Paper: arXiv 2511.08571

Authors: Mainak Singha, Jose Aguilera-Toste, Vinayak Lahiri

Abstract

We test whether simple, interpretable state variables-trend and momentum-can generate durable out-of-sample alpha in one of the world’s most liquid assets, gold. Using a rolling 10-year training and 6-month testing walk-forward from 2015 to 2025 (2,793 trading days), we convert a smoothed trend-momentum regime signal into volatility-targeted, friction-aware positions through fractional, impact-adjusted Kelly sizing and ATR-based exits. Out of sample, the strategy delivers a Sharpe ratio of 2.88 and a maximum drawdown of 0.52 percent, net of 0.7 basis-point linear cost and a square-root impact term (gamma = 0.02). A regression on spot-gold returns yields a 43 percent annualized return (CAGR approximately 43 percent) and a 37 percent alpha (Sharpe = 2.88, IR = 2.09) at a 15 percent volatility target with beta approximately 0.03, confirming benchmark-neutral performance. Bootstrap confidence intervals ([“2.49, 3.27”]) and SPA tests (p = 0.000) confirm statistical significance and robustness to latency, reversal, and cost stress. We conclude that forecast-to-fill engineering-linking transparent signals to executable trades with explicit risk, cost, and impact control-can transform modest predictability into allocator-grade, billion-dollar-scalable alpha.

Complexity vs Empirical Score

  • Math Complexity: 4.5/10
  • Empirical Rigor: 9.2/10
  • Quadrant: Street Traders — empirical and applied, lower math complexity

Why this score: The paper is highly data-driven with a detailed walk-forward backtest, specific cost/impact modeling, and robust statistical validation, but the mathematics, while precise, is relatively accessible (mainly Kelly sizing and basic signal smoothing) rather than conceptually dense or advanced.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Identify durable, interpretable alpha in liquid gold futures using simple state variables"] --> B["Data & Methodology<br>2015-2025 OOS Walk-Forward<br>Rolling 10yr Train / 6mo Test<br>2,793 Trading Days"]
  B --> C["Signal Generation<br>Smoothed Trend-Momentum Regime State"]
  C --> D["Position Sizing & Execution<br>Volatility-Targeted (15%)<br>Fractional Kelly Sizing<br>ATR-based Exits<br>Impact-Adjusted Costs: 0.7bps + sqrt(gamma=0.02)"]
  D --> E["Performance & Statistical Validation<br>Sharpe 2.88, Max DD 0.52%<br>CAGR ~43%, Alpha 37% (Beta ~0.03)<br>Bootstrap CI [2.49, 3.27], SPA p=0.000]<br>E --> F['Key Finding<br>Forecast-to-Fill engineering creates allocator-grade, billion-dollar scalable alpha<br>from modest predictability"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a forecast‑to‑fill trading architecture that turns simple trend‑and‑momentum signals into volatility‑targeted, friction‑aware positions in gold futures.

Source sentence
We built a full forecast-to-fill architecture to answer that question empirically, connecting a simple predictive signal to real, executable trades through a reproducible, live-like pipeline.

Assumes 6

  • Linear transaction cost of 0.7 basis points per round‑trip is assumed. execution
    Source sentence
    Linear cost: we assume a round‑trip cost of k = 0.7 basis points (0.007%) per complete buy–sell cycle.
  • Market impact follows a square‑root function with impact parameter γ = 0.02. execution
    Source sentence
    We represent this with a “square‑root” impact function, where the cost increases with the square root of trade size. We set the impact parameter to γ = 0.02, which is conservative for gold liquidity.
  • Target annual volatility is fixed at 15% and leverage is capped at 2.0. model
    Source sentence
    Leverage is capped at Wmax = 2.0, so the system can never double its exposure.
  • Daily settlement prices from LBMA and COMEX accurately reflect market prices. data
    Source sentence
    We use daily settlement prices for gold from the London Bullion Market Association (LBMA) and COMEX. These are the most reliable and liquid data sources available.
Show 2 more
  • Missing days are filled by carrying forward the last known price (not the return). data
    Source sentence
    On holidays or missing days, we simply carry forward the last known price—never the return—so that the sequence of daily moves remains realistic and continuous.
  • Rolling 10‑year training windows are frozen before testing, ensuring out‑of‑sample evaluation. statistical
    Source sentence
    Every parameter is trained on a rolling 10‑year window, frozen before testing, and evaluated strictly out‑of‑sample.

Data

Backtest sample: 2015 – 2025 · Gold futures · futures · daily

  • Gold daily settlement prices · LBMA and COMEX market-prices public
    through September 16, 2025 (out‑of‑sampl · Gold futures contracts
    Source sentence
    We use daily settlement prices for gold from the London Bullion Market Association (LBMA) and COMEX... The

Costs

Modeled Costs are part of the reported headline results.

Linear cost of 0.7 bps per round‑trip and square‑root market impact with γ = 0.02 are incorporated into the net performance figures.

Source sentence
We measure net performance after applying both linear transaction costs (0.7 bps per round trip) and square‑root impact costs with γ = 0.02.

Finds 5

Reported performance (Forecast-to-Fill gold futures trend-momentum strategy): Sharpe 2.88 · annual return 2.6% · max drawdown 0.5%

  • The strategy achieves a net Sharpe ratio of 2.88 with a maximum drawdown of 0.52% over 2,793 out‑of‑sample days. performance
    Source sentence
    We measure net performance after applying both linear transaction costs (0.7 bps per round trip) and square‑root impact costs with γ = 0.02.
  • At a 15% annual volatility target the implied CAGR is 43% and benchmark‑neutral alpha is about 37%. performance
    Source sentence
    the same process implies ∼43% annualized return and ∼37% benchmark‑neutral alpha.
  • Bootstrap 95% confidence interval for Sharpe is [2.49, 3.27] and SPA test p‑value is 0.000, indicating statistical significance. statistical
    Source sentence
    SPA tests (p = 0.000) confirm statistical significance.
  • Capacity analysis shows positive expected growth up to roughly $0.8–1.0 billion of deployable capital, about 0.07% of CME gold futures daily volume. performance
    Source sentence
    positive expected log‑growth up to approximately $0.8–1.0 billion USD of deployable capital—corresponding to only ∼0.07% of CME gold futures daily volume.
  • Active‑day expectancy is +2.58 bps per active day, translating to an annualized return of about 2.63%, matching the realized CAGR. performance
    Source sentence
    The resulting 2.63% annualized return matches the realized CAGR (2.65%).

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 64 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 1 paper in the archive (1 by independent authors)

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (24)

  1. Volatility‐Managed Portfolios (2017) — link
  2. High Frequency Trading and Market Quality (2017)
  3. The probability of back-test over-fitting (2016) — link
  4. RiskMetrics Technical Document (2015)
  5. High Frequency Trading and Price Discovery (2014) — link
  6. Time Series Momentum (2012) — link
  7. The Kelly Capital Growth Investment Criterion: Theory and Practice (2011) — link
  8. No-dynamic-arbitrage and market impact (2010) — link
and 16 more
  1. A Simple Approximate Long-Memory Model of Realized Volatility (2008) — link
  2. A Test for Superior Predictive Ability (2005) — link
  3. A Reality Check for Data Snooping (2000) — link
  4. Finance: A Selective Survey (2000) — link
  5. Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency (1993) — link
  6. Forecasting, Structural Time Series Models and the Kalman Filter (1990) — link
  7. A new interpretation of information rate (1956) — link
  8. A simple, positive semi-definite, heteroscedasticity and autocorrelation consistent covariance matrix (2014)
  9. … and the Cross-Section of Expected Returns (2013) — link
  10. The Kelly Criterion in Blackjack Sports Betting, and the Stock Market (2008) — link
  11. Optimal execution for portfolio transactions (2006)
  12. Trading and Exchanges: Market Microstructure for Practitioners (2002)
  13. Timeout: the trade has been open for 30 trading days, after which the signal’s predictive power statistically decays
  14. Volatility targeting keeps realized risk constant over time, preventing uncontrolled leverage in quiet markets
  15. Confidence mapping converts that slope into a bounded, interpretable probability that scales exposure naturally
  16. Friction-adjusted sizing ensures that every position has positive expected log-growth after trading costs

0 of 24 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.