AvaxBinanceDualTimeframeMomentumConfluenceLS

AVAX Binance USD-M Dual-Timeframe Momentum Confluence — Long-Short 4H Trend Aligned with the Daily Trend, Flat on Disagreement, ATR Trailing Exit (Single-Bar, 3-Parameter)

A LONG-SHORT, single-instrument, pure-price MOMENTUM-CONFLUENCE trend follower on AVAXUSDT.BINANCE (USD-M perpetual), 4H single-bar construction. This is a deliberate FAIR-TEST of a premise the factory has already flagged as promising-but-unproven-for-venue-reasons. My trail cleanly separates two groups: (a) choppy laggards XRP (Sharpe 0.28), DOGE (0.45, underperforms buy-hold), LTC (net-LOSING -36.5%) — FAIRLY tested on 6+ Binance years, genuinely NO edge; versus (b) strong-trending majors tested only on Hyperliquid's ~3.4y and abandoned SOLELY for insufficient history despite EXPLICITLY praised edge. AVAX-HL is the clearest example of (b): its reviewer wrote 'edge is genuinely distributed (kurtosis 6.57, skew 0.36, not an outlier mirage), positive years (2024 +11.5%, 2025 +42.4%), 13.5% drawdown, PF 1.57, decorrelated alpha' — killed only because HL's short single-regime window can't clear the significance gate. Binance USD-M has ~5y of AVAX history, which is exactly the fix that 'optimized' the ADA/BNB ports (they passed backtest review) and gives AVAX's demonstrated distributed edge a fair, significance-clearing test. AVAX is a strong-trending, high-beta major (wide per-trade fee margin), NOT a range-bound laggard. It uses the proven confluence mechanism behind the paper-stage BTC survivor (Sharpe 1.99). NOT a mean-reversion fade (L53), NOT a squeeze breakout (L54), NOT a channel breakout (outlier class), NOT a cross-sectional/ensemble basket (L52), NOT a market-neutral ratio (near-zero edge), NOT a NON-PRICE-feed gate (L46), NOT options/COIN-M (L50/L51). Fills the long-short gap (13.6% vs 86% long-only). To resist overfit (#1 killer, 308x) and verification bugs: single bar type, 3 params, fixed fast-EMA. Risk profile: ~1.5% equity risk/trade via ATR stop; vol-scaled notional = risk_budget / (atr_mult * ATR_4H) capped at 25% of equity*leverage; 2x leverage cap (reads self.config.leverage). Exactly 3 tunable parameters (slow/fast EMA ratio, daily-trend EMA length, ATR trailing multiple).

Implements the hypothesis exactly: the proven confluence mechanism (4H momentum must AGREE with the daily-horizon trend, flat on disagreement, ATR trail runs winners) given the fair, significance-clearing Binance test that AVAX's promising Hyperliquid run never got -- 12857 4H bars (2020-09-23 to 2026-08-05, ~5.9 years) versus HL's 3.4. Single bar type as specified (no extra_instruments/extra_bar_types, so the positional-resolver feed drop that cost the ETH port two iterations cannot occur); exactly 3 tunables with the fast EMA fixed at 12 and code clamps mirrored verbatim in _param_bounds; defaults are the exact CENTRE of every declared bound. Leverage 2.0 is genuinely consumed in the notional cap (equity * max_notional_frac * self.config.leverage), a fraction of CURRENT equity so losses cannot compound (L15). Layers 1-3 verified locally: static clean, all six synthetic scenarios pass with a continuous varying signal, and the real 365-day sandbox runs 172 trades (86 entries) with no zero-trade, timeout or sizing issues. THE VERDICT FROM THE FAIR TEST IS SPLIT, and the Analyst owns the call. Full sample (replaying THIS EXACT CLASS, 0.10% round-trip fees): 437 round trips (~75/yr, dense enough that walk-forward windows and a 15-day holdout populate), avg +1.58% per trade NET of fees (10x the 0.15% viability floor), profit factor 1.56, win rate 26%, kurtosis 20.1 with top-3 trades only 15.2% of gross profit -- so the HL reviewer's 'genuinely distributed, not an outlier mirage' judgement is CONFIRMED on long history, and yearly avg/trade is positive 2021 +7.38%, 2022 +1.28% (bear), 2023 +3.30%, 2024 +0.95%, 2025 +0.19%. BUT the recent regime is negative and that is where the OOS gates live: the trailing 365 days replay at -0.48%/trade and the real sandbox year returns -17.4% (Sharpe -0.97, PF 0.62, avg -0.49%/trade, DD 22.5%), with 2026 YTD -1.07%/trade. This is not a defaults artefact -- EVERY cell of the 3x4x3 parameter grid I swept is positive full-sample (avg +0.80% to +2.09%, PF 1.32-1.71) and EVERY cell is negative over the trailing 365 days (-0.18% to -1.10%/trade), so the decay is a property of AVAX's recent chop, not of the parameterisation. Read together with the SOL sibling (same mechanism, same venue: full-sample +1.37%/trade but trailing-year ~flat), the pattern is that the alt trend edge was real through 2023 and has faded across the alt complex since; if the factory wants this book, BTC/ETH remain the only names where the recent window still pays.

The fair, 5y-Binance test AVAX's trend premise was said to lack FALSIFIES the premise rather than confirming it. Sharpe is only 0.337 with a bootstrap CI-low of -0.2938 — the CI straddles zero, so the significance gate the hypothesis claimed 'more history' would clear fails outright even on 5y / 427 trades. profit_factor is 1.17 (fee-edge zone, far below ADA/BNB's 2.0+), and the edge has broken down in the recent regime: 2025 -5.1%, 2026 -18.7% YTD (April -12.7%, July -10.7%), with rolling_sharpe collapsing to -13 mid-2026 — a 2021/2023-carried curve, not a distributed edge. It is also capacity-capped: capacity_usd only $4.18M with impact_cost_pct 15.46%, so thin AVAX liquidity would consume ~15% of gross PnL — the edge exists only at toy scale even if it were real. max_drawdown 22.7% (CI to 50.9%) and information_ratio -0.46 (underperforms buy-hold AVAX) round it out. The hypothesis framed AVAX as the strong-trending 'just needs history' tier like ADA; the honest test places it in the choppy-laggard 'no edge' tier (XRP/DOGE/LTC). Unlike ADA (CI-low 0.98, PF 2.19, positive every year), this does not clear significance and is getting worse. Abandon at BACKTEST_REVIEW rather than spend the optimization budget.

Long-short, single-instrument, single-bar-type momentum-confluence trend follower on AVAXUSDT.BINANCE USD-M perpetual, 4H bars. All reads come from the same 4H series: a fixed 12-bar fast EMA, a slow EMA at 12*ema_ratio, and a daily-horizon trend EMA of 135 4H bars (~22 days). calculate_signal returns (EMA_fast - EMA_slow)/ATR14 every bar -- a continuous, volatility-scaled 4H momentum read. ENTRY only when the sign of that read agrees with the daily-horizon trend (close above/below the trend EMA); disagreement means FLAT. EXIT on a 3x-ATR trailing stop from the best close since entry or on the close crossing the trend EMA against the position. Sizing risks ~1.5% of equity at the trailing-stop distance (qty = equity*0.015/(3*ATR)), hard-capped at 25% of equity x 2x leverage. Three tunables: ema_ratio, trend_ema, trail_atr_mult.

Analyst/backtest_review: abandon (or do not spend optimization budget) — the recent regime is negative at every grid cell and that is where the OOS windows sit. BTC/ETH are the only names in this family where the recent window still pays.

RECENT REGIME NEGATIVE — recommend ABANDON at BACKTEST_REVIEW. NOT a code defect (verified), so not a QA fail, but the OOS gates live exactly where this loses. Sandbox (trailing ~365 days): total_return -17.4%, Sharpe -0.97, PF 0.62, win_rate 0.151, avg_trade_return_pct -0.487%, DD 22.5%, PSR 0.13, 2026 YTD -1.07%/trade. Developer discloses EVERY cell of the 3x4x3 grid is negative over the trailing 365 days (-0.18% to -1.10%/trade) — structural to AVAX's recent chop, not a defaults artefact. Full-sample the mechanism is genuinely positive and distributed (+1.58%/trade, PF 1.56, kurtosis 20.1, top-3 15.2%), but the walk-forward OOS windows and 15-day holdout sit in the negative recent regime, so the OOS Sharpe>=0.5 and avg_trade_return_pct>=0.15% gates will almost certainly fail.

No code change needed — an edge/regime finding for the analyst, not a defect.

Verified the strongly negative sandbox is genuine no-recent-edge, NOT a code bug. Polarity correct (direction=(EMA_fast-EMA_slow)/ATR: rising -> LONG when px>trend_ema), both legs lose symmetrically (long_win 0.105, short_win 0.188), and no L17 code-defect signatures (win_rate 0.151 not 0, |Sharpe| 0.97 not >5, PF 0.62 not 0). Mechanism identical to the correct SOL/DOGE siblings. The loss is AVAX's recent trendless chop whipsawing a correct trend follower.

For live deployment, persist _side/_extreme/_entry_atr or reconstruct from the actual open position rather than from the current signal.

should_exit() infers _side from the live signal sign on restart (_side==0) and re-seeds _entry_atr/_extreme from current values. Unreachable in backtest; only a live mid-position crash-restart risk.

Distributed across years in the early sample (2021 +37.7%, 2023 +32.9%), return_kurtosis 13.1 not a single-day mirage

avg_trade_return_pct 1.75% is above fees in gross terms; 427 trades is a measurable sample

Sharpe 0.337 with CI-low -0.2938 — the CI STRADDLES ZERO; the significance gate the hypothesis claimed history would clear fails outright even on 5y/427 trades

profit_factor 1.17 — down in the fee-edge zone, far below ADA/BNB (2.0+)

Recent regime deeply losing and worsening: 2025 -5.1%, 2026 -18.7% YTD; rolling_sharpe collapses to -13 mid-2026

Capacity blocker: capacity_usd only $4.18M with impact_cost_pct 15.46% — thin AVAX liquidity, edge exists only at toy scale

max_drawdown 22.7% (CI to 50.9%), information_ratio -0.46 (underperforms buy-hold AVAX)

AvaxBinanceDualTimeframeMomentumConfluenceLS gave AVAX's Hyperliquid-praised trend edge the fair ~5-year Binance test its short-history version lacked, betting AVAX was an ADA-like 'just needs history' major rather than a range-bound laggard. The longer test settled it against the premise: +69.6% but Sharpe just 0.337 with a confidence interval straddling zero, profit factor 1.17, underperforming buy-and-hold AVAX, carried by 2021/2023 and losing badly in the recent regime (2026 -18.7% YTD, rolling Sharpe to -13), with a capacity of only ~$4M at 15.5% impact. The analyst abandoned it at backtest review, concluding the honest test falsified the edge and placed AVAX firmly in the no-edge tier with XRP/DOGE/LTC rather than with ADA. It never reached optimization, analysis, or risk review.

A Hyperliquid backtest that looks like distributed edge can be an artifact of a short single-regime window — given a fair 5-year test, AVAX's trend premise was falsified (Sharpe CI straddling zero, PF 1.17, recent regime deeply losing, thin capacity), showing the promising-but-unproven asset actually belongs with the no-edge laggards rather than the ADA tier.

The analyst abandoned it at backtest review: the fair long-history test falsified rather than confirmed AVAX's trend premise — the significance gate the hypothesis claimed more history would clear failed outright (Sharpe 0.337, CI straddling zero, PF 1.17 in the fee-edge zone), the edge has broken down in the recent regime, and thin liquidity caps it at toy scale (~$4M, 15.5% impact). The honest test places AVAX in the choppy-laggard no-edge tier, not the ADA-like just-needs-history tier.

A long-short, single-instrument dual-timeframe momentum-confluence trend follower on AVAXUSDT.BINANCE USD-M 4H (3 parameters) that aligned a fast/slow EMA spread with a ~22-day trend EMA — all from one bar series — going flat on disagreement and trailing winners with an ATR stop, giving AVAX's Hyperliquid-praised trend edge the fair ~5-year Binance test its short-history version lacked.

The backtest (AVAXUSDT.BINANCE 4H, 2142 data days from 2020-09) returned +69.6% over 427 trades with profit factor 1.17, avg_trade_return_pct 1.75%, max drawdown 22.7% (CI to 50.9%). But Sharpe was only 0.337 with a CI-low of -0.294 (straddling zero), it underperformed buy-and-hold AVAX (information ratio -0.46), the return was carried by 2021/2023 with a deeply losing recent regime (2025 -5.1%, 2026 -18.7% YTD, rolling Sharpe to -13), and capacity was just $4.18M with 15.5% impact cost.