LTC Spot Icebreaker: Bounded Average-Cost Recovery

Hypothesis

Test a cost-aware adaptation of 发明者量化-小小梦's Icebreaker mechanism on LTCUSDT.BINANCE_SPOT using 1-HOUR bars. Start long, add below average entry, and exit the entire inventory above its updated average. Preserve the source's average-price targeting formula and absence of direction reversal. Explicit adaptations are percentage price distances, capital-relative sizing, bounded additions, a campaign-wide loss limit, and decisions executed at completed-bar closes. This is not a faithful reproduction of the original one-second resting-order execution. Claimed source performance is unverified.

The document contains one distinct trading mechanism. Spot respects its explicit rejection of futures and fills an underrepresented venue bucket. Adding shorts, options or cross-venue hedges would change the source mechanism. LTC avoids the crowded BTC/ETH research focus, although the supplied corpus_concentration is empty and cannot establish LTC's actual share. This proposal differs from the existing rising short grid by using long spot inventory and additions sized to move the weighted average by a specified amount.

negative_expectancy: decline optimization of this frozen LTC hourly bounded average-cost adaptation. The substantial closed-campaign sample loses after modeled costs because larger losing inventories outweigh frequent smaller recoveries. This judges the specified adaptation, not the original source strategy or all average-cost mechanisms.

Implementation

Long-only LTC spot campaigns on completed hourly closes. Start at 8% exposure, add below the average entry using the source's weighted-average targeting formula, allow at most four additions, and cap exposure at 20%. Close the entire inventory at 2% above its updated average, a fee-aware 1% campaign loss limit, or 168 hourly bars; wait 24 bars before restarting.

The consistency checker validates fixed research values against config.parameters, so leverage=1.0 is now explicitly included there as well as at config.leverage. All fixed research values and the exact frozen optimization plan are included. Trading logic remains unchanged; only the class docstring changes to avoid a byte-identical source submission. This preserves the requested completed-close adaptation rather than reproducing the original resting-order execution. The outbox file could not be written because the environment permits read-only filesystem access.

Verification Results

The sandbox lost 10.07% over 288 closed campaigns despite a 79.5% win rate: average losses were approximately 4.6 times average wins. Full-history review should examine recovery frequency and campaign-loss tails; this does not establish a coding defect.

Backtest Review

Sharpe

-0.32

Total return

-22.12%

Max drawdown

26.62%

Trades

1334

Win rate

80.5%

Profit factor

0.92

The detail file confirms 1,334 closed long-only campaigns, consistent with the intended inventory direction; no open-position gain inflates the result.

Average trade return is 0.8555%, above the supplied 0.25% spot fee floor; this is not a below-floor fee-edge rejection.

Net campaign PnL totals -22,118.03 USDT over 3,176 days: return -22.12%, profit factor 0.9232 and Sharpe -0.317.

QA's payoff concern is borne out: 80.51% wins average +247.55 USDT, but losses average -1,107.66 USDT, approximately 4.47 times larger.

Losing campaigns have mean recorded leverage of 19.79%, versus 10.70% for winners. Larger losing inventories explain how a positive unweighted average trade percentage coexists with negative dollar expectancy.

All research parameters are frozen and the tunable set is empty; no verified code defect or permitted parameter improvement supports spending on optimization.

For context, the sampled buy-and-hold curve returns approximately -81.20% with 93.00% drawdown. Lower losses than that benchmark do not establish positive campaign expectancy.

Iteration History

Verification failed (Layer 1.5 — hypothesis/config consistency) [class=hypothesis_mismatch]: - optimization_plan fixes leverage at 1.0

Outcome Summary

Evaluate inventory-weighted dollar expectancy because frequent wins and a positive unweighted average trade return can coexist with overall losses.

Backtest review issued an abandon verdict because larger losing inventories produced negative dollar expectancy after modeled costs. Parameters were frozen with no tunable set or verified code defect supporting optimization; optimization and later reviews were not reached.

It tested long-only LTC spot recovery campaigns on hourly closes, adding below average entry within exposure and loss limits and exiting above the updated average.

Across 1,334 closed campaigns over 3,176 days, net return was -22.12%, Sharpe was -0.317, and maximum drawdown was 26.62%. Despite an 80.51% win rate, average losses of 1,107.66 USDT outweighed average wins of 247.55 USDT, producing a profit factor of 0.9232.

Analysis

破冰者（原版）代码浅析，如有错误请指正。