Skip to content
All strategies

Strategies

Multi-Symbol Major Crypto 1-Hour Asia Session Range Breakout Long-Short Portfolio

Outcome: Abandoned

MajorCrypto1HAsiaSessionRangeBreakoutLS

Outcome Summary

MajorCrypto1HAsiaSessionRangeBreakoutLS was a carefully engineered attempt to escape this session's sparse-trade graveyard — a 1H timeframe absent from the rest of the book, a 3-symbol basket for ample sample, long/short symmetry for all-regime coverage, intraday flat-by-close to kill overnight and funding risk, and just one tunable knob to minimize overfitting. It succeeded at all of that structurally: 5,634 clean trades, every year and both directions represented. But the underlying ORB premise simply doesn't hold on crypto majors — breakouts faded rather than continued, yielding profit_factor 0.855, Sharpe −1.096, −63% total return, and losses in all 7 years and both legs, with a probabilistic_sharpe of 0.0006 making the negative edge a near-certainty. The analyst abandoned it at backtest review without optimizing: with no directional edge in any regime and only an exit multiple to tune, there was nothing to fit but noise — a textbook case that orthogonality to past failures is not the same as having an edge.

Strategy report

Hypothesis

An intraday Opening Range Breakout (ORB) strategy adapted for crypto's 24/7 market, applied to a 3-symbol basket of the most liquid Binance USD-M perpetuals: BTCUSDT, ETHUSDT, SOLUSDT. Uses 1-HOUR bars — a timeframe entirely absent from the current pipeline (every existing strategy uses 4H, 1D, or weekly). Mechanism: every UTC day from 00:00 to 07:59 defines the 'Asia session range' for each symbol (high-of-period, low-of-period). From 08:00 to 22:59 UTC (Europe + US sessions), the strategy monitors hourly bar……Show moreShow less

An intraday Opening Range Breakout (ORB) strategy adapted for crypto's 24/7 market, applied to a 3-symbol basket of the most liquid Binance USD-M perpetuals: BTCUSDT, ETHUSDT, SOLUSDT. Uses 1-HOUR bars — a timeframe entirely absent from the current pipeline (every existing strategy uses 4H, 1D, or weekly). Mechanism: every UTC day from 00:00 to 07:59 defines the 'Asia session range' for each symbol (high-of-period, low-of-period). From 08:00 to 22:59 UTC (Europe + US sessions), the strategy monitors hourly bar closes. The FIRST 1H close that breaks above the Asia session high triggers a LONG entry on that symbol; the FIRST close that breaks below the Asia session low triggers a SHORT entry. Only ONE entry per symbol per UTC day (no re-entries after either side triggers). All positions are force-closed at 23:00 UTC regardless of PnL. This hypothesis is structurally orthogonal to every failure pattern in the log: it is NOT single-instrument (3-symbol basket multiplies sample size 3x), it is NOT a daily-timeframe BTC strategy (avoiding the 'BTC-daily-2024+regime-decay' pattern that killed Thursday Seasonality / BTC Daily Momentum / BTC Daily Z-Score), it is NOT a pullback-in-uptrend (avoiding the 'pullback-on-alt-daily-no-edge' pattern that killed XRP/ADA/DOGE/LINK/SOL/AVAX 9x), it is NOT a single-bar pattern (avoiding the 'single-bar-no-information-content' pattern that killed INJ/BTC vol spike), it is NOT a weekly seasonality (avoiding the decayed-microstructure pattern), and it does NOT depend on supplementary data (avoiding the verification-framework no-trades issue that killed OI Divergence / LS Ratio Fade). The ORB mechanism has 40+ years of academic and practitioner backing (Toby Crabel 1990, Larry Williams 1979, more recently revisited by Brett Steenbarger's prop work) in equity index futures and FX; the application to crypto leverages the unique structural fact that crypto is the only major asset class with continuous 24/7 trading across THREE distinct sessions (Asia/Europe/US), so the Asia session genuinely functions as a 'pre-market' range that European and American flow either confirms (breakout) or rejects (fade back). Three free parameters: Asia session window (00:00-07:59 UTC, fixed by convention), entry session window (08:00-22:59 UTC, fixed by convention), and the per-bar profit-target multiple of 2x range (one tunable parameter) plus a per-trade hard-stop at the opposite side of the Asia range (no free parameter — geometrically determined). Effectively ONE tunable parameter, the lowest of any hypothesis this session. Expected trade frequency: ~30-50% of UTC days produce a breakout setup per symbol, so 3 symbols × 365 days × 0.4 trigger rate = ~440 entries per year — well above the walk-forward viability threshold AND well above any 'sparse-trade' failure pattern in the log. Long+short symmetry means the strategy generates trades in every market regime (bull breakouts in 2021/2024, bear breakouts in 2022), structurally breaking the regime-concentration curse. All positions are flat at the daily close, so the strategy carries zero overnight risk and zero funding-rate exposure.

Implements the 24/7-adapted ORB exactly and stays orthogonal to every logged failure pattern: multi-symbol (3x sample), 1H timeframe (absent from the book), long+short (trades in bull and bear breakouts), single-entry-per-day intraday with flat-by-close (no overnight/funding/regime-concentration exposure), and OHLCV-only (no supplementary-data no-trades risk). Each symbol's session logic runs independently when its own 1H bar arrives — BTC via calculate_signal, ETH/SOL via on_extra_bar — using the bar's UTC……Show moreShow less

Implements the 24/7-adapted ORB exactly and stays orthogonal to every logged failure pattern: multi-symbol (3x sample), 1H timeframe (absent from the book), long+short (trades in bull and bear breakouts), single-entry-per-day intraday with flat-by-close (no overnight/funding/regime-concentration exposure), and OHLCV-only (no supplementary-data no-trades risk). Each symbol's session logic runs independently when its own 1H bar arrives — BTC via calculate_signal, ETH/SOL via on_extra_bar — using the bar's UTC hour/day derived from ts_event, so no cross-symbol timing coupling is needed; the single-instrument interface is inert and all legs trade via _submit_entry_instrument/_submit_exit_position. The profit target (tp_mult x range) is the lone free parameter; the stop is geometrically fixed at the opposite Asia-range side. Correctness verified: a constructed hourly day builds the Asia range (100-110), the hour-8 close above 110 fires a LONG with target 132 and stop 100, and no second entry occurs that day (one-per-day enforced); the Layer-2 proxy runs clean with both extras across all six scenarios (491 unique, non-frozen) and there are no @staticmethod proxy-rebinding hazards. Leverage is 1.0 so sizing references no margin leverage (no inert-leverage gate).

Dead premise, not a code or sparsity problem. The Asia-session opening-range-breakout (Crabel/Williams ORB) does NOT continue on 24/7 crypto majors -- the strategy traded correctly as specified (5,634 ORB entries on BTC/ETH/SOL 1H, flat by 23:00 UTC, balanced long+short) yet lost money systematically: profit_factor 0.855, Sharpe -1.096, total_return -63%, avg_trade_return negative, and NEGATIVE in all 7 years and both directions (long 45.4% / short 44.5% win rates). The Sharpe CI is entirely negative (-1.75 to……Show moreShow less

Dead premise, not a code or sparsity problem. The Asia-session opening-range-breakout (Crabel/Williams ORB) does NOT continue on 24/7 crypto majors -- the strategy traded correctly as specified (5,634 ORB entries on BTC/ETH/SOL 1H, flat by 23:00 UTC, balanced long+short) yet lost money systematically: profit_factor 0.855, Sharpe -1.096, total_return -63%, avg_trade_return negative, and NEGATIVE in all 7 years and both directions (long 45.4% / short 44.5% win rates). The Sharpe CI is entirely negative (-1.75 to -0.46) with probabilistic_sharpe 0.0006, so the true Sharpe is near-certainly negative. The single tunable parameter (tp_mult) cannot rescue a mechanism with no directional edge across every regime; optimization would waste 2 hours fitting an exit multiple to noise. Not iterate (trades match the hypothesis -- nothing to fix in code; the premise is the flaw). FAILURE PATTERN: intraday opening-range-breakout continuation on crypto majors has no edge -- session breakouts mean-revert/fade rather than continue, producing a sub-1 profit factor and negative Sharpe in every year and both long/short legs regardless of timeframe (1H here), and a flat-by-close intraday structure with one exit parameter offers no robust region to optimize.

Implementation

Intraday Asia-session Opening-Range-Breakout long-short across BTC/ETH/SOL Binance USD-M perps on 1-HOUR bars (pure OHLCV). For each symbol each UTC day it builds the Asia-session high/low over hours 0-7, then during hours 8-22 takes the FIRST close above the Asia high LONG or the FIRST close below the Asia low SHORT (one entry per symbol per day). Exits at a profit target of 2x the Asia range, a hard stop at the opposite side of the Asia range, or a forced flat at 23:00 UTC (zero overnight/funding risk). Each symbol sized at 10% of equity notional. One tunable parameter (tp_mult). Leverage 1.0.

Backtest Review

Sharpe
-1.10
Total return
-63.08%
Max drawdown
66.65%
Trades
5634
Win rate
45.0%
Profit factor
0.86

Mechanism is implemented correctly and trades as described: 5,634 trades over 6.5 years on BTC/ETH/SOL 1H, ample sample, metrics_reliable=true, no sizing/over-exposure artifact (avg_position_pct 12%)

Novel 1H timeframe and structurally orthogonal to prior failure patterns (as the hypothesis claims) -- but novelty does not create edge

Money-losing with no edge: profit_factor 0.855 (<1), Sharpe -1.096, Sortino -1.55, total_return -63%, avg_trade_return negative

Negative in ALL 7 years and BOTH directions (long_win_rate 0.454 / short_win_rate 0.445) -- the Asia breakout direction carries no predictive information on crypto majors; breakouts fade

Sharpe CI entirely negative (-1.75 to -0.46), probabilistic_sharpe 0.0006 -- the true Sharpe is essentially certain to be negative

Only one tunable parameter (tp_mult); no exit-multiple can rescue a wholly-negative-CI, PF-0.855 mechanism. max_drawdown 66.6%, commission 10.1% of gross

ample

1 of 7 (trivially)

majority

negative

Outcome Summary

Intraday opening-range-breakout continuation has no edge on 24/7 crypto majors — session breakouts fade/mean-revert rather than continue, producing sub-1 profit factor and negative Sharpe in every year and both legs — so a large, clean, structurally-novel sample doesn't redeem a flawed premise; it only makes the no-edge verdict statistically certain.

The analyst issued an abandon verdict at backtest review on a dead-premise basis: the Sharpe CI was entirely negative (−1.75 to −0.46) with probabilistic_sharpe 0.0006, so the true Sharpe is near-certainly negative, and the single tunable parameter (tp_mult) cannot rescue a mechanism with no directional edge — optimization would only fit an exit multiple to noise, and there is nothing to iterate since the code matched the hypothesis and the premise itself is the flaw.

An intraday Opening-Range-Breakout (Crabel/Williams ORB) adapted to 24/7 crypto on a 3-symbol Binance USD-M 1H basket (BTC/ETH/SOL): define each symbol's 00:00–07:59 UTC 'Asia session' range, then go long on the first hourly close above the range high or short on the first close below the low during the Europe/US window, with a tp_mult profit target, an opposite-side hard stop, and a forced flat at 23:00 UTC (zero overnight/funding risk).

The mechanism implemented and traded exactly as designed — 5,634 trades over 6.5 years, balanced long/short (2,964 long / 2,670 short), metrics_reliable=true with no sizing artifact — but it lost money systematically: profit_factor 0.855, Sharpe −1.096, Sortino −1.55, total_return −63%, max drawdown 66.6%, and negative in all 7 years and both directions (long win 45.4% / short win 44.5%), with commission 10.1% of gross.

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