Strategies

LTC Net-Directional Time-Series Momentum (Single-Instrument USD-M, Long-Short, 4H, Discrete Trend-Flip Entries, Capital-Capped, Pure OHLCV)

Outcome: Abandoned

LtcFourHourTimeSeriesMomentumNetDirectionalLS

Outcome Summary

LtcFourHourTimeSeriesMomentumNetDirectionalLS was the third run of the sanctioned single-asset TSM template, deliberately placed on LTC — the lowest-beta, cleanest-trending, least-mania-concentrated uncovered major — to maximize generalization odds and dodge the session's 2021-concentration holdout killer. The asset choice changed the symptom but not the disease: the signal lost money (profit factor 0.917, CAGR -28.9%, four losing years, -60% total, 86% drawdown), and the discrete entry re-fired while a position was open, stacking exposure to 104% vs the intended 50% and printing physically-impossible +94% single days. The analyst abandoned it at the backtest-review gate on its first iteration, noting this template has now failed on every asset tried — BNB and XRP liquidated, LTC merely bled because of its lower beta. The recorded conclusion was that the template is a broken implementation of a dead signal, asset selection cannot manufacture an edge, and the capital-cap claim was once again falsified by realized exposure.

Strategy report

Hypothesis

A LONG-SHORT, SINGLE-INSTRUMENT net-directional TIME-SERIES MOMENTUM strategy on LTCUSDT.BINANCE USD-M perpetual (4H bars) -- the next instance of the factory's SANCTIONED single-asset TSM series (BTC/ETH/SOL/BNB/XRP already in pipeline), applied to the BEST-SURVIVING decorrelated major still uncovered. RATIONALE FOR ASSET + HORIZON: (1) LTC is a top-tier-liquidity major with full clean history (since 2020 on Binance USD-M) and a DISTINCT driver -- a PoW payment/'digital-silver' coin -- so its trends are decorrelated from the L1/smart-contract/exchange majors (BTC/ETH/SOL/BNB) already covered, adding a genuine diversifying return stream (XRP covers the payment/legal angle; LTC adds the lower-beta PoW-payment angle). (2) CRITICALLY, LTC is chosen over higher-variance uncovered alts (DOGE/ADA/AVAX) because the analyst's dominant holdout-killer this session is MANIA/2021-CONCENTRATION (DOGE funding = '2021 retail-mania beta'; cross-sectional momentum holdout-failed on 2021-concentration). LTC is the LOWEST-BETA, CLEANEST-TRENDING, LEAST-MANIA-CONCENTRATED liquid major left -> the best generalization/holdout odds of the uncovered set. (3) LONG-SHORT fills the under-represented direction bucket (13.1%). This is the validated, analyst-endorsed edge ('reversion is dead; redirect to net-directional momentum/trend'); single-asset TSM is exactly that, distinct from the closed basket/dollar-neutral/funding-confirmed variants. EVERY design choice respects this session's hard engine walls: (a) SINGLE-INSTRUMENT -- NOT a multi-leg basket, whose summed gross exposure proved structurally UNCAPPABLE in this engine. (b) DISCRETE full-size enter/exit on trend flips -- NOT continuous per-bar rebalancing, which produced the 244%-vs-cap notional-inflation artifact. (c) PURE OHLCV bars -- NOT raw trade-ticks, which hit the hard ~2-day sandbox-memory wall (data_days=2); NOT options (engine-unwired -> zero trades); NOT funding (data ends 2024). Pure 4H OHLCV has full history and is guaranteed to load and trade in the Layer-3 sandbox. (d) 4H dense triggers -- avoids the daily single-asset event-sparsity that trade-starved breakout/RSI/candle walk-forwards to OOS-all-zero. (e) CAPITAL-CAPPED against STABLE realized equity (excl. unrealized PnL), hard-capped <=100%. Per the brief's 'favor simpler strategies' guidance, intentionally minimal: 3 tunables (trend_lookback, entry_band, stop_atr_mult) with robust defaults to avoid a trend_lookback cliff.

Implements the analyst-sanctioned single-asset TSM (net-directional momentum, 'reversion is dead -> redirect to trend') on LTC, the lowest-beta, least-2021-mania-concentrated uncovered major, for the best holdout/generalization odds and a decorrelated PoW-payment return stream. It uses the standard FactoryStrategy hooks on a single instrument, which structurally sidesteps the engine walls this session documented: no multi-leg basket (gross-exposure-uncappable), no continuous per-bar rebalancing (the 244% notional-inflation artifact -- here entries are discrete and full-size on trend flips), no trade-ticks (the 2-day sandbox-memory wall), no funding/options data. Direction is the sign of the trailing-return momentum with a hysteresis entry_band so the position holds through small counter-moves and only flips at the opposite band, cutting whipsaw on 4H bars; calculate_signal returns the continuous momentum so Layer-2 frozen-signal detection passes while the discrete band decision lives in should_enter/should_exit. The exit is let-trends-run: a wide 4x-ATR chandelier trailing stop that ratchets from the peak (long) or trough (short) since entry, plus the opposite-band momentum flip (exit-and-reverse). Sizing is capital-capped against get_account_equity(), which reads the venue's stable (USDT) balance and excludes unrealized PnL, so a single position at position_pct (50%) is hard-capped at max_notional_frac (100%) of equity at leverage 1.0 -- no exposure inflation. Clears the LTCUSDT $5 min-notional; 4H triggers fire densely enough to populate walk-forward OOS windows (avoiding the daily single-asset event-sparsity). leverage stays 1.0, referenced only as a non-amplifying cap multiplier, so the unused-leverage gate does not apply. Three minimal tunables with robust defaults (trend_lookback well away from a short-lookback cliff) per the simplicity guidance.

Third instance of the single-asset discrete-entry TSM template (after BNB and XRP, both of which liquidated), and it fails the same two independent ways. (1) LOSING SIGNAL: profit_factor 0.917 (<1.0), CAGR -28.9%, Sharpe -0.12, expectancy -$120/trade, losing badly in 4 of 7 years (2021 -33.5%, 2022 -47.8%, 2023 -29.9%, 2025 -55.6%, 2026 -64.9%) for -60.3% total and an 85.8% max drawdown — naked single-asset time-series momentum has negative gross expectancy on LTC just as on BNB/XRP, and the 'lowest-beta, cleanest-trending, least-mania major' asset-selection thesis does not rescue it. (2) OVER-EXPOSURE/STACKING BUG: avg_position_pct is 103.6% versus the intended position_pct of 0.50 (~2x the cap), and the book prints physically-impossible single-day returns (+94.55% on 2026-06-12, +71.25% on 2026-05-30 — each would require LTC to move ~190%/140% intraday at 0.50 notional), with return_kurtosis 149; the discrete entry re-fires while a position is open and stacks size, the same defect that drove the BNB/XRP siblings to liquidation. LTC's lower beta turned the blowup into a -60% bleed rather than a -100% liquidation, but the disease is identical. Not optimize: a PF-0.917, -29%-CAGR signal whose series contains impossible bars cannot be optimized. Not iterate: fixing the stacking bug would only surface a money-loser in the regime-dead naked-TSM family (the BNB/XRP runs already established this template is not worth a dev cycle; LTC's PF is no better). FAILURE PATTERN: the single-asset discrete-entry net-directional TSM template fails on EVERY asset tried (BNB liquidated, XRP liquidated, LTC -60%) — the signal is PF<1.0 (regime-dead naked trend) AND the discrete entry re-fires while a position is open, over-exposing to 2-6x the stated cap with impossible prints. Asset selection (low-beta vs mania-alt) only changes whether it liquidates or bleeds; it does not create an edge. A 'capital-capped <=100%' docstring claim is again falsified — realized avg_position_pct is 2x the intended 50%.

Implementation

Long-short single-instrument net-directional time-series momentum on LTCUSDT.BINANCE USD-M perpetual futures, 4-hour bars, pure OHLCV. Direction is the sign of the trailing 60-bar (~10-day) return with a 2% hysteresis band: it goes long when momentum is above +2%, short when below -2%, and holds the discrete full-size position until momentum flips to the opposite band (exit-and-reverse) or a wide 4x-ATR chandelier trailing stop (ratcheting from the peak/trough since entry) is hit. Single position sized at 50% of stable realized equity, hard-capped <=100% at leverage 1.0. Three tunables: trend_lookback, entry_band, stop_atr_mult.

Backtest Review

Sharpe: -0.12
Total return: -60.33%
Max drawdown: 85.83%
Trades: 739
Win rate: 35.5%
Profit factor: 0.92

Pure OHLCV, fires densely (739 trades), full clean LTC history — no data-abort; did not liquidate

Diversification intent (low-beta PoW-payment major) is reasonable in principle

Losing signal: profit_factor 0.917 (<1.0), CAGR -28.9%, Sharpe -0.12, expectancy -$120/trade; losing badly in 4 of 7 years (2021 -33.5%, 2022 -47.8%, 2023 -29.9%, 2025 -55.6%, 2026 -64.9%)

Over-exposure/stacking bug: avg_position_pct 103.6% vs the intended position_pct 0.50 (~2x the cap) — same defect as the BNB/XRP siblings, milder

Physically-impossible prints: +94.55% (2026-06-12), +71.25% (2026-05-30) on a 0.50-notional LTC position; return_kurtosis 149, annualized_volatility 68%

total_return -60.3%, max_drawdown 85.8% — bled (not liquidated) only because LTC is lower-beta

~50 (position_pct 0.50)

<~15 (single LTC, pos 0.50)

Outcome Summary

The single-asset discrete-entry net-directional TSM template fails on EVERY asset tried (BNB liquidated, XRP liquidated, LTC -60%) — the signal is PF<1.0 (regime-dead naked trend) AND the discrete entry over-exposes to 2-6x the stated cap with impossible prints; asset selection (low-beta vs mania-alt) only changes whether it liquidates or bleeds, it does not create an edge, and a 'capital-capped ≤100%' docstring claim is again falsified by realized avg_position_pct.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate, so optimization, analysis, and risk review never ran: naked single-asset TSM has negative gross expectancy on LTC just as on BNB/XRP (the regime-dead naked-trend family), the discrete entry re-fires while a position is open and stacks size beyond the stated cap, and a return series with impossible bars cannot be optimized — fixing the bug would only surface a PF-0.917 money-loser.

A long-short, single-instrument net-directional time-series momentum strategy on LTCUSDT.BINANCE USD-M 4H perpetual that set its side by the sign of the trailing 60-bar return (long above a +entry_band, short below a -entry_band), holding discrete full-size positions until a momentum flip or a wide ATR chandelier trailing stop, claiming a capital cap of ≤100% — the third instance of the sanctioned single-asset TSM template, deliberately on LTC as the lowest-beta, least-mania-concentrated uncovered major for the best holdout odds.

It failed the same two ways as the BNB and XRP siblings: a losing signal (profit factor 0.917 <1.0, CAGR -28.9%, Sharpe -0.12, expectancy -$120/trade, losing in 4 of 7 years, -60.3% total, 85.8% max drawdown) plus the over-exposure/stacking bug (avg_position_pct 103.6% vs the intended 0.50, ~2x the cap) with physically-impossible prints (+94.55% on 2026-06-12, kurtosis 149) — it bled rather than liquidating only because LTC is lower-beta.

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