SolDailyFundingConfirmedTrendLS

SOL USD-M Daily Funding-Confirmed Trend, Long-Short (Single-Instrument BINANCE Perp, Daily Bars + Multi-Year Funding History, Trade WITH the Trend Only When Funding CONFIRMS Healthy Directional Demand, Exit on Funding/Price Divergence, Tail-Safe Sizing, Low-Parameter)

A LONG-SHORT, SINGLE-INSTRUMENT, DAILY trend-following strategy on SOLUSDT.BINANCE (USD-M perpetual) that extends the factory's SEPARATELY-VALIDATED, PROMOTED mechanism EthPerpDailyFundingConfirmedTrendLS — trade WITH the trend, but only when perpetual FUNDING CONFIRMS the move's directional demand — to another core, cleanly-trending major (SOL, whose pure-price TS-momentum sibling is already at paper_stage, so SOL is proven to trend). This is the deliberate, evidence-driven choice after the design space collapsed: the analyst has explicitly established that (a) funding-CONFIRMED trend WORKS (promoted) while funding-EXTREME CONTRARIAN fade does NOT (closed out); (b) single-name PRICE-ONLY trend clones on non-core alts all fail (BNB/LTC/LINK/AVAX/DOT); (c) all intraday/contrarian fades wipe out; and (d) any liquidation/OI-SIGNAL strategy is unbacktestable on the current ~April-2026 forced-flow data wall. This hypothesis sits squarely in a WORKING family on a CORE trender with BACKTESTABLE deep funding history (the funding-extreme run already confirmed ~6 years / 149 funding observations are available, so funding-as-signal is fully backtestable, unlike liquidations/OI). It is engineered around every confirmed landmine: SINGLE-INSTRUMENT / SINGLE-VENUE / SINGLE-ACCOUNT (no multi-instrument equity/return aggregation bug); funding used only as a confirming SIGNAL (no carry/accrual/cross-account issue); DAILY cadence (not the dead intraday-fade horizon); and TAIL-SAFE risk controls (leverage <=1.5, hard ATR stop, modest notional cap) specifically to avoid the LINK-style account blowup. The funding confirmation is the genuine differentiator versus plain SOL TS-momentum: it filters out 'hollow' trends (price rising on short-covering with negative funding) and exits early on funding/price divergence (exhaustion), aiming to decorrelate from and improve on the pure-price trend sibling. Low parameter count to resist overfitting.

Iteration-2 fix: Layer-2 synthetic verification failed because the strategy proxy rebinds methods and injects `self`, so the former `@staticmethod _ema(values, period)` received 3 positional args. Converted `_ema` to a normal instance method `_ema(self, values, period)` — the sole change. All existing call sites `self._ema(closes[...], period)` now pass exactly (self, values, period), matching the signature in both normal and proxied execution. No imports, signal logic, funding-confirmation logic, or sizing were altered, so Layer 1 (static) stays green and the mechanism that reached Layer 2 is preserved. The strategy implements the promoted funding-confirmed-trend family on a core trender (SOL), using funding purely as a confirming signal with tail-safe risk controls.

Failed deflated Sharpe and the forward walk-forward gates decisively on attempt 1. DSR=0.0 (vs 0.95 bar) with the optimized Sharpe 0.7606 FAR BELOW the 225-trial expected-max luck bar of 2.9479 (is_significant=false, sharpe_ci_low -0.0465 straddling 0, PBO 0.7381 >0.5) — the selected config is statistically indistinguishable from best-of-N noise. The walk-forward is is_overfitted=TRUE with avg IS 0.917 collapsing to avg OOS -1.602 (windows [-3.37, +0.68, -2.12], two deeply negative). The holdout 'pass' is SPURIOUS — its ratio is computed against a negative WF-OOS (-1.602), so a positive holdout_sharpe (0.72) cannot validate a strategy that loses money out-of-sample. This is the classic PSR-vs-DSR trap: probabilistic_sharpe 0.9835 ignores the trial count that DSR corrects for. The clean sensitivity grid (0 cliffs) measures consistency, not significance, and the entire surface sits below the luck bar. Optimization made it WORSE (base Sharpe 0.843 -> optimized 0.761), and the edge is outlier-carried and regime-concentrated (kurtosis 86-112, skew 7.3-9.3, tail_ratio 0.088; ~35% of the base return earned in the 2021 bull, with 2023 -6.4% / 2025 -4.0% recent years). Not iterate (attempt 1 of 2): the sensitivity surface is a flat 0-cliff plateau (optimizer already on its plateau) with no under-explored robust region, and the binding failure is a negative-OOS, sub-luck-bar edge — a second best-of-225 sweep re-selects the same non-significant config. Not revise_hypothesis: the funding-confirmed-trend mechanism is already validated/promoted on the ETH sibling and SOL is a proven trender (price-only TSM sibling at paper_stage), so this is the funding-confirmation VARIANT on SOL failing multiple-testing (asset/variant-selectivity), not a sound mechanism stranded on a dead target. FAILURE PATTERN: porting the promoted ETH funding-confirmed daily trend mechanism to SOL does NOT inherit promotability — the SOL arm's edge is too modest and too 2021-concentrated (kurtosis ~100) to clear best-of-225 deflation (Sharpe 0.76 vs 2.95 luck bar, DSR 0.0) and its walk-forward OOS is negative in two of three windows (avg -1.60, is_overfitted). A clean sensitivity grid, a spurious holdout ratio against a negative WF-OOS baseline, and PSR 0.98 do not rescue a strategy that fails deflated Sharpe with a CI through 0 and PBO 0.74. The promoted ETH instance remains the asset where this mechanism clears the bar; SOL does not.

Long-short daily trend-following on SOLUSDT USD-M perpetual (BINANCE). Enters WITH the trend (fast>slow EMA stack, price above/below a sloped trend-MA) only when smoothed perpetual funding CONFIRMS directional demand: longs need modest positive funding (not frothy-extreme), shorts need modest negative funding. Exits on hard ATR stop, ATR take-profit, EMA-stack flip, or funding/price divergence (exhaustion). Single-instrument/single-venue/single-account, tail-safe sizing (leverage 1.5, ATR-risk-based notional capped at leveraged equity).

Verification failed (Layer 2 — synthetic scenarios): Parameters used: ['fast_ema', 'slow_ema', 'trend_ma', 'risk_frac', 'atr_period', 'tp_atr_mult', 'min_notional', 'atr_stop_mult', 'entry_threshold', 'funding_lag_sec', 'funding_exit_min', 'ma_slope_lookback', 'funding_confirm_min', 'funding_extreme_cap', 'funding_smooth_days'] Check that __init__ sets all attributes from self.parameters.get(). - steady_uptrend: TypeError: SolDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000) - steady_downtrend: TypeError: SolDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000) - flat_ranging: TypeError: SolDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000) - volatility_spike: TypeError: SolDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000) - zero_volume: TypeError: SolDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000) - price_gap: TypeError: SolDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000)

86 trades over 5.7 years (~15/yr) is an appropriate sample for a daily long-short trend strategy — enough to optimize meaningfully without being noise-fit-prone.

Strategy is clearly functioning: 86 signaled -> 86 submitted, 0 dropped (size/cash/min-notional all clean); both long (37) and short (49) legs fire, consistent with the funding-confirmed LS hypothesis.

Genuine risk-adjusted edge: Sharpe 0.84, Sortino 1.98, profit factor 1.72, PSR 0.99, positive alpha 0.045 with low beta 0.02 and max DD 16%.

Fee-robust: commission only 1.01% of gross; sits in a validated family (promoted EthPerpDailyFundingConfirmedTrendLS sibling) on a proven trender (SOL).

Returns are heavily concentrated in 2021 (+35%) vs low single-digit years afterward — outlier/regime-carried edge (return_kurtosis 85.9, skew 7.29, tail_ratio 0.088).

Long pain profile: max drawdown duration 693 days and 11 consecutive losses.

These require the deflated-Sharpe / walk-forward / holdout gates to confirm the edge is not best-of-N noise — flagging for the full post-optimization review.

Clean sensitivity grid (0 cliffs) — parameters are individually robust, not on knife-edges

Low drawdown (max_drawdown 15.5% optimized), high recovery/calmar (5.69), Sortino 2.25 — risk controls work as designed

Well-engineered around known landmines: single-instrument/venue/account, funding-as-signal only, daily cadence, tail-safe sizing (no blowup, not liquidated)

Funding history is genuinely deep and backtestable (4,699 events) — the mechanism is measurable

Failed deflated Sharpe: DSR=0.0 vs 0.95 bar; optimized Sharpe 0.76 FAR below the 225-trial expected-max luck bar of 2.9479 (is_significant=false)

Walk-forward is_overfitted=true: avg IS 0.917 collapses to avg OOS -1.602, with OOS windows [-3.37, +0.68, -2.12] (two deeply negative)

PBO 0.7381 (>0.5) and sharpe_ci_low -0.0465 (CI through 0) — selection is more likely than not overfitting

PSR 0.9835 vs DSR 0.0 — textbook multiple-testing inflation; PSR ignores the 225-trial count DSR corrects for

Holdout 'pass' is spurious: ratio computed against a NEGATIVE WF-OOS (-1.602); holdout_sharpe 0.72 cannot validate a negative forward baseline

Optimization made it WORSE (base Sharpe 0.843 -> optimized 0.761) — no exploitable robust region

Edge is outlier-carried and regime-concentrated: kurtosis 86-112, skew 7.3-9.3, tail_ratio 0.088; return dominated by 2021 (+57% opt) with recent years flat-to-negative (2023 -6.4%, 2025 -4.0%)

SolDailyFundingConfirmedTrendLS extended the factory's promoted ETH funding-confirmed daily trend mechanism to SOL, trading long-short with the trend only when perpetual funding confirmed directional demand, engineered around known landmines with single-instrument, tail-safe sizing. The initial backtest looked healthy enough (Sharpe 0.84, profit factor 1.72, 86 trades, 16% max drawdown) that the analyst gated it to full optimization, but the 3-phase optimization exposed the edge as best-of-N noise: deflated Sharpe of 0.0 against a 0.95 bar, an optimized Sharpe of 0.76 far below the 2.95 luck bar, PBO 0.74, and a walk-forward whose in-sample 0.92 collapsed to -1.60 out-of-sample across two deeply negative windows. On its first attempt, after 2 iterations, it was abandoned because the edge was outlier-carried and concentrated in the 2021 bull, and a second sweep on the already-flat sensitivity plateau would only re-select the same non-significant configuration. The record concludes that the promoted ETH instance remains the asset where this mechanism clears the bar, while SOL does not.

A mechanism promoted on one asset (ETH) does not inherit promotability when ported to another (SOL) — an outlier-, regime-concentrated edge that clears best-of-N deflation on one target can fail it decisively on another, and a clean sensitivity grid, spurious holdout, or high PSR cannot rescue a negative out-of-sample walk-forward.

The analyst abandoned it after optimization: it failed the deflated Sharpe gate (DSR 0.0 vs 0.95, optimized Sharpe 0.76 far below the 225-trial luck bar of 2.95), the walk-forward was flagged overfitted (avg IS 0.92 collapsing to avg OOS -1.60, two of three windows deeply negative), with PBO 0.74 and a Sharpe CI through zero.

A single-instrument, single-venue daily long-short trend-following strategy on SOLUSDT.BINANCE perpetuals that only traded WITH the EMA/trend-MA trend when smoothed perpetual funding confirmed healthy directional demand, porting the already-promoted ETH funding-confirmed-trend mechanism to SOL.

The initial backtest returned 65.15% total over ~5.7 years across 86 trades with Sharpe 0.84, Sortino 1.98, profit factor 1.72 and 16.08% max drawdown; optimization raised total return to 120.64% but lowered Sharpe to 0.76, with returns heavily concentrated in 2021 (kurtosis 86–112, skew 7.3–9.3).