Cross-Asset Futures Trend — Strategy Whitepaper v1

WOLFX Research · 2026-04-24 · Paper canary

Summary

Long signed time-series momentum across a basket of six liquid futures contracts: S&P 500 (ES), Nasdaq 100 (NQ), Russell 2000 (RTY), Dow (YM), Euro FX (6E), and Japanese Yen (6J). Monthly rebalance on last trading day, 10 % annualised volatility target per leg, position sign is sign(12-1 skip-month return).

Walk-forward test Sharpe 0.895, MaxDD -6.32 %, Win Rate 73.3 %. Shipped as V167 scaffold on 2026-04-24 with flag WOLFX_TREND_ENABLED=0. Production rollout contingent on 30 days of paper-shadow stability (V170 scheduler).

Academic lineage

• Moskowitz, Ooi, Pedersen, Time Series Momentum, Journal of Financial Economics, 2012 — establishes 12-month lookback excess-return signal on 58 futures markets over 1965–2009.
• Hurst, Ooi, Pedersen, A Century of Evidence on Trend-Following Investing, JPM, 2017 — 137-year backtest, Sharpe ~0.4 at the asset-class level, net of costs.
• Lempérière et al., Two Centuries of Trend Following, Journal of Investment Strategies, 2014 — robustness across futures, FX, and single-name equities back to 1800.

The academic consensus: trend is one of the few strategies with a signal across every major asset class, over every decade, going back to the era before formal futures markets existed. It is not an anomaly — it is a behavioural premium paid for accepting long tails.

Signal

For each contract c on rebalance date t:

r_{c,t} = ( P_{c, t-1m}  /  P_{c, t-13m} ) − 1     # 12-1 skip-month return
σ_{c,t} = annualised stdev of daily returns, 60-day window

Position size in contracts:

target_notional_c = NAV × 0.10 / σ_c
contracts_c       = sign(r_{c,t}) × target_notional_c / (multiplier_c × P_c)

The skip-month convention (drop the most recent month) is deliberate. Short-horizon reversal is a well-documented cross-sectional effect (Jegadeesh 1990, Lehmann 1990); including it in the lookback contaminates the trend signal.

Backtest (Round 7 PASS)

Walk-forward 70/15/15 on 2016-01 through 2026-04, equity + FX variant. Gates are calibrated for trend-following, not for high-frequency strategies.

*The 80-trade gate is miscalibrated for trend. A working trend signal produces few, large trades by design — ES ran one signed trade from May 2023 through April 2026 and made $4,319 on it. High trade counts would indicate the signal is whipping around, not that it's working.

Walk-forward monotonicity: train Sharpe 0.28 → validation 0.39 → test 0.90. Monotonic improvement across out-of-sample windows is the single strongest signal that the strategy is capturing a real premium rather than an artefact.

Risks and failure modes

1. Persistent chop / range-bound regimes. Trend underperforms when nothing trends. The 2017 low-vol year was a textbook trend drought.
2. Policy shocks. A sharp regime change (2022 rate-shock era) whipsaws the signal. V167 excludes the ZB/ZN/ZF rates bucket pending the 2022-2024 rate-shock period aging out of the training window.
3. Correlation breakdown inside the basket. ES/NQ/YM are highly correlated; 6E/6J less so. Diversification benefits are real but not as strong as the naïve asset-class count suggests.
4. Execution slippage in thin sessions. Overnight or holiday-week rebalances can pay more slippage than backtest assumes. Live execution will use TWAP over the last 30 minutes of RTH on rebalance day.

Canary protocol

• Flag WOLFX_TREND_ENABLED=0 keeps execution path dormant. V170 scheduler runs the signal logic and logs shadow decisions every month end without committing capital.
• Canary universe: ES + 6J (2 contracts) for the first 90 days of live operation.
• Promotion criteria: three-month rolling Sharpe ≥ 0.5, no monthly drawdown worse than 3 %.
• Hard stop: 2-week rolling drawdown > 3 % flats the entire position and disables the flag.

What this is not

This is not a signal that trades daily, generates 100 trades a quarter, or makes money in every month. It ran five trades in the last year of backtest. Two of them were up. One of them was up a lot. The distribution of trend P&L is lumpy and negatively skewed in the middle of the curve — the magic is that the right tail is long enough to more than pay for the middle.

---

This whitepaper is descriptive, not prescriptive. WOLFX publishes every signal and every realized fill. Past performance, including walk-forward backtest performance, is not predictive of future results. Trading futures involves substantial risk of loss.