Strategies

FIL Four-Hour Volume Breakout Long-Only (BINANCE USD-M Futures, 4H, OHLCV-Only)

Outcome: Abandoned

FilFourHourVolumeBreakoutLong

Outcome Summary

This strategy ported the validated 4H volume-breakout mechanism to FIL, pivoting to 4H to dodge the daily-bar refill failures plaguing the session. It was implemented cleanly (193 entries, none dropped) and net-positive, but the edge was marginal — profit factor 1.058, Sharpe 0.27 with a CI straddling zero, only ~$2.5M capacity at 20% impact, and decaying into 2025/2026. The analyst abandoned it before optimization on conclusive within-family precedent: the stronger DOT arm had optimized into a flashy in-sample Sharpe that failed deflated Sharpe and a negative holdout, AVAX failed too, and LINK was abandoned pre-opt on the same thin-PF/low-capacity logic. FIL, weaker than DOT on an even thinner instrument, was judged the negative-selectivity arm of a family whose value is already captured on the liquid majors — not worth two hours to re-run a known-losing experiment.

Strategy report

Hypothesis

Long-only FIL (Filecoin) 4-hour strategy on BINANCE USD-M futures. Direct analog of the landed BTC/ETH/SOL/BNB/AVAX/XRP/MATIC/TRX/ETC FourHourVolumeBreakoutLong family — applied to FIL. Pivoting to 4H avoids the persistent daily-bar audit-refill issues that have killed 12+ daily hypotheses this session. Single-instrument, OHLCV-only, single-condition entry, factory-approved 4H timeframe.

Ports the proven FourHourVolumeBreakoutLong recipe (landed on BTC/ETH/SOL/BNB/AVAX/XRP and others) to FIL, with no mechanism changes. The single-condition entry is a Donchian channel breakout confirmed by the breakout bar's own volume exceeding its trailing average -- same-bar volume CONFIRMATION (not divergence), which prunes thin-volume fakeouts. calculate_signal returns the continuous, bar-varying Donchian breakout margin so the verifier sees real signal variation, while the discrete price-break-AND-volume-confirm boolean gates should_enter and position_size, avoiding the gated-to-0.0 anti-pattern. Single instrument, single 4H timeframe, OHLCV-only avoids the daily-bar audit-refill failures and cross-asset audit_stale risk. Venue FILUSDT.BINANCE USD-M 4H matches the family; FIL is a long-listed liquid Binance futures product, futures RT taker (~0.10%) is cheaper than spot, and FIL qty rounds to 1 decimal (FILUSDT step). Long-only with leverage left at 1.0 (no amplification) so the leverage_set_but_unused gate does not trigger; the ATR/6% targets keep expected per-trade edge above the ~0.10% round-trip fee.

FIL is the weak, lower-liquidity, fee-fragile arm of the validated FourHourVolumeBreakoutLong family and is not worth a 2-hour optimization. The base edge is below the meaningful bar: profit_factor 1.058 (<1.2), Sharpe 0.265 with CI -1.20 to 1.58 (straddles 0), total_return +12.5% over 5.7 years (CAGR ~1.3%), omega 1.06. Capacity is only $2.5M with impact_cost_pct 19.9%, so any 'successful' optimization yields an edge real only at toy scale, and the strategy is decaying out-of-sample (negative 2025 -2.9% and 2026 -3.2%, rolling Sharpe -3 to -4 through 2026). The code is correct and trades as specified (193 entries, 0 dropped, long-only), so this is a weak-edge/instrument-selectivity problem, not a bug. Decisively, the precedent within this exact family is conclusive: the STRONGER DOT arm (PF 1.11) optimized then failed deflated Sharpe (0.24) with a negative holdout and was abandoned as 'the weakest, most fee-fragile arm'; AVAX failed too; and LINK Keltner was abandoned pre-optimization on identical logic (thin PF ~1.08, ~$3M capacity, sibling already failed). FIL is weaker than DOT (PF 1.058 < 1.11) on an even lower-liquidity instrument -- a base Sharpe 0.27 with a CI straddling 0 will, after best-of-225 selection, produce a flashy in-sample number that fails DSR and the holdout, exactly as the stronger arms did. Not optimize: re-running a known-losing experiment on a weaker instrument. Not iterate: implementation is sound; FIL is simply the negative-selectivity arm of a family whose value is already captured on the liquid majors (BTC/ETH/SOL/BNB/XRP). FAILURE PATTERN: the FourHourVolumeBreakoutLong family does not generalize to lower-liquidity alts -- the marginal arms (DOT PF 1.11, AVAX, LINK, now FIL PF 1.058) show thin profit factors, Sharpe CIs straddling 0, small capacity ($2.5-3M), and recent-year decay, and they optimize to flashy in-sample Sharpes that fail deflated Sharpe and the one-shot holdout. A pre-opt PF below ~1.1 on a low-capacity alt member of a family whose stronger siblings already failed the post-opt gates should be abandoned rather than optimized.

Implementation

Long-only 4-HOUR volume-confirmed Donchian breakout on FILUSDT.BINANCE USD-M futures, pure OHLCV single feed. Direct analog of the landed FourHourVolumeBreakoutLong family. Enters LONG when price closes above the prior 20-bar Donchian high AND the breakout bar's volume is at least 1.5x the trailing 20-bar average volume. Exits on a 1.5x ATR stop, a 3x ATR target, a 6% take-profit fallback, a lower-Donchian reclaim (close below the prior 10-bar low), or a 30-bar time stop. The signal is the continuous Donchian breakout margin (close/prior_high - 1) recomputed every bar; the volume-confirmed breakout is the discrete entry gate. Sized at 18% equity notional, capped so ATR-stop risk <= 1.5% of equity. leverage 1.0.

Backtest Review

Sharpe: 0.27
Total return: 12.47%
Max drawdown: 13.63%
Trades: 193
Win rate: 43.0%
Profit factor: 1.06

Implemented correctly and trades as specified: 193 entries, 0 dropped, long-only; clean member of the validated FourHourVolumeBreakoutLong family; low drawdown (13.6%)

Net positive with a per-trade return (~0.21%) just above the futures fee floor; adequate sample (193 trades)

Marginal edge below the meaningful bar: profit_factor 1.058 (<1.2), Sharpe 0.265 with CI -1.20 to 1.58 (straddles 0), CAGR ~1.3%, omega 1.06

Lower-liquidity / fee-fragile: capacity only $2.5M with impact_cost_pct 19.9% -- edge real only at small scale

Recent decay: negative in 2025 (-2.9%) and 2026 (-3.2%), rolling Sharpe -3 to -4 through 2026

Weaker than the already-abandoned DOT arm (PF 1.058 < 1.11) of this exact family; DOT/AVAX failed the post-opt DSR/holdout gates and LINK was abandoned pre-opt on identical thin-PF/low-capacity/sibling-failed logic

Outcome Summary

The FourHourVolumeBreakoutLong family does not generalize to lower-liquidity alts — its marginal arms (DOT, AVAX, LINK, now FIL) show thin profit factors, Sharpe CIs straddling zero, small capacity, and recent-year decay, and optimize to flashy in-sample Sharpes that fail the forward gates; a pre-opt PF below ~1.1 on a low-capacity alt whose stronger siblings already failed should be abandoned rather than optimized.

The analyst abandoned it at backtest review as the weak, lower-liquidity arm of the family rather than a bug — the code traded exactly as specified, but a PF 1.058 / Sharpe 0.27 edge with a CI straddling zero on a $2.5M-capacity instrument would, after best-of-225 selection, only produce a flashy in-sample number that fails deflated Sharpe and the holdout, exactly as the stronger DOT (PF 1.11) and AVAX arms already did post-optimization and as LINK was abandoned pre-optimization on identical logic.

A long-only volume-confirmed Donchian breakout on FILUSDT 4H perpetual futures (pure OHLCV) — a direct analog of the landed BTC/ETH/SOL/BNB/AVAX/XRP FourHourVolumeBreakoutLong family applied to Filecoin — entering long when price broke above the prior 20-bar high on breakout-bar volume ≥1.5x the 20-bar average, exiting on ATR stop/target, a percentage take-profit fallback, a lower-Donchian reclaim, or a time stop.

Over a 193-trade sample (~all long, none dropped) the edge was marginal: profit factor 1.058, Sharpe 0.265 (CI -1.20 to 1.58), total return +12.5% over 5.7 years (CAGR ~1.3%), expectancy +$38/trade (~0.21%, just above the fee floor), 13.6% max drawdown; it was fee-fragile (capacity only $2.5M at 19.9% impact) and decaying out-of-sample (2025 -2.9%, 2026 -3.2%, rolling Sharpe -3 to -4 through 2026).

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