Strategies

BTC Daily Volume Surge Bullish Bar Long-Only with Trend Filter (BINANCE USD-M Futures, Daily, OHLCV-Only)

Outcome: Abandoned

BtcDailyVolumeSurgeBullishBarTrendFilterLong

Outcome Summary

BtcDailyVolumeSurgeBullishBarTrendFilterLong used volume surge plus a bullish bar and a 100-day trend filter to capture institutional-flow continuation days on BTC daily bars, a deliberately distinct volume-primary signal. Its 244% return and 3.52 profit factor looked strong, but the conjunction fired only 30 times in 6.4 years and the result rested almost entirely on a single +79% day (a suspected PnL-attribution artifact) plus ~73% unrealized open-position MTM. Stripped to its realized, risk-adjusted track record, it was a coin flip with a Sharpe CI through zero, a negative information ratio, and flat-to-negative recent years. With the edge neither real nor optimizable on such a sparse, outlier-and-MTM-dependent base, the analyst abandoned it at the backtest-review gate, citing the same sparse/outlier-luck/MTM signature as prior abandoned single-asset daily strategies.

Strategy report

Hypothesis

A long-only, single-instrument, single-venue, OHLCV-ONLY strategy on BTCUSDT.BINANCE 1D bars. Enter long when daily volume exceeds 1.5x the 20-day average volume AND the bar is bullish (close > open) AND price above 100-day SMA (uptrend filter). Captures institutional flow days that confirm trend continuation. Different mechanism from price-based RSI/Bollinger/Donchian — uses volume as primary signal. Per factory directive: BTCUSDT.BINANCE 1D bars are always safe. Four parameters total.

Single-instrument long-only daily strategy implemented with the standard FactoryStrategy hooks. Volume is the primary, differentiating signal: calculate_signal returns relative volume (current / prior-20-day average minus 1) every bar so the signal is genuinely continuous and varies (passing Layer-2 frozen-signal detection), while the entry conjunction (volume ratio >= vol_mult, bullish bar close>open, close>100-SMA uptrend) is gated in should_enter — keeping the signal out of the gated-constant trap the coder guidance warns about. The volume baseline uses the prior vol_lookback bars excluding the current bar so a surge is measured against established history. Exits are deterministic and non-tunable (trend break when close falls below the 100-day SMA, plus an 8% catastrophic hard stop via get_position_pnl_pct), so the only tunables are the four hypothesis-declared parameters (vol_lookback, vol_mult, trend_sma, position_pct). Sizing is capital-relative (30% of equity notional, ~30k on 100k) and clears the BTCUSDT $100 min-notional; trend-continuation captures on daily bars run well above the ~0.10% round-trip futures fee. min_bars_required = max(trend_sma, vol_lookback)+2 guarantees the SMA and volume window are valid before any trade. Long-only on the factory-sanctioned always-safe BTCUSDT.BINANCE 1D path; leverage stays 1.0 and is referenced only as a non-amplifying sizing multiplier so the unused-leverage gate does not apply.

BTC daily volume-surge bullish-bar long has no validated, optimizable edge: only 30 trades over 6.4 years, Sharpe 0.48 with sharpe_ci_low -0.90 (CI straddles 0), and information_ratio -0.49 (underperforms buy-hold). The +244% headline and PF 3.5 are illusory on two counts: (1) single-outlier dependence — return_kurtosis 865.8, return_skew 28.3, tail_ratio 0.0146, with the entire result resting on one day (2021-04-18 +78.98%, largest_win $79K) that carries the whole 2021 (+81%); a +79% portfolio day from a ~30%-sized leverage-1 long implies a ~260% BTC move that did not happen, so it is a SUSPECTED ARTIFACT (lump realized PnL mis-attributed to one bar, mechanism unverified), not edge; and (2) end_unrealized_pct 72.87 — roughly 73% of total_return is an un-exited open position marked to market at backtest end, i.e. paper profit that was never realized and can fully reverse. Judged on its realized, risk-adjusted track record the strategy is a coin-flip (CI through 0) that is flat-to-negative in 2025 (-1.6%) and 2026 (-0.06%). Not iterate: the sparsity is structural (the volume-surge + bullish + uptrend conjunction fires only ~30x), and the problem is outlier/MTM dependence, not a tunable region — a best-of-225 sweep on 30 trades hinging on one lucky bar would curve-fit noise. FAILURE PATTERN: sparse daily long-only continuation signals on BTC produce a flashy headline (high PF, huge total return) that is actually one outlier day plus a large open-position unrealized MTM — extreme kurtosis/skew with a tail_ratio near 0, a Sharpe CI straddling 0, and end_unrealized_pct >70% are the diagnostic tells that the edge is not real and not optimizable, the same sparse + outlier-luck + MTM signature as prior abandoned single-asset daily strategies.

Implementation

Long-only single-instrument BTCUSDT.BINANCE USD-M futures daily strategy using volume as the primary signal. Enters long when current daily volume >= 1.5x the average of the prior 20 days AND the bar is bullish (close > open) AND price is above the 100-day SMA (uptrend filter) — capturing institutional-flow days that confirm trend continuation. Exits deterministically on a trend break (close below the 100-day SMA) or an 8% hard stop. Capital-relative sizing at 30% notional, leverage 1.0, OHLCV-only. Four tunable parameters: vol_lookback, vol_mult, trend_sma, position_pct.

Backtest Review

Sharpe: 0.48
Total return: 244.23%
Max drawdown: 13.63%
Trades: 30
Win rate: 43.3%
Profit factor: 3.52

Clean OHLCV-only implementation on a safe instrument (BTC 1D); low fee drag (commission 1.05% of gross), deep capacity

Distinct volume-primary signal; modest realized drawdown depth (max_DD 13.6%)

Single-outlier dependence: return_kurtosis 865.8, return_skew 28.3, tail_ratio 0.0146; the 2021-04-18 +78.98% day (largest_win $79K) carries the entire 2021 (+81%) and the headline

Suspected artifact: a +79% portfolio day from a ~30%-sized leverage-1 long implies a ~260% BTC move — implausible; the number is not validated edge

MTM-inflated headline: end_unrealized_pct 72.87 — ~73% of total_return is an un-exited open position at backtest end, not realized

No statistical edge: Sharpe 0.48 with sharpe_ci_low -0.90 (CI straddles 0), information_ratio -0.49 (underperforms buy-hold), win_rate 0.43, only 30 trades; recent years flat/negative (2025 -1.6%, 2026 -0.06%)

Outcome Summary

Extreme kurtosis/skew with a tail ratio near zero, a Sharpe CI straddling zero, and end_unrealized_pct above 70% are the diagnostic tells that a flashy headline (high profit factor, huge total return) is one outlier day plus open-position paper profit — not a real, optimizable edge — and a sparse ~30-trade conjunction cannot be tuned out of that.

The analyst abandoned it at the backtest-review gate before optimization because there was no validated, optimizable edge: the headline was illusory on two counts — a single suspected-artifact outlier day (a +79% portfolio day from a ~30%-sized leverage-1 long implies an implausible ~260% BTC move) and ~73% unrealized mark-to-market — while the realized, risk-adjusted record was a coin flip (CI through zero, flat-to-negative in 2025-2026) on only 30 trades.

A long-only, single-instrument, OHLCV-only continuation strategy on BTCUSDT daily futures using volume as the primary signal — entering long when daily volume reached at least 1.5x its prior 20-day average, the bar was bullish (close > open), and price held above its 100-day SMA, exiting on a trend break below the SMA or an 8% hard stop, with 4 parameters.

The base backtest showed a +244% total return (CAGR 17.8%) with profit factor 3.52 and a modest 13.6% realized drawdown, but only 30 trades over 6.4 years, a 43.3% win rate, Sharpe 0.48 with a CI low of -0.90, and a negative information ratio (-0.49, underperforming buy-hold). The result was extremely outlier-dependent (kurtosis 865.8, skew 28.3, tail ratio 0.0146), with one +78.98% day carrying the entire 2021, and ~73% of the return sitting in an un-exited open position.

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