SuiFourHourNBarLowerLowReversalLong

SUI 4H N-Bar Lower-Low Reversal Long

A minimal long-only single-instrument mean-reversion strategy on SUIUSDT perpetual futures using 4-hour bars and ONLY OHLCV data (no funding, no L/S ratios, no liquidations, no pairs) to keep verification simple and deterministic. The hypothesis: SUI is a high-beta altcoin (avg 4H true range ~2-3%) that exhibits frequent intraday flush-and-recover patterns — panic selling drives price to a new N-bar low, but if that same bar closes bullishly (close > open) it signals absorption by buyers and a high-probability tactical bounce. The strategy waits for two simultaneous conditions on a closed 4H bar: (1) the bar's low prints a new 20-bar low, AND (2) the same bar closes green. It then enters long at the next bar's open with a tight ATR-buffered stop and a 1.5R fixed target, plus an 18-bar (3-day) time stop. Holding period is short (median ~6-12 bars), per-trade move target is ~2-3% which comfortably exceeds the ~0.10% round-trip futures fee. Trade frequency expected ~6-10 setups per month on SUI 4H — sparse enough to avoid fee drag, frequent enough for statistical significance over a 12-month backtest. Three parameters total (lookback=20, target_R=1.5, max_bars=18) — minimal overfitting surface area.

Implements the hypothesis exactly with its three headline parameters (lookback=20, target_R=1.5, max_bars=18) plus a minimal ATR config for the stop. The edge — a new N-bar low that closes green signalling buyer absorption — is a genuinely binary pattern, so should_enter reads stored flags (_armed) rather than a fragile threshold, while calculate_signal still returns a continuous, varying conviction score (break depth + greenness in ATR units) to satisfy the 'continuous signal' rule and give the verifier a non-frozen signal. The R-multiple framework is fully wired: the stop is captured at the trigger bar (one ATR below its low), the 1.5R target and 18-bar time stop are enforced in should_exit, and position_size risks 1.5% of equity over that exact R distance (capped at 30% notional), so the ~2-3% per-trade target comfortably clears the ~0.10% round-trip futures fee. Long-only with no shorts matches the bullish absorption thesis and the empirical refutation of crypto-major shorts. Kept on SUIUSDT.BINANCE USD-M futures 4H as specified (perpetual, fee-cheaper round-trip than spot), single-instrument with no supplementary data — deterministic verification and no multi-instrument staleness risk; leverage=1.0 since sizing uses no margin amplification, avoiding the leverage_set_but_unused gate.

Dead premise with no edge -- not a code or sparsity problem. The 'fade new N-bar lows that close green' absorption-reversal traded correctly as specified (103 long entries on SUI 4H) but LOSES money: profit_factor 0.929 (<1), Sharpe -0.29, Sortino -0.37, total_return -4.36%, expectancy negative, and negative in 3 of 4 years (2024 -0.2, 2025 -4.4, 2026 -1.0). The new-low-green-close signal carries no predictive information -- avg_win ($1,371) ~ avg_loss ($1,288) with win_rate 0.466, so it bleeds net after fees -- and sharpe_ci_low -2.19 / probabilistic_sharpe 0.378 mean the true Sharpe is more likely negative than positive. The 3 tunable parameters (lookback, target_R, max_bars) cannot convert a sub-1 profit factor into an edge; optimization would fit noise on a negative-edge signal. Not iterate (the code is correct and matches the hypothesis; the premise is the flaw). FAILURE PATTERN: single-bar 'flush-and-recover' / new-low-green-close mean-reversion fades on a high-beta crypto alt (SUI) have no edge -- the absorption signal does not predict a tradeable bounce, yielding a symmetric payoff (avg_win ~ avg_loss), sub-1 profit factor, and net-negative returns across regimes, the same no-edge signature as the Keltner/RSI(2)/Bollinger fade family on majors.

Long-only single-instrument mean-reversion on SUIUSDT.BINANCE perpetual 4H bars using only OHLCV. On each closed bar it checks two simultaneous conditions while flat: (1) the bar's low prints a new 20-bar low, and (2) the same bar closes green (close > open) — a flush-and-recover absorption signal on a high-beta altcoin. It then enters long at market (~next-bar open) with a stop one ATR(14) below the signal bar's low, a fixed 1.5R take-profit, and an 18-bar (3-day) time stop. calculate_signal returns a continuous reversal score (low-break depth in ATRs plus bar greenness) every bar so the signal varies; the binary setup is captured in an _armed flag checked by should_enter. Position size = equity * 1.5% risk / R (capped at 30% of equity notional), tying size to the actual stop distance.

Implemented correctly and trades as specified: 103 long-only new-low-green-close reversal entries, clean entry_diag (103 signaled = 103 submitted, 0 dropped), metrics_reliable=true, adequate sample

No edge / money-losing: profit_factor 0.929 (<1), Sharpe -0.29, Sortino -0.37, total_return -4.36%, expectancy -48.9 (negative)

Signal carries no information: avg_win ($1,371) ~ avg_loss ($1,288), win_rate 0.466 -> net-negative after fees

Negative in 3 of 4 years (2024 -0.2, 2025 -4.4, 2026 -1.0); sharpe_ci_low -2.19 (CI centered negative), probabilistic_sharpe 0.378

~0.5+ (pre-opt)

~62 WF floor

1 of 4

majority

negative (-48.9)

SuiFourHourNBarLowerLowReversalLong was a deliberately minimal bet — just three parameters — that SUI's high beta would make new-20-bar-low bars that close green into reliable 'absorption' bounces. The code did exactly that across 103 trades, but the signal turned out to carry no information: avg_win and avg_loss were nearly identical, win_rate sat at 46.6%, and after fees the strategy bled to profit_factor 0.929, Sharpe −0.29, and a −4.36% return that was negative in 3 of 4 years. With the Sharpe CI centered below zero and only three knobs to turn, the analyst abandoned it at backtest review without optimizing — there is no parameter that converts a coin-flip signal into an edge. It extends the session's mean-reversion-fade graveyard from the majors (Keltner, RSI(2), Bollinger) to a high-beta alt, confirming that flush-and-recover absorption fades simply don't predict bounces in crypto.

Single-bar 'flush-and-recover' / new-low-green-close mean-reversion fades on a high-beta crypto alt have no edge — the absorption signal does not predict a tradeable bounce, producing a symmetric payoff (avg_win ≈ avg_loss), sub-1 profit factor, and net-negative returns across regimes — the same no-edge signature as the Keltner/RSI(2)/Bollinger fade family on majors, now confirmed on an altcoin.

The analyst issued an abandon verdict at backtest review on a dead-premise basis: the new-low-green-close absorption signal carries no predictive information (symmetric win/loss, sub-1 profit factor, net-negative after fees), the Sharpe CI is centered negative so the true Sharpe is more likely negative than positive, and the 3 parameters cannot convert a sub-1 profit factor into an edge — the code matches the hypothesis, so the premise itself is the flaw and there is nothing to iterate.

A minimal long-only, OHLCV-only mean-reversion on SUIUSDT.BINANCE 4H bars (3 parameters): wait for a closed bar that simultaneously prints a new 20-bar low AND closes green (close > open) — read as buyer 'absorption' of panic selling on a high-beta alt — then enter long at the next open with an ATR-buffered stop, a 1.5R fixed target, and an 18-bar (3-day) time stop.

It implemented and traded correctly with an adequate sample — 103 long entries, clean entry diagnostics, metrics_reliable=true — but had no edge and lost money: profit_factor 0.929, Sharpe −0.29, Sortino −0.37, total_return −4.36%, expectancy −48.9, win_rate 46.6%, and negative in 3 of 4 years (2024 −0.2%, 2025 −4.4%, 2026 −1.0%). The payoff was symmetric — avg_win $1,371 ≈ avg_loss $1,288 — with sharpe_ci_low −2.19 and probabilistic_sharpe 0.378.