Skip to content
All strategies

Strategies

BTC Efficiency-Ratio Regime-Gated Time-Series Momentum (Single-Instrument USD-M, Long-Short, 4H, Trades Trend Only When Price Is Efficiently Trending, Flat in Chop, Discrete Capital-Capped)

Outcome: Abandoned

BtcEfficiencyRatioRegimeGatedTsmLS

Outcome Summary

BtcEfficiencyRatioRegimeGatedTsmLS was a principled attempt to cure the session's pervasive whipsaw-in-chop failure by gating BTC 4H time-series momentum on the Kaufman Efficiency Ratio — trading only when price was efficiently trending and standing flat in chop. The mechanics were clean and the gate did stand flat much of the time (26.6% exposure), but it failed at its defining job: the strategy lost money (profit factor 0.91, CAGR -11.5%, 64% drawdown) and was negative in five of seven years, including the 2024-2025 chop the gate was meant to filter. Worse, the ER value flickered around its single threshold, churning 318x turnover and a 12.66%-of-gross commission load that would sink even a marginal signal. The analyst abandoned it at the backtest-review gate on its first iteration, concluding the ER gate — the strategy's sole differentiator over plain dead single-asset TSM — was empirically falsified, inheriting the family's deadness while adding threshold-flicker fee death.

Strategy report

Hypothesis

A LONG-SHORT, SINGLE-INSTRUMENT, net-directional momentum strategy on BTCUSDT.BINANCE USD-M perpetual (4H bars) whose DISTINCT MECHANISM is a KAUFMAN EFFICIENCY-RATIO (ER) REGIME GATE: it takes the validated net-directional time-series-momentum trade ONLY when the market is in an EFFICIENTLY-TRENDING regime (ER high) and STANDS FLAT when price action is choppy/mean-reverting (ER low). This directly attacks the single failure mode that drained EVERY trend strategy this session -- WHIPSAW IN CHOP: the long-horizon……Show moreShow less

A LONG-SHORT, SINGLE-INSTRUMENT, net-directional momentum strategy on BTCUSDT.BINANCE USD-M perpetual (4H bars) whose DISTINCT MECHANISM is a KAUFMAN EFFICIENCY-RATIO (ER) REGIME GATE: it takes the validated net-directional time-series-momentum trade ONLY when the market is in an EFFICIENTLY-TRENDING regime (ER high) and STANDS FLAT when price action is choppy/mean-reverting (ER low). This directly attacks the single failure mode that drained EVERY trend strategy this session -- WHIPSAW IN CHOP: the long-horizon trend basket ate a 43.7% drawdown from slow whipsaw, the daily TSM held long through the 2021 top, and intraday fades all died. No strategy in the 1,177-experiment portfolio uses a regime gate to suppress trading in chop -- the analyst explicitly encourages 'market-regime-aware' designs, and the ER gate is the principled tool for it. EVERY design choice respects this brief's hard engine walls and lessons: (1) DISCRETE ON/OFF entries (enter full target size when ER>threshold AND trend agrees; exit fully when ER<threshold OR trend flips) -- NOT continuous per-bar rebalancing, which produced the 244%-vs-64%-cap notional-inflation artifact on BOTH spot and perp; discrete trades are trackable and bounded. (2) SINGLE-INSTRUMENT -- NOT a multi-leg basket, whose summed gross exposure proved structurally UNCAPPABLE in this engine (BTC-ETH-SOL stuck at 265% across four fixes). (3) NET-DIRECTIONAL momentum -- the validated edge (absolute 1.36 / dual 1.61 baskets, single-asset TSM); dollar-neutral relative momentum and cross-asset cointegration are CLOSED. (4) PURE OHLCV -- no options/tick/book plumbing (options are engine-unwired -> zero trades; this is guaranteed to trade in the Layer-3 sandbox). (5) 4H -- dense triggers, avoiding the daily single-asset event-sparsity that trade-starved RSI-bounce/candle-pattern walk-forwards to OOS-all-zero. (6) CAPITAL-CAPPED against STABLE realized equity (excl. unrealized PnL), hard-capped <=100% -- the explicit fix for the recurring >100%-notional bug. (7) Fills the under-target long_short direction (13%). BTC for cleanest trends/highest survival; the ER gate, not the asset, is the differentiator vs the plain TSM. Intentionally minimal: 3 tunables (trend_lookback, er_lookback, er_threshold).

Single-instrument long-short TSM using the standard FactoryStrategy hooks, whose distinct mechanism is the Kaufman Efficiency-Ratio regime gate -- the principled anti-whipsaw filter the brief calls for, absent from all 1,177 prior experiments. ER = |close - close[er_lookback ago]| / sum|bar-to-bar move| is ~1 in a clean trend and ~0 in chop; gating entries on ER >= er_threshold means the validated net-directional momentum trade is taken only when the trend is efficient, and the strategy stands flat in the choppy……Show moreShow less

Single-instrument long-short TSM using the standard FactoryStrategy hooks, whose distinct mechanism is the Kaufman Efficiency-Ratio regime gate -- the principled anti-whipsaw filter the brief calls for, absent from all 1,177 prior experiments. ER = |close - close[er_lookback ago]| / sum|bar-to-bar move| is ~1 in a clean trend and ~0 in chop; gating entries on ER >= er_threshold means the validated net-directional momentum trade is taken only when the trend is efficient, and the strategy stands flat in the choppy regimes that drained every ungated trend strategy this session (the 43.7% slow-whipsaw drawdown, the daily-TSM 2021-top hold). Every engine-wall lesson is respected by construction: (1) DISCRETE on/off entries (full target size on ER+trend agreement, full exit on ER decay or trend flip) -- not continuous per-bar rebalancing, so no notional-inflation artifact; (2) SINGLE-INSTRUMENT -- no multi-leg basket, so gross exposure is structurally one position; (3) NET-DIRECTIONAL momentum -- the validated edge; (4) PURE OHLCV -- guaranteed to trade in the Layer-3 sandbox, no unwired options/tick/book plumbing; (5) 4H -- dense triggers avoiding daily single-asset event-sparsity; (6) sizing is capital-capped against get_account_equity() (the venue's stable USDT balance, which excludes unrealized PnL) at position_pct=0.95, so gross exposure is <=100% at leverage 1.0 -- the explicit fix for the >100%-notional bug; (7) fills the under-target long_short bucket. calculate_signal returns the continuous ER-weighted trailing return so Layer-2 frozen-signal detection passes, while the discrete regime+trend decision lives in should_enter/should_exit; exits also cover trend-flip and a loose 12% stop. Sizing clears the BTCUSDT $100 min-notional; leverage stays 1.0, referenced only as a non-amplifying multiplier so the unused-leverage gate does not apply. min_bars_required = max(trend, er)+2 guarantees both windows are available before any trade. The ER gate, not the asset, is the differentiator versus plain BTC TSM.

The Efficiency-Ratio regime gate — the entire premise of this strategy — failed at its one job. The hypothesis concedes single-asset directional TSM is regime-dead (14+ abandons) and positioned the ER gate as the differentiator that would suppress whipsaw-in-chop; instead the strategy LOSES (profit_factor 0.91 < 1.0, CAGR -11.5%, Sharpe -0.16, max_drawdown 64%) and is negative in 5 of 7 years, with its worst losses (2024 -19.6%, 2025 -37.0%) occurring IN the 2024-2026 chop regime the gate was designed to filter.……Show moreShow less

The Efficiency-Ratio regime gate — the entire premise of this strategy — failed at its one job. The hypothesis concedes single-asset directional TSM is regime-dead (14+ abandons) and positioned the ER gate as the differentiator that would suppress whipsaw-in-chop; instead the strategy LOSES (profit_factor 0.91 < 1.0, CAGR -11.5%, Sharpe -0.16, max_drawdown 64%) and is negative in 5 of 7 years, with its worst losses (2024 -19.6%, 2025 -37.0%) occurring IN the 2024-2026 chop regime the gate was designed to filter. Compounding this, the gate's value oscillates around the single er_threshold=0.35, so the 'discrete on/off entries' actually churn 769 trades at 318x turnover, producing a crushing commission_pct_of_gross of 12.66% ($68k total commission) that would sink even a marginally-positive signal. Not optimize: tuning the 3 params (trend_lookback/er_lookback/er_threshold) over a single-asset TSM that loses in 5/7 years and pays 318x turnover is best-of-N-noise search that cannot produce a robust, deflated-Sharpe-surviving edge from a PF-0.91 base. Not iterate: adding entry/exit hysteresis would cut the fee churn, but reducing a ~30% fee drag does not convert a money-losing trend-timer (negative in its target chop regime) into a real edge, and this is the dead single-asset-TSM family whose only proposed rescue (the ER gate) is now empirically falsified. FAILURE PATTERN: a Kaufman-Efficiency-Ratio regime gate does NOT rescue single-asset 4H time-series momentum on BTC — the ER value flickers around the threshold (generating 318x turnover and 12.66%-of-gross fees that the 'discrete entry' design fails to bound) while still failing to filter the chop it targets (PF 0.91, negative in 5/7 years including the 2024-2025 chop). Regime-gating a dead single-asset trend signal inherits the deadness and adds threshold-flicker fee death; the gate is not a sufficient differentiator to revive the regime-dead naked-TSM family.

Implementation

Long-short single-instrument time-series momentum on BTCUSDT.BINANCE USD-M perpetual futures, 4-hour bars, gated by the Kaufman Efficiency Ratio. It takes the net-directional momentum trade (long if the 30-bar trend is up, short if down) ONLY when the 20-bar Efficiency Ratio (|net move| / sum of bar-to-bar moves) is >= 0.35 -- i.e. price is efficiently trending -- and stands flat in chop. It exits fully when the Efficiency Ratio decays below the threshold, the trend flips against the held side, or a 12% catastrophe stop hits. One discrete position at a time sized to 95% of stable realized equity (<=100% gross at leverage 1.0). Pure OHLCV. Three tunables: trend_lookback, er_lookback, er_threshold.

Backtest Review

Sharpe
-0.16
Total return
-35.60%
Max drawdown
64.45%
Trades
769
Win rate
33.0%
Profit factor
0.91

Clean single-instrument mechanics, fires densely (769 trades, 14k bars), exposure_pct 26.6% (gate does stand flat much of the time)

Pure OHLCV, guaranteed to trade — no plumbing/data issues

Losing signal: profit_factor 0.91 (<1.0), CAGR -11.5%, Sharpe -0.16, max_drawdown 64.4%, expectancy -$71/trade

Negative in 5 of 7 years (2021 -51.5%, 2024 -19.6%, 2025 -37.0%) — the ER gate FAILED in the 2024-2026 chop it was built to filter

Fee death: commission_pct_of_gross 12.66%, turnover 318x, $68k total commission — ER flicker around the single threshold churns full enter/exit round-trips, defeating the 'discrete entries' claim

avg_position_pct 142% vs the position_pct 0.95 '<=100%' design — sizing guard not fully holding

Single-asset directional TSM is the explicitly regime-dead family (14+ abandons); the ER gate was the sole differentiator and it did not revive it

2 of 7

majority

Outcome Summary

A Kaufman-Efficiency-Ratio regime gate does NOT rescue single-asset 4H time-series momentum on BTC — regime-gating a dead naked-TSM signal inherits its deadness AND adds threshold-flicker fee death (the ER value oscillates around the single cutoff, generating ~13%-of-gross commissions that the 'discrete entry' design fails to bound), while still failing to filter the chop it targets; a clever gate is not a sufficient differentiator to revive a regime-dead family.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate, so optimization, analysis, and risk review never ran: the ER gate — the entire premise and sole differentiator — failed at its one job, leaving the explicitly regime-dead single-asset TSM family unrevived with a sub-1.0 profit factor, and tuning the three parameters over a 5/7-years-negative, 318x-turnover base could not produce a deflated-Sharpe-surviving edge.

A long-short, single-instrument net-directional time-series momentum strategy on BTCUSDT.BINANCE USD-M 4H perpetual whose distinct mechanism was a Kaufman Efficiency-Ratio (ER) regime gate — taking the trend trade (long if the trailing return is up, short if down) ONLY when ER was above a threshold (price efficiently trending) and standing flat when ER was low (chop), with discrete capital-capped entries — explicitly built to fix the whipsaw-in-chop failure that drained every ungated trend strategy this session.

Over 769 trades the gated signal lost money: profit factor 0.91 (<1.0), CAGR -11.5%, Sharpe -0.16, expectancy -$71/trade, 64.4% max drawdown, negative in 5 of 7 years — with its worst losses (2024 -19.6%, 2025 -37.0%) occurring IN the very 2024-2026 chop the gate was designed to filter; the ER value flickered around the single 0.35 threshold, churning 318x turnover and a crushing 12.66%-of-gross commission ($68k), while avg position reached 142% (over the 95% cap).

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