BtcDailyFundingConfirmedTrendLS

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

A LONG-SHORT, SINGLE-INSTRUMENT, DAILY trend-following strategy on BTCUSDT.BINANCE (USD-M perpetual) that completes the factory's SEPARATELY-VALIDATED, PROMOTED funding-confirmed-trend family — promoted on ETH (EthPerpDailyFundingConfirmedTrendLS), pending on SOL (my prior submission) — by applying it to BTC, the deepest-liquidity, cleanest-trending, most-validatable core major. This is the evidence-forced choice after the design space collapsed to exactly two working, backtestable mechanisms: (1) daily price trend/momentum on BTC/ETH/SOL (promoted), and (2) funding-CONFIRMED trend (promoted ETH). The analyst has explicitly and repeatedly established the boundaries: funding-CONFIRMED trend WORKS while funding-EXTREME CONTRARIAN fade is CLOSED OUT (no edge); single-name PRICE-ONLY trend clones on non-core alts ALL fail (BNB/LTC/LINK/AVAX/DOT) but the trend edge is real on core majors; all intraday/contrarian fades wipe out; and any liquidation/OI-SIGNAL strategy is unbacktestable on the ~April-2026 forced-flow data wall. This hypothesis lives entirely inside the two WORKING families, on the single most-validatable asset, with BACKTESTABLE deep funding history (the funding-extreme run confirmed ~6 years / 149 funding observations exist for BTC, 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); momentum direction (trade WITH the trend, the working direction, never contrarian); and TAIL-SAFE controls (leverage <=1.5, hard ATR stop, modest notional cap) to prevent the LINK-style account blowup. The funding confirmation is the genuine differentiator versus the promoted plain BTC TS-momentum: it filters out 'hollow' trends (price rising on short-covering with non-confirming funding) and exits early on funding/price divergence (exhaustion), aiming to decorrelate from and improve the pure-price BTC trend sibling. Low parameter count to resist overfitting.

Iteration 2 fixes the Layer-2 failure: the verifier's strategy proxy does not preserve the `@staticmethod` binding, so `self._ema(values, period)` was passing `self` as a third positional argument and raising TypeError in all six synthetic scenarios. The minimal, surgical change is to convert `_ema` from a static method into a normal instance method (`def _ema(self, values, period)`), so `self` is correctly bound and the call passes exactly two positional args. No other logic, imports, parameters, or signal behavior changed — all earlier passing layers remain intact. The strategy still implements the funding-confirmed daily trend hypothesis on BTC: continuous EMA-spread signal, funding confirmation gating entries, divergence/trend/ATR exits, capital-relative tail-safe sizing consuming the 1.5x leverage.

Failed deflated Sharpe and the forward holdout decisively on optimization attempt 1. DSR=0.0003 (vs 0.95 bar) with the optimized Sharpe 0.739 FAR below the 225-trial expected-max luck bar of 2.0094 (is_significant=false, sharpe_ci_low -0.0382 straddling 0, PBO 0.6164 >0.5) — after multiple-testing correction the selected best-of-225 config is statistically indistinguishable from best-of-N noise; probabilistic_sharpe 0.9762 vs DSR 0.0003 is the textbook PSR-vs-DSR selection-inflation trap. Decisively, the untouched time-ordered HOLDOUT FAILED and is NEGATIVE (holdout_sharpe -0.9946, ratio -1.239 vs WF-OOS 0.8027) — the recent regime loses money on data the optimizer never saw, and the first walk-forward OOS window is already negative (-0.786). Sensitivity FAILED with 3 cliff parameters (trend_ma, tp_atr_mult, funding_smooth_obs) and the entire grid sits ~0.27-0.92, uniformly below the luck bar. The strategy carries a NEGATIVE information_ratio (base -0.674, optimized -0.6187) with benchmark_meaningful=true — it loses to its own BTC buy-hold risk-adjusted — and its edge is outlier-carried (kurtosis 55.6->79.8, skew 4.1->6.9, tail_ratio 0.05) and decaying (annual 2023 -5.5%, 2025 -10%, 2026 -2%). Not iterate (attempt 1 of 2): the binding failures are a negative forward holdout plus best-of-225 insignificance with PBO 0.62, and the whole sensitivity surface is below significance, so there is no robust region to tune toward — a second best-of-225 sweep re-selects the same sub-significant, decaying config. Not revise_hypothesis: BTC is a fine instrument and this is a forward-generalization/multiple-testing failure, not a proven mechanism stranded on a dead target — the funding-confirmed-trend family already has a promoted ETH sibling and BTC already has a promoted plain TS-momentum sibling, so adding funding confirmation on BTC created no distinct promotable edge. FAILURE PATTERN: porting the promoted ETH funding-confirmed-trend mechanism to single-asset BTC does NOT inherit promotability — the BTC port produces the same single-name daily-trend deflation signature (DSR 0.0003, optimized Sharpe 0.739 below the 2.01 luck bar, PBO 0.62, 3 sensitivity cliffs), a NEGATIVE forward holdout (-0.9946), and a negative information ratio vs its own buy-hold, with the edge outlier-carried and decaying since 2023. A non-overfit-flagged walk-forward and a positive OOS average measure consistency, not significance or forward-generalization, and do not rescue a strategy whose forward holdout loses money.

Long-short daily trend-following on BTCUSDT USD-M perpetual (daily bars), trading WITH the price trend (fast/slow EMA stack plus sloped 100-day MA) only when the funding rate confirms genuine directional demand (positive funding for longs, negative for shorts). Exits on hard ATR stop, ATR take-profit, funding/price divergence (exhaustion), or EMA-stack trend invalidation. Tail-safe: leverage 1.5x, hard ATR stop, 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', 'conf_funding', 'entry_thresh', 'exit_funding', 'min_notional', 'atr_stop_mult', 'ma_slope_lookback', 'funding_smooth_obs'] Check that __init__ sets all attributes from self.parameters.get(). - steady_uptrend: TypeError: BtcDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000) - steady_downtrend: TypeError: BtcDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000) - flat_ranging: TypeError: BtcDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000) - volatility_spike: TypeError: BtcDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000) - zero_volume: TypeError: BtcDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000) - price_gap: TypeError: BtcDailyFundingConfirmedTrendLS._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735697040000)

Strategy functions correctly: 62 signals -> 62 entries, 0 dropped; both long (35) and short (27) legs fire.

Trades implement the stated mechanism — funding-confirmed, WITH-trend, long-short daily — no mechanism mismatch or coding-bug smell.

Genuine (if modest) edge that survives fees: profit_factor 1.35, expectancy +$395/trade, avg_position ~20% equity, max_dd contained at 14.3%.

Trade count (62 over ~6.5y) is appropriate for a filter-gated daily trend strategy and adequate for a parameter sweep; sits inside a promoted family (ETH funding-confirmed-trend already promoted).

Modest headline Sharpe 0.45 with sharpe_ci_low -0.46 (CI straddles 0) — significance is the open question, to be settled by deflated-Sharpe/holdout in optimization.

Return is regime-concentrated: 2020 (+15%) and 2024 (+23%) carry it while 2023 (-5.5%), 2025 (-10%), 2026 (-2%) are negative; return_kurtosis 55.6 / skew 4.1 signals outlier dependence.

information_ratio -0.67 (benchmark_meaningful=true) — loses to buy-hold risk-adjusted; the funding filter must prove it decorrelates from the plain BTC TS-momentum sibling.

Single-instrument/single-venue/single-account design avoids the multi-instrument aggregation and funding-accrual pitfalls; funding used only as a confirming signal.

Walk-forward not overfit-flagged (avg IS 1.15 -> avg OOS 0.80) and low base drawdown (14.3%); optimized max_dd 10.1%.

Clean execution: 62 base / 72 optimized trades, no dropped/min-notional issues, avg_trade_return well above fee floor.

Untouched time-ordered HOLDOUT FAILED and is NEGATIVE: holdout_sharpe -0.9946, ratio -1.239 vs WF-OOS 0.8027 — the recent regime loses money out-of-sample.

Fails multiple-testing robustness decisively: DSR=0.0003, is_significant=false, sharpe_ci_low -0.0382 (CI straddles 0), PBO=0.6164 (>0.5); optimized Sharpe 0.739 far below the 225-trial expected-max luck bar of 2.0094.

Sensitivity FAILED with 3 cliff parameters (trend_ma, tp_atr_mult, funding_smooth_obs); entire surface sits ~0.27-0.92, no robust region above significance.

NEGATIVE information_ratio (base -0.674, optimized -0.6187) with benchmark_meaningful=true — loses to its own BTC buy-hold risk-adjusted.

First OOS window negative (-0.786); edge is outlier-carried (kurtosis 55.6/79.8, skew 4.1/6.9, tail_ratio 0.05) and decaying (annual 2023 -5.5%, 2025 -10%, 2026 -2%).

The BTC USD-M Daily Funding-Confirmed Trend strategy aimed to complete a validated family by trading with the BTC trend only when funding confirmed real directional demand, filtering out 'hollow' trends and exiting on funding/price divergence. Its first backtest looked reasonable — 37.38% return, Sharpe 0.45, 62 clean trades, contained 14.3% drawdown — earning an 'optimize' verdict despite regime-concentrated, outlier-carried returns. But the 3-phase optimization exposed the edge as luck: the untouched time-ordered holdout was outright negative (-0.9946 Sharpe), deflated Sharpe was 0.0003 against a 0.95 bar, PBO was 0.62, and the optimized Sharpe of 0.739 fell far short of the 2.01 best-of-225 luck bar, with three sensitivity cliffs and a negative information ratio versus buy-hold. The analyst abandoned it, concluding that layering funding confirmation onto BTC — which already had a promoted momentum sibling — created no promotable edge and merely reproduced the family's single-name deflation-and-decay failure pattern.

Porting a promoted mechanism (ETH funding-confirmed trend) onto a different single asset does not inherit its promotability — BTC already had a promoted plain TS-momentum sibling, so adding funding confirmation produced no distinct, forward-generalizing edge and reproduced the single-name daily-trend deflation signature.

The analyst voted abandon after optimization attempt 1 because the strategy failed the deflated-Sharpe and forward-holdout tests decisively — a negative out-of-sample holdout, best-of-225 statistical insignificance (PBO 0.62), 3 sensitivity cliffs, and a negative information ratio versus its own BTC buy-hold left no robust region to tune toward.

A long-short, single-instrument daily trend-following strategy on BTCUSDT Binance perps that traded with the price trend only when the funding rate confirmed genuine directional demand, porting the promoted ETH funding-confirmed-trend mechanism to BTC.

The initial backtest cleared the review gate with a 37.38% total return, Sharpe 0.45, 62 trades (35 long / 27 short), profit factor 1.35 and 14.3% max drawdown, but optimization collapsed: the untouched holdout was negative (holdout_sharpe -0.9946, ratio -1.239 vs WF-OOS 0.80), deflated Sharpe was 0.0003, PBO 0.6164, and the optimized Sharpe of 0.739 sat far below the 225-trial luck bar of 2.0094.