dual-timeframe momentum confluence). Rather than re-skin that exact 4H+1D moving-average-confluence recipe (which overfits when ported to noisier alts), this expresses the same underlying edge — directional persistence after a genuine regime shift — through a DIFFERENT, mechanically simpler signal: a Donchian channel breakout. The entry fires on a NEW N-bar high/low (a real range expansion / new information event), which produces LARGE per-trade captures that clear fees, unlike the sub-fee mean-reversion, VWAP, opening-range, funding-gated, OI-gated and liquidation-fade classes the meta-learner has closed (L70/L71/L80/L83/L85/L89). Instrument choice is deliberate diversification away from the saturated BTC/ETH/SOL/ADA names: LTCUSDT is a top-tier-liquid, 2020-listed major with full multi-year history that has been barely tested in this factory, reducing best-of-N overfit selection pressure. Only 3 parameters (channel length, trend-filter length, ATR trailing multiple) to fight the overfit deaths that killed the more heavily-parameterized trend variants (BtcEfficiencyGatedChannelBreakout, BtcTrendPullbackContinuation).

Implements the hypothesis exactly: a mechanically simple range-expansion trend entry (new N-bar Donchian extreme) aligned with a slow absolute-trend filter, ATR trailing exit, and a deliberately tiny tuned surface — channel_length, trend_length, atr_trail_mult (declared in _param_bounds together with the structural atr_period/threshold/risk clamps the code enforces). The signal is continuous and varies every bar (channel position), with the breakout decision thresholded in should_enter at |signal| >= 1, so nothing is gated behind a constant. Only one conjunct sits on top of the breakout (the trend filter), which keeps firing frequency high — an offline replay of this exact logic on the catalog's LTCUSDT 12H history produces ~85-125 trades over 6.5 years across the parameter box, well clear of the zero-trade sandbox risk, with per-trade captures of several percent that dwarf the ~0.10% round-trip taker cost. Honest caveat for the Research Lead/Analyst: in that same offline replay the mechanism's PnL on LTC is roughly flat-to-negative over most of the parameter box (only very wide trailing multiples, ~5-6 ATR, come out positive), so I would not expect this instrument to carry the trend-persistence edge as strongly as BTC/ADA did — I have implemented it as specified and left the trail multiple's box wide enough (2.5-6.0) for the optimizer to reach that region rather than pre-selecting it.

Negative expectancy on a reliable 79-trade / 6.5-year sample: profit_factor 0.759 (<1), total_return -7.08%, avg_trade_return_pct -1.04% (below ZERO, far under the 0.15% USD-M fee floor), expectancy -$312/trade, Sharpe -0.02. Commission is only 1.93% of gross, so this is gross-negative signal, not fee-fragility. The trades correctly implement the hypothesis (breakout aligned with trend filter, balanced 38L/41S, ATR chandelier), so this is a fair test of the premise — and the premise is falsified on the carrier instrument: the strategy loses WORST in the stressed vol tercile (Sharpe -0.71, -25.2%), the exact regime a breakout trend-follower is supposed to exploit, and loses in 4 of the last 5 calendar years. QA's edge concern is confirmed on full history: the developer's own replay found LTC flat-to-negative across most of the parameter box with positivity only at extreme ~5-6 ATR trailing multiples (box-edge = overfit, not a robust region). Optimizing 3 parameters cannot invert a PF-0.759, below-zero-expectancy signal — there is no positive gross edge to tune toward. Diversification-away-from-BTC/ADA onto a weaker carrier instrument did not produce an edge. abandon_class=negative_expectancy.

Long/short Donchian channel breakout on LTCUSDT.BINANCE USD-M perpetual, 12H bars. Every bar it computes a continuous volatility-context decision variable: the close's position inside the PRIOR channel_length-bar high/low range, rescaled to [-1, +1] (>+1 = the bar closed above the prior N-bar high, <-1 = below the prior N-bar low). It goes long on a new N-bar high while price is above its slow SMA(trend_length) absolute-trend filter, short on a new N-bar low while below it, and exits with a chandelier ATR trailing stop (close atr_trail_mult ATRs back from the best high/low reached since entry) — the only exit, so winners run and losers are cut at roughly one trailing-stop width. Size risks risk_pct of equity over the chandelier stop distance, hard-capped at max_notional_frac of equity; leverage 1.0 (futures venue is required only because the strategy trades both directions).

If the base's own warmup gating is relied upon elsewhere, confirm the override does not shorten the intended warmup; otherwise leave as-is.

min_bars_required(self) overrides a framework method (Layer-1 base_shadow warning). This appears deliberate (returns a sensible warmup of ~102 bars covering channel_length+1, trend_length, atr_period+1), and the SMA/ATR/channel guards in should_enter independently prevent premature signals, so it is not a correctness risk. Flagged for awareness only.

Sandbox is a losing unoptimized draw (total_return -5.8%, Sharpe ~0.0, PF 0.94, avg_trade_return_pct -0.97%, 78 trades). The developer's own honest note (config rationale) confirms an offline replay shows LTC flat-to-negative across most of the parameter box, with only very wide trailing multiples (~5-6 ATR) turning positive; regime breakdown shows the strategy loses specifically in the stressed vol tercile (-24.9%, Sharpe -0.70). This is diversification-away-from-BTC/ADA with an admittedly weaker carrier instrument. The analyst should verify the optimizer can actually reach the ~5-6 ATR trail region and that the edge is not confined to a single window before spending full optimization; if the positive region is only the box edge, this is likely an overfit/no-edge death rather than a promotable trend strategy. Advisory only — does not block.

Clean, minimal 3-parameter trend-following design; trades correctly implement the hypothesis (Donchian breakout aligned with trend filter, balanced 38 long / 41 short, ATR chandelier exit)

Reliable, decisive sample: 2361 days of full LTC history, 79 trades, metrics_reliable=true

Negative expectancy on the full sample: profit_factor 0.759 (<1), total_return -7.08%, avg_trade_return_pct -1.04% (BELOW ZERO, far under the 0.15% Binance USD-M fee floor), expectancy -$312/trade, Sharpe -0.02, Sortino -0.02, CAGR -4.54%

Commission is only 1.93% of gross — this is NOT fee-fragility, the breakout signal is gross-negative on LTC

Loses specifically in the regime the hypothesis targets: stressed vol tercile Sharpe -0.71 / -25.2% return; losing calendar years in 2022 (-15.0%), 2023, 2024, and 2025 (-19.0%)

QA edge concern borne out: the developer's own offline replay shows LTC flat-to-negative across most of the parameter box, with positivity only at extreme ~5-6 ATR trailing multiples (box-edge = overfit, not a robust region)

LtcVolNormalizedDonchianBreakoutLS-51f36af783

A trend edge that survived on BTC/ADA does not transfer to a weaker carrier like LTC — a PF<1, below-zero signal can't be rescued by tuning 3 parameters.

Abandoned at the backtest-review gate — a gross-negative signal (commission only 1.93% of gross) with no edge to optimize; optimization never ran.

A long-short Donchian channel breakout on LTCUSDT.BINANCE USD-M 12H bars, trend-filtered with an ATR chandelier exit, aiming to capture directional persistence after regime shifts on a less-tested major.

Net-negative over 2361 days / 79 trades: total return -7.08%, Sharpe -0.02, PF 0.759, avg_trade_return_pct -1.04% (below the 0.15% fee floor); worst in the stressed vol tercile.