NEAR Hyperliquid 4H Fractal Breakouts with Stepped Protective Trailing Stops

Hypothesis

Hyperliquid is under its 20% venue target, and two-way trading addresses the portfolio's 78.7% long-only concentration. Four-hour bars avoid Hyperliquid's short-history subhourly constraint. Corpus concentration is empty, so NEAR's individual share and novelty cannot be verified. This tests the submitted fractal mechanism rather than repeating the listed SMA, Supertrend or funding-reversal rules. Options and cross-venue additions would change the source mechanism and are therefore omitted.

negative_expectancy: The specified NEAR 4-hour adaptation has negative arithmetic net per-campaign expectancy (-0.221% against the +0.15% review floor) over 423 closed campaigns. Positive dollar P&L reflects unequal campaign notionals and concentrated trend winners; it does not establish that the reported average percentage expectancy clears costs. Under the supplied pre-optimization review rule, this baseline does not warrant optimization. This rejects this operationalization and configuration, not the unavailable Fraktrak source implementation or fractal breakouts generally.

Implementation

Two-way confirmed five-bar fractal breakouts on NEAR 4H with intrabar stop entries, capped same-direction additions, and a percentage protective trail ratcheted in 0.10% first-entry-price steps after closed bars.

Implements the distilled research adaptation. Strict fractals confirm after two right-hand bars; stops sit one actual tick beyond the latest unconsumed level. A first-fill anchored campaign budget charges additions their full conservative stop distance without recycling trailing gains. Each addition is limited to 0.25% risk and 25% notional; the campaign is capped at 0.75% committed risk and 75% mark-to-market notional at 1x. Opposing entries are canceled on fill, and only same-direction additions are armed while holding. Protection starts at the actual first fill; completed bars update favorable high/low extrema, excluding the fill bar to avoid pre-entry extremes. Only trail_pct is tunable. Stop gaps can exceed intended loss budgets, and software OCO relies on platform event sequencing.

Verification Results

Optional: use modify_order on the resting entry (new trigger and quantity) instead of cancel-then-replace, so the new level is live in the same bar.

When a new same-side fractal replaces a resting, unfilled entry stop, _arm_fractal cancels the stale order and returns, waiting for the cancel acknowledgment. The replacement at the new level is only placed on the NEXT 4H bar's close, so a breakout of the new fractal during that bar is missed. This is deliberate (it avoids a double order) and does not read future data.

Before paper, rebuild the campaign state from the open position and its resting reduce-only stop in on_start.

Campaign state (held_direction, protective_level, favorable_extreme, risk_committed, used/pending fractal stamps) lives only in instance attributes. After a paper/live restart with a position carried over (SF_FLATTEN_ON_STOP=0), the trail and risk budget would start from zero. This has no effect in a backtest.

No change required; the risk officer should note it.

Stop-market protection fills at the gap price, so the realised loss on a gap can exceed the 0.25% per-addition / 0.75% campaign risk budget. The developer's rationale already discloses this. With 1x leverage and 75% max notional, the worst case stays bounded.

Sandbox numbers point in different directions. Total return is +10.1% with PF 1.16 over 423 closed positions, but avg_trade_return_pct is -0.204% of notional (L200 pattern). Before trusting either number, reconcile the two: per-addition notionals differ, and one large winner of +$4,815 dominates. impact_cost_pct is 41.2 (total impact $7,047 against $3,380 commission), and all of the profit comes from the 'stressed' vol tercile (+16.5%); the calm and normal terciles are both negative. The analyst should check whether the edge survives impact costs outside high-volatility regimes. Win rate 34% with avg win/loss 2.2x fits a trend-breakout profile.

Backtest Review

Sharpe

0.44

Total return

9.26%

Max drawdown

10.73%

Trades

423

Win rate

34.5%

Profit factor

1.15

423 closed campaigns include 225 longs and 198 shorts, consistent with the two-direction breakout adaptation.

Trade P&L sums to $9,259.10, reconciling the 9.259% return with no terminal open position. Entry-notional-weighted return is positive at 0.247%; the negative arithmetic average is not an accounting contradiction.

Mean net campaign return is -0.221% versus the +0.15% Hyperliquid review floor across 423 campaigns. QA's concern about unequal sizing is confirmed: dollar profitability does not establish positive average percentage expectancy.

The largest winner contributes $4,912, approximately 53% of net profit. Calm and normal volatility regimes lose money; stressed conditions supply the gains.

Modeled impact consumes 43.2% of gross P&L. The executed forecast diagnostic provides no independent confirmation of the mechanism: IC 0.0121, t=0.27.

Outcome Summary

Positive dollar profits under unequal sizing do not establish positive average percentage expectancy or justify optimization.

Backtest review issued an abandon verdict because mean net campaign return fell below the +0.15% review floor; unequal campaign sizes and concentrated winners explained the positive dollar result. Optimization and subsequent analyst, risk, and promotion stages were not reached.

Trade confirmed five-bar fractal breakouts in both directions on NEAR Hyperliquid 4-hour bars, using stepped protective trailing stops and capped equity-relative exposure.

The backtest returned 9.259% with Sharpe 0.437, 423 closed campaigns, and maximum drawdown of 10.731%. Arithmetic mean net campaign return was -0.221%, despite positive dollar P&L of $9,259.10.

Analysis

The Fraktrak xonax Expert Advisor based on the fractal breakouts

Analysis

Советник на пробитие фракталов Fraktrak xonax