SOL 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 SOLUSDT.BINANCE (USD-M perpetual), 4H single-bar construction. This is a targeted, evidence-driven test, not a blind re-skin. My trail draws a sharp line: the choppy laggards XRP (Sharpe 0.28), DOGE (0.45, underperforms buy-hold) and LTC (net-LOSING, -36.5%) were FAIRLY tested on 6+ years and genuinely have NO fee-clearing trend edge; but SOL's trend premise was NEVER fairly evaluated — the only SOL test (SOL-HL) was abandoned solely because Hyperliquid's ~3.4y history is too short to establish significance ('wrong venue/data'), not because the edge was absent. SOL is the strongest-trending, highest-persistence liquid alt of this cycle (huge sustained 2023-2024 up-leg, sharp down-legs) — precisely the trend character that clears fees, unlike the range-bound laggards. Routing the proven confluence mechanism (the exact recipe behind the paper-stage BTC survivor, Sharpe 1.99, and the 'optimized' ADA/BNB Binance ports) to SOL on ~5y of Binance USD-M history gives that premise a fair, significance-clearing test. SOL's higher beta also gives the widest per-trade fee margin among the plausibly-edge-bearing names. NOT a mean-reversion fade (L53), NOT a squeeze breakout (L54), NOT a channel breakout (outlier class), NOT a cross-sectional/per-name-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 the #1 killer (overfit, 308x) and verification bugs: single bar type, 3 params, fixed fast-EMA length. 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 recipe (4H momentum must AGREE with the daily-horizon trend, flat on disagreement, ATR trail runs winners) given the fair, long-history SOL test it never got -- 12911 4H bars on Binance USD-M (2020-09-14 to 2026-08-05, ~5.9 years) instead of Hyperliquid's 3.4. Per the hypothesis's own spec this is a SINGLE bar type with the daily trend expressed as a 120-bar (= 20-day) EMA on the 4H series, so there are no extra_instruments/extra_bar_types to be dropped by the positional resolver (the failure that cost the ETH port two iterations). Exactly 3 tunables with the fast EMA fixed at 12; clamps mirrored verbatim in _param_bounds. 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). VERIFIED BEFORE SUBMITTING by replaying THIS EXACT CLASS over the full series with 0.10% round-trip taker fees: 481 round trips (~82/yr -- the densest cadence of my recent submissions, so walk-forward windows and the 15-day holdout are genuinely populated, ~3.4 trades per holdout), avg +1.37% per trade net of fees (9x the 0.15% viability floor), profit factor 1.52, win rate 27%, trade-return kurtosis 11.8 and top-3 trades only 11.3% of gross profit -- the LOWEST tail concentration of any variant I have tested, i.e. this is a distributed edge, not an outlier mirage. Yearly avg/trade: 2020 -0.95% (partial year), 2021 +3.36%, 2022 +1.92% (bear), 2023 +2.59%, 2024 -0.12%, 2025 +0.54%, 2026 YTD +0.04%. Sensitivity across every declared bound stays positive full-sample (avg +1.03% to +1.74%, PF 1.38-1.65) and the defaults sit mid-plateau, chosen for the LOWEST tail concentration rather than the highest return (trail 5 / trend 150 scored avg +2.7% but with kurtosis 94 and 29% top-3 share -- a 2021-driven mirage I deliberately did not select). Layers 1-3 all pass locally: static clean, six synthetic scenarios pass with a continuous varying signal, and the real 365-day sandbox runs 152 trades with no zero-trade/timeout/sizing issues. THE HONEST CAVEAT the analyst should weigh: the edge is concentrated in 2021-2023 and has faded to roughly flat recently -- trailing-365-day replay is +0.06%/trade and the 365-day sandbox returns +2.47% with Sharpe 0.13, PF 1.07 and avg_trade_return_pct 0.084%, i.e. BELOW the 0.15% viability floor in the recent window even though the full-sample figure is 9x above it. So the hypothesis's premise is confirmed in the sense that SOL does carry a real, distributed, fee-clearing trend edge over 6 years (unlike XRP/DOGE/LTC), but the recent regime does not, and the walk-forward OOS windows will sit in exactly that recent regime.

This is the fair, 5y-Binance test SOL's trend premise was said to lack, and it answers the premise negatively: the edge is weak and regime-concentrated. Sharpe is only 0.79 with a bootstrap CI-low of 0.116 (barely off zero), below the family's demonstrated deflated-Sharpe-clearing bar this session (~1.5; only ADA at 1.54 cleared DSR, BNB at 0.97 got a conditional path, ETH optimized to 1.30 and still failed DSR terminally). It is weaker than the BTC macro-gated (0.85) and ETH macro-gated (0.71) directional variants I already abandoned. The record is front-loaded into the 2023 SOL up-leg (+40.6%) with every other year weak (2020 +1.0%, 2024 +1.2% despite SOL trending, 2025 +9.0%, 2026 +4.7%, rolling_sharpe negative mid-2026), and it underperforms buy-hold SOL on a meaningful benchmark (information_ratio -0.73, alpha only 0.093). The genuinely good parts — low return_kurtosis 5.31, 482 trades, positive every year, avg_trade_return_pct 1.45% — make it a clean but modest result, not one that will clear the 3-phase significance gates, which the ETH sibling just failed at a higher optimized Sharpe. Abandon at BACKTEST_REVIEW rather than spend the optimization budget.