AtomDailyMultiWeekTrendContinuationLong200Sma

ATOM Daily Multi-Week Trend Continuation Long with 200-SMA Regime Filter

A long-only single-instrument trend-continuation strategy on ATOMUSDT perpetual futures using daily bars and OHLCV-only data. Architecturally mirrors the highly successful LINK 1D Multi-Week Trend Continuation (Sharpe 2.55, promoting stage), applied to ATOM (Cosmos) for portfolio diversification across L1 ecosystem narratives. ATOM has 5+ years of clean Binance USD-M perp data and exhibits classic Cosmos-ecosystem cycle dynamics: extended consolidations punctuated by multi-month rally phases driven by inter-blockchain communication (IBC) growth, ATOM 2.0 upgrades, staking yield expansions, and Cosmos Hub revenue mechanisms. Crucially, ATOM is structurally different from BTC/ETH/SOL/BNB (which are dominated by exchange/institutional flows) and from LINK (oracle-narrative driven) — it captures a separate macro-driver basket (Cosmos / IBC / Tendermint chains). Single-dominant-filter design (200-day SMA + EMA stack momentum), explicitly NOT a multi-condition AND-gate (the recurring failure mode of AVAX Calm-Regime / DOT Pullback / SOL TOM, all abandoned for sparse-signal + outlier-dependency reasons). Calibrated for 5-12 entries/year (the empirically validated success zone of LINK Multi-Week Trend), using the same proven '200-SMA regime + EMA-stack momentum + clear continuation trigger' construction that produced Sharpe 2.55 on LINK. Uses ONLY OHLCV — no funding rates (which can structurally disappear, killing BTC Negative Funding), no proxies (which produce bait-and-switch failures), no calendar anomalies (which decay in 24/7 markets). ATOM has not been used in any current or pending pipeline strategy.

Architecturally mirrors the promoted LINK Multi-Week Trend Continuation (Sharpe 2.55) on ATOM, diversifying the portfolio into the Cosmos/IBC L1 macro-driver basket (distinct from BTC/ETH/SOL/BNB exchange-flow assets and LINK's oracle narrative); ATOM is unused elsewhere in the pipeline. Single-dominant-filter design (200-SMA regime gate) plus a clear continuation trigger (fresh 20-day-close-high breakout) and an ATR trailing exit — explicitly NOT a multi-condition AND-gate, the failure mode the hypothesis warns against. Verified by trace: entry requires close above the 200-SMA AND a fresh breakout (rejecting below-SMA, no-breakout, and stale-breakout cases), and the ATR Chandelier trailing stop computes the correct exit level. One deliberate robustness upgrade over the reference: rather than returning 0.0 on no-entry (which would freeze the signal in synthetic flat/down scenarios where the entry never fires and fail Layer-2 frozen-signal detection), calculate_signal returns a continuous regime+ROC proxy that varies every bar while the boolean entry decision is read separately by should_enter, so it can never leak a false entry. Position-state bookkeeping tracks the high-water mark and bars-held by detecting the open position each bar. Quantity precision is delegated to the engine's make_qty (no hardcoded step size). OHLCV-only — no funding (which can structurally vanish), no proxies, no calendar anomalies. Venue BINANCE USD-M futures per the proven ATOMUSDT.BINANCE-1-DAY path; long-only, leverage 1.0; the ATR trailing stop lets winners run while ~0.10% round-trip fees are negligible against multi-week trend legs. Calibrated for ~5-12 entries/year, the validated success zone of the LINK template.

Net-loser single-name trend follower with no edge and too few trades to optimize. Profit_factor 0.83 (<1.0), total_return -5.22%, expectancy -$476/trade, and Sharpe ~0.01 with CI [-0.625, 0.646] straddling zero — the mechanism has negative expectancy, not a tunable one. It loses in five of seven years (2020/2021/2022/2025/2026), and the two positive years don't offset the rest. The promoted-LINK trend-continuation construction does not transfer to ATOM. It is a single-name OHLCV trend-follower with PF < 1 — the zero-survivor trend-clone class (L9), and the third ATOM trend-follower reviewed in a row (Supertrend, Parabolic SAR) with the same dead-single-name profile. With only 38 trades in 6.6 years the sample is too thin to fit parameters against; the optimizer would curve-fit the handful of outlier winners (kurtosis 36.5) and fail OOS/holdout. No parameter tuning flips a negative-expectancy, sub-1.0-PF trend follower positive. Not worth 2 hours of optimization.

Long-only ATOMUSDT.BINANCE daily multi-week trend-continuation strategy, OHLCV-only. Enters long when close is above the 200-day SMA (long-term regime gate) AND closes above the prior 20-day close-high (fresh continuation breakout, where the prior bar had not already exceeded that level). Exits primarily via an ATR Chandelier trailing stop (close < highest-high-since-entry - 3xATR14) that ratchets up with new highs to let multi-week trends run, plus +30% take-profit, -8% stop-loss, and a 60-bar time-stop safety net. Sizes 30% of equity notional per trade. Leverage 1.0.

CLEAN RESTART 2026-09-04 — this run's verdict history and learning records were removed and it was restarted from verification. Its previous abandonment came from the pipeline, not from the market: the Layer-2 harness mis-bound @staticmethod helpers (fixed), QA issued terminal performance verdicts on an unoptimized smoke test (removed — QA now judges correctness only), and sandbox timeouts came from backtest-slot starvation (fixed). The hypothesis and the strategy code are unchanged. Verify the code on its merits; performance is decided later by the full backtest and the optimizer.

Compute the breakout level as of the PRIOR bar (max of closes ending at bars[-3], i.e. bars[-(n+2):-2]) and test bars[-2].close against THAT, or delete the guard since flatness already blocks re-entry.

The 'fresh continuation breakout' guard is dead code. brk = max(prior 20 closes) INCLUDES bars[-2], so `if bars[-2].close > brk: return False` is always False and never rejects. It doesn't implement the stated intent (reject a breakout the prior bar already exceeded). Harmless — being flat already prevents consecutive-new-high re-entry, and the core entry (close>SMA200 AND close>prior-20-high) is correct — but misleading.

Clean OHLCV-only construction, no supplementary-data dependency or coverage wall; deep price history (6.6 years)

Single-dominant-filter design (avoids the multi-condition AND-gate sparse-signal failure mode); clean execution (38 signaled, 38 submitted, no drops)

Net loser: profit_factor 0.83, total_return -5.22%, expectancy -$476/trade, win_rate 0.24

Sharpe ~0.01 with CI [-0.625, 0.646] straddling zero — no edge; information_ratio -0.28

Loses in 5 of 7 years (2020, 2021, 2022, 2025, 2026 all negative); only 2023/2024 positive

Only 38 trades in 6.6 years — too thin to fit parameters; return_kurtosis 36.5 (outlier-dependent)

Single-name OHLCV trend-follower with PF < 1 — zero-survivor trend-clone class (L9); third ATOM trend-follower in a row with the same dead profile

The LINK Sharpe-2.55 mechanism does not transfer to ATOM

data_unavailable: backtest data for ATOMUSDT.BINANCE could not be collected after 5 retries: - bars (bar_type=ATOMUSDT.BINANCE-1-DAY-LAST-EXTERNAL): got 0, needed 50, or 90+ days span The data layer (not the strategy) is the blocker — re-evaluate this hypothesis once collection is fixed / the symbol is backfilled.