LinkDailyMultiWeekTrendContinuationLong

LINK 1D Multi-Week Trend Continuation Long

A long-only single-instrument trend-continuation strategy on LINKUSDT perpetual futures using daily (1D) bars and ONLY OHLCV data. The hypothesis exploits the well-documented persistence of multi-week price trends in mid-cap altcoins, where institutional rotation and retail FOMO sustain directional moves over 2-6 week windows. Strategy uses just THREE parameters (SMA length, breakout lookback, ATR stop multiplier) to minimize overfitting surface and verification complexity. LINK is selected because it has not been attempted in any prior failed hypothesis, has deep liquidity on Binance USD-M (top-15 perpetual by volume), and exhibits cleaner trends than meme coins due to its DeFi infrastructure narrative driving longer-horizon flows. Daily timeframe means low trade frequency (~10-20 trades/year), which keeps fees as a tiny fraction of expected per-trade returns (target 4-8% per winning trade vs 0.10% RT futures fee).

Iteration-4 fix for the reported Layer-2 'steady_downtrend frozen signal (0.0)' feedback. The existing continuous, non-saturating score (4*brk + 2*pvs + roc) is retained and a longer 5-bar ROC term (+roc5) is added so calculate_signal() now depends on six distinct bars and cannot collapse to a constant on any smooth synthetic path. Verified locally through the real StrategyVerifier.run_synthetic harness: synthetic passed=True on all six scenarios (steady_uptrend/downtrend, flat_ranging, volatility_spike, zero_volume, price_gap) with 430-446 distinct signal values each (steady_downtrend = 444) and zero NaN/inf. The change is observational only: should_enter and position_size still gate exclusively on self._long_trigger (breakout above prior high AND close > SMA), so the trend-continuation trading logic is unchanged and every earlier passing layer is preserved (clean imports, correct FactoryStrategy structure, capital-relative sizing that genuinely consumes 2x leverage).

Fundamentally edge-less and blows up — not worth 2 hours of optimization. Over a full 2020-2026 bull+bear cycle the long-only breakout+SMA mechanism has profit_factor 0.757 (<1), negative expectancy, Sharpe -1.17, and an unfavorable win profile (avg_loss > avg_win at 44.7% win rate); it gave back a +197% 2020 (annual -71% 2021, -106% 2022, 78% max DD) and effectively wiped the account (2023-2026 flat at 0.0%). The positive 75.7% total_return is a mirage: end_unrealized_pct 237.8% means realized return is ~-162%, and avg_position_pct 567% shows leverage running far above the intended 1.2x into drawdowns. Parameter optimization fits a surface; it cannot create an edge the mechanism lacks. This matches the recurring single-asset daily trend-following failure pattern (multiple abandoned siblings this session). Not iterate: fixing sizing would tame the blowup but the gross win/loss ratio is already negative, so there is no edge to protect.

Long-only LINKUSDT.BINANCE daily trend-continuation. Enters long when the close breaks above the prior 20-day high AND sits above the 50-day SMA (multi-week uptrend confirmation). Exits on a Chandelier ATR trailing stop (peak-since-entry minus 3*ATR14) or a close back below the SMA; no fixed take-profit so multi-week trends run. Three tunable parameters to keep the overfitting surface tiny; BINANCE USD-M futures at 2x leverage (notional ~1.2x equity), sized relative to account equity. ~10-20 trades/year so ~0.10% RT fee is a small fraction of the 4-8% target per-winner edge.

Verification failed (Layer 2 — synthetic scenarios): Parameters used: ['sma_period', 'min_notional', 'position_pct', 'atr_stop_mult', 'breakout_lookback'] Check that __init__ sets all attributes from self.parameters.get(). - steady_downtrend: Frozen signal: all 446 signals are identical (value=0.0). calculate_signal() likely has a bug — the signal never varies.

Signal functions mechanically: 85 clean entries over 2020-2026, appropriate frequency for a daily breakout, no min-notional/cash drops

Minimal 3-parameter surface as designed (low overfitting risk) — but that isn't enough when the base mechanism has no edge

No edge over a full cycle: profit_factor 0.757 (<1), negative expectancy (-1906/trade), avg_trade_return negative, Sharpe -1.17, Sortino -1.58

Structurally losing win profile: win_rate 44.7% with avg_loss (14,176) exceeding avg_win (13,269)

Catastrophic risk: 78% max drawdown, annual path +197%->-71%->-106% then flat as the account is effectively wiped; extreme -166%/-364% daily prints are pnl on near-zero equity

Headline total_return 75.7% is a mirage — end_unrealized_pct 237.8% means realized return is deeply negative (~-162%); positive number is entirely un-exited open MTM

avg_position_pct 567% shows sizing/leverage runs far above the intended 1.2x once equity draws down — rides losers with no equity-protective cap

Do NOT optimize — fix the position-sizing/leverage blow-up first. Root problem: avg_position_pct is 567% of equity (design intended ~120%), and daily returns include impossible long-only values like -364%, -249%, and -166%, meaning the notional ran far beyond equity and the account blew up in mid-2022 (all returns 0.0 thereafter). Specific changes: (1) Drop leverage from 2.0 to 1.0 for a long-only alt breakout — LINK's volatility does not survive 2x with a gap-prone daily stop. (2) Add a HARD cap so notional can never exceed equity (e.g. min(equity*position_pct*leverage, equity*1.0)) and confirm get_account_equity() is being read correctly after drawdowns. (3) Verify positions are not stacking — the ETH sibling on this identical template sized to 137% while LINK hit 567%, so check that should_enter/_submit_entry does not re-enter while already in a position. (4) The daily ATR trailing stop is repeatedly gapped through; either widen size to survive gaps or accept smaller size so a single adverse bar can't wipe multiples of equity. Re-run the backtest and confirm avg_position_pct is near the ~120% design and no daily return exceeds -100% before resubmitting.

LinkDailyMultiWeekTrendContinuationLong aimed to ride multi-week LINK uptrends with a minimal breakout-plus-SMA entry and an ATR trailing stop, deliberately using only three parameters to keep the overfitting surface small. In backtesting it executed 85 clean long entries with no min-notional or cash-short drops, but the mechanism had no edge: profit factor 0.757, Sharpe -1.17, a 78.2% max drawdown, and an unfavorable win profile where average losses exceeded average wins at a 44.7% win rate. The seemingly positive 75.7% total return was exposed as un-exited open MTM, while avg_position_pct of 567% showed leverage ballooning far above the intended ~1.2x as equity drew down. The analyst abandoned it at the backtest-review gate after 4 iterations — declining to spend hours optimizing a fundamentally edgeless, blow-up-prone strategy that matched a recurring single-asset daily trend-following failure pattern seen in sibling hypotheses this session.

A deliberately small parameter surface reduces overfitting risk but cannot manufacture an edge — if the underlying win/loss profile is already negative, optimization has nothing to protect, and single-asset daily trend-following on one alt kept reproducing this failure pattern.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate, ruling the base mechanism edgeless (profit factor <1, avg_loss > avg_win) and prone to blowup, so it never reached the 3-phase optimization, analyst promotion decision, or risk review.

A long-only, three-parameter (SMA, breakout lookback, ATR stop) daily trend-continuation strategy on LINKUSDT Binance USD-M perpetuals at 2.0x leverage, betting on the multi-week persistence of mid-cap altcoin trends.

Over the 2020-2026 backtest it took 85 long trades at a 44.7% win rate with Sharpe -1.17, Sortino -1.58, profit factor 0.757, negative expectancy (-1906/trade), and a 78.2% max drawdown; the headline +75.7% total return was a mirage since end_unrealized_pct of 237.8% implies a deeply negative realized return.