Strategies

Spot Range-Acceptance Retests on BCH, ETC and ATOM Hyperliquid Perps

Outcome: Abandoned

SpotRangeAcceptanceRetests

Outcome Summary

SpotRangeAcceptanceRetests turned an FMZ essay about falsifiable trading decisions into a rule: after a Binance spot breakout on BCH, ETC or ATOM, wait for a confirmed retest of the frozen range boundary before trading the Hyperliquid perp. On 893 days of clean overlapping 4H data it closed 188 trades for -2.12% total return and a Sharpe of -0.26, averaging -0.026% per trade against the 0.15% floor. Most of the drag was modeled impact of $12.6k, which exceeded the roughly +0.42% pre-cost profit. That pre-cost profit came only from 2024 and the high-volatility regime, and ATOM and ETC lost money overall. The analyst abandoned it at backtest review after one iteration, citing negative expectancy and a signal with no measurable predictive power, so optimization and later stages were never reached.

Strategy report

Hypothesis

Derived by the Research Lead from the FMZ essay on falsifiable trading decisions and avoiding additions to losing positions: https://www.fmz.com/bbs-topic/973. The supplied local article was read; it supplies a philosophy of explicit invalidation, not these entry rules. This operationalization tests whether a completed spot breakout followed by a successful later retest of its frozen boundary predicts multi-day continuation. Use synchronized 4-HOUR bars for three independent spot/perpetual pairs. Trade only Hyperliquid perps, both long and short; Binance spot instruments are signal sources. No trade occurs on the initial breakout. Market orders execute at the completed confirmation bar close, the capability registry equivalent of next-bar-open execution on continuous venues. Intrabar protective stops are matched by the platform, with no exchange-hosting requirement. Baselines: range_bars=30, retest_window_bars=12, retest_band_atr=0.35. Predicted frequency is 12-24 completed campaigns per asset/year, not a measured rate. Use all actual overlapping history, with no assumption that Hyperliquid 4H history can be backfilled indefinitely.

The largest direction imbalance is long-only 77.8% versus <=55%, so this symmetric long/short design addresses it while increasing Hyperliquid exposure (11.4% versus >=20%), multi-instrument scope and cross-venue research (8.3% versus >=15%). It does not fill the larger options venue deficit because this proposal tests spot price acceptance rather than option valuation. corpus_concentration was empty; read-only snapshot counts identify BTC as dominant, so BTC and ETH are excluded. Prior-art searches found spot-led Donchian entries and several failed-auction reversal baskets, but no BCH/ETC/ATOM retest contract. The substantive change is entry AFTER a frozen boundary is retested and subsequently accepted, not entry on a breakout or a reversal after its failure; volume-share filters and fixed first-shock entries from recent failures are absent. Scientific history motivates this contract, not a pre-test performance verdict. Binance is signal-only because the test is about spot boundary acceptance; execution stays on the underrepresented Hyperliquid venue.

negative_expectancy. This rejects the Research Lead's operationalization of spot range-acceptance retests on BCH, ETC and ATOM Hyperliquid 4H perps, with Binance spot as the signal source. It does not reject anything in the FMZ essay, which supplies a philosophy and no entry rules. On a reliable sample of 188 closed positions over 893 days, avg_trade_return_pct is -0.026% against the +0.15% Hyperliquid floor. PF is 0.921, Sharpe -0.26 and total return -2.12%. Win rate is 0.32 with a payoff ratio of 1.92. QA's concern is borne out on the full-history ledger, and the notional weighting checks out. Measured gross before costs is about +0.42% of entry notional and about +0.31% after commission and funding, but modeled impact ($12.6k, 151% of gross; capacity $44k) makes it negative. Sizing is fixed by the contract. That pre-impact gross does not show a stable edge: it comes from 2024 and the stressed-vol tercile. 2025 lost $11.0k, the calm and normal regimes both lose, ATOM and ETC are net losers, and the executed IC is -0.047 (t = -0.6). The tunable retest parameters have no information signal to tune toward and do not affect the impact drag. A smaller-size reframe would cut impact but would rest on a regime-dependent, noise-level signal, so I do not recommend one. No accounting or data defect is asserted.

Implementation

Symmetric BCH, ETC and ATOM spot breakout-retest-acceptance campaigns executed only on Hyperliquid perpetuals. Breakout boundaries and 14-bar SMA true-range ATR are frozen; a later range-touch retest must be followed by a separate close beyond the breakout buffer within the retest window. Market entries at the completed confirmation close use the registered continuous-market next-open equivalent. Intrabar OUO brackets protect the mapped boundary-minus/plus-0.5-ATR stop and 3R target; spot close invalidation or 30 synchronized 4H bars also closes a campaign. No additions; six-bar post-close cooldown, at most two concurrent positions.

Implements the Research Lead operationalization, not trading rules from the philosophical essay. Acceptance means a subsequent close beyond the frozen boundary by the specified breakout buffer, no farther than 1.5 frozen ATR. The retest touches the symmetric 0.35-ATR band and closes no farther inside than that band. Stops translate the spot invalidation boundary using the confirmation perp/spot price ratio; sizing risks at most 0.5% of primary Hyperliquid equity and caps each name at 15%, so two names use at most 30% gross. Scalar campaign state persists across restarts; bounded spot histories are indicators only. Every signal leg must share the exact primary close stamp. All fixed values and the three named tunables retain their specified meanings and baselines; no supplementary feeds or history assumptions are introduced.

Verification Results

Sandbox net return is -1.76% across 188 closed positions; modeled impact is 138.5% of gross profit. Average trade return is slightly positive (0.00386% of notional) despite negative aggregate PnL, so examine notional weighting and cost attribution in the full-history ledger before judging economics. These smoke-test results do not establish a correctness failure.

Backtest Review

Sharpe: -0.26
Total return: -2.12%
Max drawdown: 8.10%
Trades: 188
Win rate: 32.4%
Profit factor: 0.92

The mechanism is implemented as written. No entry happens on the initial breakout. Entry comes only after a spot retest of the frozen boundary and acceptance back beyond the buffer, with brackets sized off the invalidation stop.

Trades are balanced: 88 long and 100 short across BCH, ETC and ATOM. The sample is reliable: 188 closed positions over 893 days of full overlapping history, with clean input audits.

Risk is contained: max drawdown 8.1% against the 12% contract constraint, with peak leverage 0.31.

Fee-floor comparison: avg_trade_return_pct is -0.026% against the +0.15% Hyperliquid floor. PF is 0.921, Sharpe -0.26 and total return -2.12%.

Payoff shape: win rate 0.32 with a payoff ratio of 1.92 gives negative expectancy of -$22.5 per trade.

Cost decomposition from the ledger, per $2.67M of entry notional: gross before all costs is about +$11.3k (+0.42%). Commission is $2.2k, funding $0.76k and modeled impact $12.6k. Impact alone is 151% of gross and turns the result negative. Capacity is only $44k against roughly $14k positions. Sizing (risk_pct, max_notional_fraction) is contract-fixed.

The gross edge is regime-concentrated and not stable. Year PnL is 2024 +$5.8k, 2025 -$11.0k and 2026 +$1.0k. Only the stressed-vol tercile is profitable (+5.4%); calm (-4.1%) and normal (-3.1%) both lose. The drawdown has lasted 607 days.

By asset: ATOM loses $5.2k and ETC $0.9k; only BCH is positive (+$1.9k).

The executed forecast IC is -0.047 (t = -0.62), so the signal carries no information. The tunables range_bars, retest_band_atr and retest_window_bars have nothing to tune toward, and none of them changes the impact drag.

Outcome Summary

A pre-cost edge that appears in only one volatility regime and one year, with a signal that is statistically noise, does not justify reworking the strategy at smaller size: the smaller size would cut impact costs but would still rest on an unreliable signal.

The analyst abandoned it at backtest review for negative expectancy, so optimization never ran. Before costs it made about +0.42% of entry notional. Modeled market impact ($12.6k, 151% of the gross profit) turned that negative. The pre-cost profit also came mostly from 2024 and from the high-volatility third of the period, and the signal's information coefficient was -0.047 (t = -0.62), which is indistinguishable from noise.

It used Binance spot 4H breakouts on BCH, ETC and ATOM as the signal, waited for a successful retest of the frozen breakout boundary, and then traded continuation long or short on the matching Hyperliquid perps.

Over 893 days (2024-05-02 to 2026-10-11) it closed 188 positions (88 long, 100 short). Total return was -2.12%, Sharpe -0.26, profit factor 0.921, win rate 32% and max drawdown 8.1%. The average trade returned -0.026% of notional, against the +0.15% Hyperliquid floor.

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