Time-Series Momentum in Futures — a backtest with Sharpe near 1.0 that live CTAs never matched

Time-series momentum (TSMOM) — go long a futures contract after 12 months of positive excess return, short after negative — produced an academic Sharpe near 1.0 across 58 markets from 1965–2009. Live CTA trend-following funds tracking the same idea since 2000 have delivered Sharpe 0.61. This page holds both numbers side by side: the backtest that made the case, and the investable reality that followed it.

The backtest: Moskowitz-Ooi-Pedersen 2012

Moskowitz, Ooi & Pedersen (*Journal of Financial Economics*, 2012) tested a signal on 58 liquid futures contracts — 24 commodity, 12 currency pairs, 9 equity index futures, 13 government bond futures — over 1965–2009 (44 years).

Signal

long positions on assets with positive excess return over the trailing 12 months, short on negative — position size scaled inversely to volatility using GARCH estimation, rebalanced monthly.

Results, diversified portfolio across all 58 markets

- Annualized alpha (Fama-French adjusted): 20.7% - Sharpe ratio: ~1.0 — the figure consistently cited for this diversified portfolio (Foxholm's citation of MOP 2012 reports "about 1.0"). An earlier draft of this page cited 1.31, sourced to a Quantpedia derivation; that figure could not be reconciled with the reported 15.74% volatility (it implies a ~1.03 Sharpe, not 1.31) and is superseded here as unverified. - Annualized volatility: 15.74% - Maximum drawdown: -33.87% - Excess return: >1%/month across the portfolio, 1985–2009

Breadth

all 58 contracts individually showed positive predictability from their own trailing 12-month return; 52 of 58 were statistically significant at the 5% level — this is not one or two markets carrying the result.

A single-strategy Sharpe near 1.0 for a diversified TSMOM portfolio combining all 58 markets is the number most often quoted as "the" backtest result.

Post-publication decay and a contested confound

McLean & Pontiff (2016, "Does Academic Research Destroy Stock Return Predictability?") found post-publication returns run about 50% smaller than in-sample returns — a general anomaly-decay finding that alphaarchitect.com applies to TSMOM specifically. On that same trend-following literature:

- Each year further from original publication adds roughly 5 percentage points of Sharpe decay; publication year alone explains ~30% of the variance in that decay. - After June 2010, the Sharpe ratio for both TSMOM variants tested dropped below 0.7. - Applied to actual trend-following ETFs (not the academic backtest universe), in-sample Sharpe was already only 0.1–0.2, and out-of-sample performance went negative.

(unverified, contested) Some researchers argue the MOP 20.7%-alpha / Sharpe-near-1.0 result is largely an artifact of the volatility-scaling in the position-sizing rule rather than genuine time-series momentum: strip out vol-scaling and TSMOM's alpha converges toward plain buy-and-hold. This is flagged in the source as a research interpretation, not a settled finding — treat the original MOP numbers as upper-bound estimates until this is resolved.

Live CTA track record vs the backtest

The Société Générale CTA Index is the standard proxy for what trend-following managed futures funds actually delivered, live, net of the strategies' implementation choices (though the disclosed figures are gross of hedge fund fees):

- 2000–May 2024 annualized return: 4.8% - 2000–May 2024 Sharpe ratio: 0.61 (vs. S&P 500 Sharpe 0.44 over the same window — CTAs still beat equities risk-adjusted, just far below the backtest) - 2022 (a "crisis alpha" year, equities down): SG CTA Index +27.3%, its best year since the index's 2000 inception - 2022 dispersion across individual managed futures funds: -7.28% to +58.09% — the index average hides enormous manager-level variance - Managed futures AUM: ~$350B of the ~$7T total hedge fund universe

The live Sharpe of 0.61 sits well below the 1965–2009 MOP backtest Sharpe of ~1.0 — roughly a 40% haircut. This gap (backtest to live-fund reality) is the single most important number on this page for anyone using TSMOM as a live signal rather than a historical curiosity; it is an inferred comparison across two different sources (returnstacked.com's SG CTA data and quantpedia's MOP extraction), not a single study's head-to-head test, so treat the ratio as directional rather than exact.

Diversification value

managed futures show empirically low, often near-zero-or-negative rolling 12-month correlation to equities, but that correlation is regime-dependent — meaningfully positive in equity bull markets (2006–2007) and meaningfully negative in equity bear markets (2008). Adding a 20% managed-futures allocation to a 60/40 stock/bond portfolio improved risk-adjusted returns over January 2000–May 2019, and the funds performed well specifically when equity drawdowns exceeded 20%. The asymmetric, drawdown- triggered negative correlation (not a stable constant correlation) is the mechanism behind that diversification benefit.

Transaction costs and capacity limits

TSMOM in liquid futures is cited as surviving realistic costs: 0–3 basis points of slippage/execution cost is the range found in the transaction-cost literature for well-implemented futures TSMOM, and the strategy is reported to remain profitable net of that cost band. See Transaction Cost Accounting — the arithmetic that separates a real edge from a paper one for how that number is built up market by market.

Capacity is the binding constraint once size grows, and it is worse in commodities specifically:

- The top 10 commodity futures markets hold ~70% of total commodity futures liquidity; energy alone is 55–65% of it. - A $1B commodity allocation forced into a concentrated subset by capacity limits showed a 17% Sharpe deterioration versus the diversified version. - In basis-point terms: roughly -1.6% annual performance drag per unit of portfolio volatility (e.g., a 12%-vol program loses ~1.6%/year) when capacity forces suboptimal diversification. - Across the full 69-commodity-futures universe, average per-market Sharpe is only 0.15–0.20 (±0.1) — the diversified-portfolio Sharpe of ~1.0 depends on combining many weak, largely uncorrelated individual bets, not on a few strong ones. - Commodities contributed ~50% of trend-following CTA returns over the five years through December 2024, despite being the segment most exposed to this liquidity concentration. - Trend-following ETFs, which are structurally limited to a subset of available markets, face an estimated ~4%/year drag before fees purely from forced concentration — a large chunk of the backtest-to-live gap documented above traces to exactly this mechanism.

Why this page matters for the wiki's objective

TSMOM is the cleanest worked example in this wiki of the gap the objective asks every page to quantify — backtested Sharpe ~1.0 vs. live Sharpe 0.61, a ~40% haircut that is explained, piece by piece, by post-publication decay, forced concentration under capacity limits, and (per the contested volatility-scaling critique) possibly by an inflated original estimate. Anyone asking "does trend- following actually work" gets a number-backed answer here instead of a backtest quoted in isolation.

What does NOT work

- Trend-following ETFs as a retail proxy for the backtest: in-sample Sharpe of only 0.1–0.2 even before going out of sample, then negative out-of-sample — the forced-concentration mechanism above is the documented cause, not bad luck. - Reading the MOP 20.7%-alpha/Sharpe-near-1.0 numbers as what a live portfolio would earn today: the post-2010 Sharpe for the same signal family dropped below 0.7, and the SG CTA Index's realized 2000–2024 Sharpe is 0.61 — under half the original backtest. - Treating TSMOM as a pure momentum effect independent of volatility scaling: unverified but flagged — the position-sizing rule itself may be doing much of the work.

Related

- Transaction Cost Accounting — the arithmetic that separates a real edge from a paper one — the 0–3bp survival threshold cited above comes from the same cost-accounting discipline this page's companion concept page builds out in general. - Out-of-Sample vs Post-Publication Decay: The Two Numbers That Tell You If a Premium Is Real — the McLean-Pontiff decay mechanism applied here to TSMOM is the general case this page is one instance of. - Cross-Sectional Momentum in Equities — the strongest documented anomaly, and how much of it survives costs — the equity-market sibling of momentum; compare its ~10%-to-~2% post-1990s decay against TSMOM's backtest-to-live Sharpe halving to see whether momentum decay is a single phenomenon or two separate ones. - Moving-Average Crossover Rules in Equities — widely claimed, thinly sourced: what we could and could not verify — a superficially similar trend-signal in equities that mostly did NOT survive out of sample; the contrast is instructive on why futures TSMOM fared better. - Computing a Strategy's Transaction-Cost Threshold — the four-step check that decides whether a documented edge is tradable — for computing whether a specific TSMOM implementation clears the 0–3bp survival band documented here.