DOTDailyTimeSeriesMomentumRegimeFilteredLS

DOT USD-M Daily Time-Series Momentum, Long-Short with Long-Term Regime Filter (Single-Instrument BINANCE Perp, Pure OHLCV, Validated-Family Trend Edge Extended to a Decorrelated, Deep-History L1 Major, Wide ATR Stop, Low-Parameter)

A LONG-SHORT, SINGLE-INSTRUMENT, DAILY time-series momentum strategy on DOTUSDT.BINANCE (USD-M perpetual) that extends the factory's ONE empirically-validated, promoted edge — daily trend-following on majors — to a deep-history L1 major the long-short trend book does not yet cover and that adds a decorrelated return stream (DOT's Polkadot/parachain ecosystem drivers differ from the BTC/ETH/SOL/BNB/LINK/AVAX sleeves). It deliberately uses the SAME proven construction as the promoted BTC/ETH/SOL TS-momentum siblings (price-vs-moving-average trend sign) rather than a novel signal, because the repeated lesson of this session is that ONLY the validated daily-trend family survives — every orthogonal variant has either lacked edge (intraday VWAP/taker fades wiped out, multi-day reversal negative) or hit an engine wall (multi-instrument aggregation, options settlement, HL short-history overfit). This is therefore a deliberately conservative, high-survival-probability extension, engineered around every confirmed landmine: SINGLE-INSTRUMENT / SINGLE-VENUE / SINGLE-ACCOUNT (no multi-instrument equity/return aggregation artifact); PURE OHLCV (no funding / OI / liquidations supplementary-data dependency, avoiding the OI cache-gap and funding-accrual issues seen elsewhere); DEEP MULTI-YEAR Binance history (walk-forward + holdout span multiple real regimes, unlike the HL 155-day data wall); DAILY bars (the validated cadence — NOT the intraday-bar breakout/fade variants that are chop-dominated and negative on majors). It is built to PASS where LTC trend FAILED: LTC failed because it is a range-bound laggard that chopped a naked trend-cross (16.8% win rate). DOT, by contrast, has exhibited SUSTAINED DIRECTIONAL REGIMES on BOTH sides (the 2021 parachain-narrative advance and the multi-year 2022-2024 decline) — sustained trends, not chop — which a LONG-SHORT trend book monetizes in both directions, and a LONG-TERM REGIME FILTER keeps it flat during directionless transitions. Low parameter count (~3-4 knobs) to resist overfitting.

Directly extends the factory's one empirically-validated edge -- daily trend-following on majors -- to DOT, a deep-history L1 major the long-short trend book does not yet cover, adding a decorrelated return stream. It deliberately clones the proven AVAX/BTC/ETH/SOL daily-trend construction (price-vs-MA trend sign + 200d rising/falling regime gate + wide ATR stop) rather than a novel signal, because the repeated session lesson is that only the validated daily-trend family survives. It is engineered around every confirmed landmine: SINGLE-INSTRUMENT/SINGLE-VENUE/SINGLE-ACCOUNT (no multi-instrument aggregation artifact), PURE OHLCV (no funding/OI/liquidation cache gaps), DEEP multi-year Binance history (walk-forward + holdout span multiple regimes), and DAILY bars (the validated cadence, not chop-dominated intraday variants). It is built to pass where LTC trend failed: LTC chopped a naked trend-cross because it is a range-bound laggard (16.8% win rate); the sloping-200d-MA regime filter forces this book flat during directionless transitions and only trades sustained DOT regimes on both sides (the 2021 parachain advance and the 2022-2024 decline), which a long-short trend book monetizes in both directions. Low parameter count resists overfitting.

The validated daily-trend family does not extend to DOT — the backtest is a structural loser, not a tuning candidate. Negative expectancy (-$836/trade), profit_factor 0.55, omega 0.76, Sharpe -0.37, and total_return -28.8% over 60 trades spanning 2020-2026 (multiple real regimes). The decisive tell is the long leg: long_win_rate 5.9% (1 win in 17) — this is precisely the LTC whipsaw failure mode (16.8% win rate) the hypothesis claimed DOT plus the 200-day rising/falling regime + slope filter would avoid, and the gate demonstrably did not prevent it; longs are entered in 'confirmed uptrends' and chopped/stopped on DOT's brief, reversing rallies. Shorts fared modestly better (32.6%) on DOT's multi-year decline but not enough to offset. I verified the entry polarity (signal = price-vs-MA50, BUY only above a rising MA200) — there is no code/polarity bug, so the losses reflect a genuine absence of trend edge: DOT trades like the range-bound laggard LTC, not a clean-trending major. Optimization tunes parameters around an edge; with PF 0.55, negative expectancy, and a 1-in-17 long win rate persisting across every regime, there is no robust region to tune toward, and 60 trades would only let the optimizer curve-fit to the few winning shorts. FAILURE PATTERN: porting the promoted BTC/ETH/SOL daily TS-momentum long-short to a non-clean-trending L1 major (DOT) reproduces the LTC whipsaw — a catastrophic long-leg win rate (5.9%), negative expectancy, and net-negative returns across multiple regimes despite a regime/slope chop filter; the daily-trend edge is asset-specific to clean trenders and does not generalize to range-bound laggards, regardless of parameter tuning.

Long-short, single-instrument DAILY time-series momentum on DOTUSDT.BINANCE (USD-M perpetual) using the validated price-vs-moving-average trend family. The continuous signal is (close - 50d SMA)/|50d SMA|. A long-term 200-day regime filter (price above a RISING 200d MA for longs, below a FALLING 200d MA for shorts) gates all entries, keeping the book flat during directionless transitions. Exits on a wide 4*ATR stop, a momentum (signal-sign) flip, or a regime flip (price crossing back through the 200d MA). Pure OHLCV, no supplementary data, single-venue/single-account, ~6 low-overfit knobs. Sizes 50% of equity notional per leg at leverage 1.0.

Clean single-instrument/single-venue/single-account, pure-OHLCV design that avoids the multi-leg and supplementary-data landmines flagged elsewhere

Deep multi-regime history (2020-2026, 60 trades) so the negative result is well-sampled, not a thin fluke

Low fee drag (commission 3.25% of gross) — losses are signal-driven, not cost-driven

Structurally losing: total_return -28.8%, Sharpe -0.37, profit_factor 0.55, omega 0.76, expectancy -$836/trade, win_rate 25%, max_drawdown 61%, 15 consecutive losses

Long leg is broken: long_win_rate 5.9% (1 of 17) — the exact LTC-style whipsaw (16.8% win rate) the hypothesis claimed the regime/slope filter would prevent; the gate did not stop it

No edge across regimes: annual returns net negative and inconsistent (2021 -62%, 2024 -17%, 2025 -21%; only 2022/2026 positive) over 5+ years — the span meant to validate the edge instead invalidates it

DOT behaves like a range-bound laggard (the LTC failure profile), not a clean-trending major; entry polarity checks out, so this is genuine lack of edge, not a fixable bug

DOTDailyTimeSeriesMomentumRegimeFilteredLS extended the promoted BTC/ETH/SOL daily TS-momentum family to DOT, a deep-history decorrelated L1, adding a 200-day rising/falling regime-and-slope gate specifically to dodge the chop that killed the LTC trend extension. The design was clean and well-sampled, but DOT traded like a range-bound laggard: longs were entered in 'confirmed uptrends' and chopped to a 5.9% win rate, producing profit factor 0.55, -$836 expectancy, and a -28.8% net result with a 61% drawdown across multiple regimes. The analyst verified there was no polarity bug and ruled it a genuine absence of trend edge — the family does not generalize to non-clean-trending majors regardless of the slope filter or tuning. It ended after one iteration as abandoned, never advancing to optimization or risk review.

The daily trend-following edge is asset-specific to clean trenders (BTC/ETH/SOL) and does not generalize to range-bound L1 laggards: DOT reproduced the LTC failure (5.9% long win rate, negative expectancy across every regime) and a 200-day regime + slope chop filter did not prevent the whipsaw — screening for a decorrelated narrative is not a substitute for verified trend persistence on the specific asset.

It was abandoned at the pre-optimization backtest-review gate (verdict: abandon) because the validated daily-trend edge does not extend to DOT — a genuine lack of edge (entry polarity verified, no bug) across multiple real regimes with a catastrophic 5.9% long win rate, exactly the LTC whipsaw the regime/slope filter was meant to prevent — so with profit factor 0.55 and no robust region to tune, optimization and all later stages were never reached.

A single-instrument, single-venue long/short daily time-series momentum strategy on DOTUSDT.BINANCE USD-M perps — trading the price-vs-MA50 trend sign, gated by a 200-day rising/falling regime-with-slope filter (long only above a rising MA200, short only below a falling MA200) with a wide ATR stop and momentum/regime-flip exits — extending the factory's validated daily-trend family to DOT as a decorrelated, deep-history L1 major, with the slope gate explicitly added to avoid the LTC chop-whipsaw.

Over 2020-2026 and 60 trades (17 long / 43 short) it was a structural loser: total return -28.8%, Sharpe -0.37, profit factor 0.55, omega 0.76, expectancy -$836/trade, win rate 25%, 61% max drawdown, 15 consecutive losses — with the long leg broken (long win rate 5.9%, 1 of 17) and net-negative inconsistent annual returns (2021 -62%, 2024 -17%, 2025 -21%; only 2022/2026 positive); fees were not the driver (commission 3.25% of gross).