Strategies

AVAX Intraday Large-Impulse-Bar Continuation, Long-Short (Binance USD-M, 1H, High-Vol L1 Asset-Match to SOL Winner, Clamped Impulse Threshold, Tight Initial Stop, Trailing-Stop Winners, Discrete Capital-Capped)

Outcome: Abandoned

AvaxImpulseBarContinuationTrendLS1H

Outcome Summary

This strategy ported the factory's promoted impulse-bar-continuation trigger to AVAX 1H perps, chosen as the best untested high-volatility L1 match to the SOL winner and hardened against the prior LINK overfit with a clamped impulse_mult, frozen strong_frac, and only four tunable parameters. The mechanism fired cleanly across 1,455 trades, but the edge was negative net of fees: profit factor 0.964, total return -67.4%, Sharpe 0.095, and a 92.1% max drawdown with a near-liquidation late-period blow-up. The analyst abandoned it at the backtest-review gate, noting it was profitable only in the 2021/2022 bull before decaying sharply through 2023-2026 — behaving exactly like the LINK/XRP whipsaw failures the hypothesis itself named as the family's boundary. The outcome was deemed an unfixable asset-selectivity problem, since best-of-N tuning would only refit the dead bull market and fail forward.

Strategy report

Hypothesis

A LONG-SHORT, single-instrument, pure-OHLCV intraday CONTINUATION strategy on AVAXUSDT.BINANCE USD-M perpetual using 1-HOUR bars — the next per-asset instance of the factory's VALIDATED and PROMOTED impulse-bar-continuation family (ETH-1H ~1.40 promoted, SOL-1H ~0.69 paper), applying the EXACT validated single-impulse-bar trigger (NOT my ORB/sustained-drive/portfolio variants) to the untested asset selected by the now-decisive asset discriminator. The family's boundary is fully mapped: it WORKS on ETH and SOL (high-volatility, high-liquidity, clean-flow L1 majors whose intraday thrusts are large enough to clear fees and directional enough to continue) but FAILS on BTC (fee-bound-zero — moves too small to clear fees across USD-M/HL/CoinM), LINK, and XRP (whipsaw). AVAX is the best untested match to the SOL winning profile: a high-beta L1 'ETH-alternative' with high realized volatility (so intraday impulse moves are LARGE — the property that makes the fee math work, exactly what low-vol BTC lacked) and deep Binance USD-M history (so the walk-forward is well-populated, unlike the ~7-month HL intraday wall). It deliberately AVOIDS every dead/blocked path proven this session: NOT BTC (consistently fee-bound, abandoned across 3 venues — stop porting it), NOT a funding/basis/carry strategy (the entire family is fee-dominated/event-sparse — a few bps vs ~19-30 bps round-trip), NOT cross-venue (fee floor), NOT relative/cross-sectional momentum (decayed), NOT a daily-trend basket (correlated with promoted AbsoluteMomentum), NOT options (engine), NOT reversion (dead), NOT sub-1H (fee-dead), NOT HL intraday (data wall). To pre-empt the LINK overfit failure (its impulse_mult was left unbounded and the optimizer fled to a sparse corner), it bakes in the SOL-sibling's hardening: impulse_mult CLAMPED to [1.0, 1.8], strong_frac FROZEN, <=4 tunable parameters, flat-between-signals. Engine-safe: discrete enter-once/exit-once, single position, hard-capped notional, no continuous rebalance/vol-scaling.

Implements the hypothesis as the next per-asset instance of the VALIDATED impulse-bar-continuation family (ETH-1H promoted ~1.40, SOL-1H paper ~0.69). It reuses the proven SOL sibling's structure (continuous signed impulse score returned every bar, tight initial stop + ATR trailing stop asymmetry, discrete enter-once/exit-once single position, equity-relative hard-capped sizing) and applies the EXACT single-impulse-bar trigger to AVAXUSDT — the best untested match to the SOL winning profile (high-beta high-vol L1 with deep Binance USD-M history, so impulse moves are large enough to clear the ~10-11bps round-trip, unlike fee-bound BTC). LINK-overfit hardening is baked in: impulse_mult is hard-clamped to [1.0,1.8] so the optimizer cannot flee to a sparse high-threshold corner, strong_frac is frozen, and only 4 parameters are tunable (impulse_mult, atr_period, trail_atr_mult, max_hold_bars). It avoids every dead path noted in the hypothesis: not BTC, not funding/basis/carry, not cross-venue, not cross-sectional, not options, not reversion, not sub-1H, not HL-intraday. Routed to BINANCE USD-M futures because the strategy goes short; leverage left at 1.0 and referenced in sizing so the config/engine agree (no leverage_set_but_unused risk).

No edge on AVAX: profit_factor 0.964 (<1) and negative expectancy over 1455 trades net of fees, Sharpe 0.095 with CI straddling 0, plus a 92.1% max drawdown and a late-period equity blow-up (kurtosis 1115, avg_position_pct 112%). The mechanism fires correctly (1455 signaled = 1455 submitted), so this is not a code bug — it is a genuinely negative, regime-decayed edge: profitable only in the 2021/2022 bull (+98%/+61%) then -51.6%/-15.8%/-86.9% across 2024-2026. This is exactly the whipsaw failure the hypothesis itself names as the impulse-continuation family's boundary (works on ETH/SOL, fails on LINK/XRP); AVAX is behaving like the failures, an asset-selectivity problem optimization cannot fix — best-of-N tuning would only refit the dead 2021-2022 bull and fail forward. Not worth 2 hours of optimization.

Implementation

Long/short single-instrument intraday CONTINUATION strategy on AVAXUSDT.BINANCE USD-M perpetual using 1-HOUR bars. Each completed 1H bar is scored continuously as impulse = body/ATR (signed). When that thrust is abnormally large (|impulse| >= impulse_mult, clamped to [1.0,1.8]) AND clean (body fills >= strong_frac of the bar's range), the strategy enters WITH the impulse (long on a large clean up-bar, short on a large clean down-bar). Risk is managed with a tight initial ATR stop, an ATR trailing stop that ratchets to let winners run, and a max-hold time cap. Discrete enter-then-exit-to-flat, single position, sized off stable realized equity and hard-capped at 40% of equity notional (leverage 1x).

Backtest Review

Sharpe: 0.10
Total return: -67.36%
Max drawdown: 92.07%
Trades: 1455
Win rate: 33.5%
Profit factor: 0.96

Healthy trade count (1455 trades) and clean entry diagnostics (1455 signaled = 1455 submitted, no min-notional/size drops) — the signal mechanism itself fires correctly, so this is not a coding/no-trade bug

Engine-safe discrete design with the LINK-overfit hardening (clamped impulse_mult, frozen strong_frac, 4 tunables) as intended

Negative edge: profit_factor 0.964 (<1), total_return -67.4%, negative expectancy/avg_trade, Sharpe 0.095 with CI straddling 0 (-1.46 to 0.61) — loses money net of fees over 1455 trades

Catastrophic risk: max_drawdown 92.1%, recovery_factor -0.87, 905-day drawdown duration; late-period equity blow-up (return_kurtosis 1115, daily prints of +112%/+537%, avg_position_pct 112%) indicates near-liquidation, not a tradeable curve

Severe regime decay: profitable only in 2021 (+98%) and 2022 (+61%), then -2% / -51.6% / -15.8% / -86.9% across 2023-2026 — the impulse-continuation edge has died on AVAX

Matches the family's own predicted failure boundary: the hypothesis says continuation WORKS on ETH/SOL but FAILS via whipsaw on LINK/XRP; AVAX is behaving like the whipsaw failures, i.e. asset-selectivity

Outcome Summary

The impulse-bar-continuation edge is asset-selective and regime-decayed — AVAX behaved like the family's known LINK/XRP whipsaw failures rather than the ETH/SOL winners, so a high-volatility L1 profile alone does not guarantee the trigger works, and tuning would only refit the dead 2021-2022 bull.

It was abandoned at the pre-optimization backtest-review gate with an 'abandon' verdict, failing every criterion (Sharpe 0.095<0.5, max_drawdown 92.1%, total_return -67.4%, profit factor 0.964, negative avg trade return); the analyst, risk, and optimization stages were never reached.

A long-short, single-instrument intraday continuation strategy on AVAXUSDT.BINANCE 1H perps applying the factory's validated impulse-bar-continuation trigger — entering in the direction of an abnormally large, clean-bodied (ATR-normalized) impulse bar with a tight initial stop and trailing stop — betting AVAX's high realized volatility would match the profile that worked on the promoted ETH and paper SOL instances.

Over 1,577 days it fired 1,455 trades (725 long / 730 short) but produced a negative edge: total return -67.4%, profit factor 0.964, 33% win rate, negative expectancy (-$59.78/trade), Sharpe 0.095 (CI -1.46 to 0.61), and a 92.1% max drawdown with a late-period equity blow-up (avg position 112%); it was profitable only in 2021 (+98%) and 2022 (+61%) then sharply negative across 2023-2026.

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