Week 51: Managed Futures and Trend-Following — Diversifying Alpha and Crisis Convexity

---

1. Why This Is Important

If you have read Week 47 you already know one of the strangest facts in asset management: a trend-following CTA index returned roughly **+14% in 2008 while the S&P 500 lost −37%** and the dutifully-balanced 60/40 lost −22%. In 2022 the SocGen CTA Index ran +20% while 60/40 had its worst calendar year since 1937. In March 2020, with the S&P down a third in twenty trading days, the same index was up mid-single-digits. The pattern is not luck. It is the structural signature of trend-following — the strategy that almost the entire quantitative CTA industry runs in some form.

You need to understand managed futures for four reasons:

**It is the only systematic strategy with reliable crisis

convexity.** Tail-hedging via puts (Week 47) buys explicit convexity for a known premium drag. Trend-following manufactures convexity from price action: as markets break down it goes short, so a sustained bear market is exactly when its P&L compounds. Different mechanism, similar shape, much smaller carry cost. That is rare enough that it deserves its own sleeve in any serious portfolio.

It is genuinely uncorrelated to stocks and bonds. The CTA-vs-

S&P long-run correlation is roughly zero, with the conditional correlation turning sharply negative in extended drawdowns. This is the property the 60/40 lost in 2022 and that bonds lost in 2008-by-default. A barbell only works if the two ends are actually different. CTAs are the cleanest "different" you can buy in liquid public markets.

It is now retail-accessible at institutional cost. For three

decades trend-following was 2-and-20, $5M minimums, gated. Today DBMF, KMLM, FMF, and AHLT replicate the SocGen CTA Index at roughly 0.85 to 0.95% all-in expense ratios with daily liquidity. The fee compression is the single biggest change in liquid alternatives over the last five years.

It is the textbook L5 alpha sleeve. Of the small handful of

places alpha can come from, trend-following is the most empirically validated at the systematic end (alongside size and value), with out-of-sample evidence going back to 1880 across futures markets that did not yet exist when the studies were funded. The strategy is not a secret — and it still works, because the source of the return (the volatility tail wagging the dog plus cross-asset momentum) is hard for marginal capital to arbitrage away without accepting the same drawdowns CTAs accept in chop markets.

This week's job is to wire that into a portfolio, not just to admire it from a distance.

---

2. What You Need to Know

2.1 What "Trend-Following" Actually Is

A textbook trend-follower is a rules-based system that goes long an asset when its recent return is positive and short when it is negative, sized inversely to that asset's recent volatility. The canonical version is time-series momentum: at month-end, for each of roughly 50–80 liquid futures contracts (S&P, Nasdaq, FTSE, JGB, Bund, US 10-year, EUR, JPY, gold, copper, crude, corn, …), compute the 12-month total return; if positive, hold a long position; if negative, hold a short. Each contract is sized so that it contributes the same risk in volatility units to the portfolio. Lookbacks are usually a blend (1-month, 3-month, 12-month) and the trade is rebalanced daily or weekly.

That is the entire algorithm. There is no forecast of fundamentals, no view on Fed policy, no DCF. The strategy says: *if this thing has been going up, ride it; if it has been going down, ride that too*. It is momentum applied across asset classes, with the key innovation being that the universe is not just stocks. When stocks fall and bonds rally and the dollar squeezes and crude tanks, all four trades line up in the same direction, and the convex payoff in a crisis is the sum of those four legs.

2.2 The "Crisis Alpha" Pattern

Sébastien Page and the AQR research team coined the phrase *crisis alpha* for what trend-following does in extended bear markets. The mechanism is mechanical: most large equity drawdowns are not single days — they are trends. The S&P 500 drawdown of 2007–2009 played out over 17 months. The 2000–2002 drawdown over 31 months. 2022 over 9 months. Trend-followers caught all three, because each of those drawdowns was preceded by a regime in which most futures across asset classes were trending in the same direction long enough for the 12-month signal to flip and the position to size up.

The flip side is sideways or chop markets, which are bad for trend. 2011-2015 was the longest stretch of CTA underperformance in 40 years — five years of roughly zero return while equities ran 70%. 2014 was the redemption year (commodity trends finally re-engaged), 2015 was flat, 2016-2019 mixed. Then 2022 paid back the entire 2011-2019 gap in a single year for the index.

That is the deal: in calm markets you pay a small or zero opportunity cost; in extended trending crises you get paid a multiple of your allocation. The few months that matter dominate the long-run return distribution, and trend-following is structurally long those months.

2.3 The SocGen CTA Index and What "Average CTA" Means

The SocGen CTA Index (formerly Newedge CTA Index) is the industry benchmark: equally-weighted composite of the 20 largest reporting CTA managers, rebalanced annually, AUM-weighted in older versions. It goes back to 2000. The trend-following sub-index goes back to 2000 as well, but proxy series stitched from BarclayHedge and other databases extend the picture to 1990.

A few calendar-year landmarks since 2000:

Year	SocGen CTA	S&P 500 (TR)	Notes
2002	+18.3%	−22.1%	dot-com tail, USD weakness, bond rally
2008	+13.1%	−37.0%	crisis alpha, all asset trends
2011	−4.4%	+2.1%	euro-debt chop, no follow-through
2014	+15.7%	+13.7%	dollar surge, oil collapse, JGB rally
2015	0.0%	+1.4%	reversal year, wrong-footed in Q3
2018	−2.9%	−4.4%	feb Volmageddon + Q4 chop
2020	+1.9%	+18.4%	COVID round-trip neutralised trend
2022	+20.5%	−18.1%	dollar/rates/energy all trending

The arithmetic mean since 2000 is roughly 4–5% with about 10–11% volatility — Sharpe a hair under 0.4. That is much lower than US equities' Sharpe, and that is the point: you are not buying CTAs for return-density. You are buying them for the shape of the return, specifically the negative-correlation-when-it-matters property.

2.4 Capacity, Fees, and the Retail Vehicles

Until ~2019 the only way to buy this strategy was a direct allocation to a CTA fund: $1M to $5M minimums, 1+10 to 2+20 fee schedules (typical: 1.5% management + 17.5% performance), monthly liquidity at best. The good systematic CTAs (AHL, Winton, Aspect, Campbell, Graham, Millburn, Transtrend) ran tens of billions and delivered the bulk of what the index showed.

The 2019 launch of DBMF (iMGP DBi Managed Futures Strategy) was the inflection. The fund replicates the SocGen CTA Index using a 40-factor regression on the index returns themselves, then runs a mirror portfolio of liquid futures that targets the same factor loadings. The result is roughly 0.85% all-in (no performance fee, no lockups), daily liquidity, 1099 reporting (no K-1), Section 1256 60/40 capital gains tax treatment on the futures sleeve.

The current retail CTA menu, April 2026:

Ticker	Vehicle	ER	AUM	Style
DBMF	iMGP DBi Managed Futures	0.85%	~$1.6B	Index-replication of SocGen CTA
KMLM	KFA Mount Lucas Managed Futures	0.92%	~$0.6B	Long-only-trend on commodities-FX-rates
FMF	First Trust Managed Futures	0.95%	~$0.3B	50/30/20 equity/comm/FX trend
AHLT	AHL Trend ETF (Man Group)	0.95%	~$0.4B	Direct-from-AHL trend, multi-tenor
AQR Managed Futures (mutual fund)	QMHIX/QMHRX	1.18%	~$3.8B	AQR full-strength trend, mutual-fund wrapper

The mutual fund version of AQR's strategy (QMHIX) is the closest thing to "buy AQR's trend desk in your IRA." DBMF is the cheapest proxy for the index average. KMLM has the longest standalone live record of the ETFs (since 2020). For most investors at the start, a single position in DBMF or KMLM at 10–15% of portfolio is the right default.

2.5 Why Trend Is "Long-Vol" Even Though It Buys No Options

A trend-follower never buys a put or a call. It also has no convexity built into a single trade — a long futures position is linear in the underlying. The convexity comes from the signal: as prices move enough, the position is added to (more long if rallying hard, more short if collapsing). If a drawdown lasts long enough for the signal to flip from long to short, the realised P&L on the short position grows quadratically with the size of the move — same qualitative shape as a put.

The mathematical sketch: if you imagine a perfect trend-follower that always holds the right sign at every instant, the P&L is the integral of |dS|, which is path-dependent and dominated by big moves. That is the signature of a long-straddle — a long volatility position. Trend- following synthesises a straddle out of price changes, paying for it in chop markets (when the signal whipsaws) rather than option premium (when the underlying drifts).

Practical implication: trend-following is correlated to realised volatility, not implied volatility. It does badly when realised vol collapses (2017, 2019), well when realised vol stays elevated and directional (2022). In years with high implied vol but no follow- through (2018), it can lose despite VIX being elevated.

2.6 Where Trend Sits in the Four-Tranche Framework

Slot the strategy into the four-tranche structure:

Tranche	Target	Role of CTA sleeve
Growth (60–70%)	Stocks: VTI, factor tilts	unchanged
Income (10–20%)	Bonds, JEPI, MLPs	unchanged
Stores of Value (5–15%)	Gold, T-bills, USD cash	partially overlaps with CTA
Opportunistic / Tactical (5–15%)	CTA + tail hedges + alt sleeves	here

For a 100k portfolio with the L5 default 80/10/10/0:

• 80k VTI/VOO,
• 10k bonds,
• 10k DBMF or KMLM as the "stores of value plus opportunistic"

blend,

• 0–5k SPY OTM puts if you want explicit tail (Week 47).

A more aggressive split is 70/10/15/5 with the 15 being entirely CTA. That is approximately the AQR house-portfolio recommendation for endowment-style mandates.

The wealth-path chart shows the empirical reason. From 1990 to April 2026 a 100% S&P portfolio compounds to the highest terminal value but has the worst drawdown profile (−51% in 2009, −24% in 2022). A 60/40 has lower terminal but a less severe drawdown (−30% / −20%). A 50/30/20 with 20% in a synthetic CTA proxy lands between the two on terminal wealth but with the best drawdown profile of the three (−22% / −10%) — exactly what you would expect from a sleeve with −0.2 unconditional correlation that goes more negative in crashes.

2.7 Sizing, Fees, and the Behavioural Trap

Three rules of thumb I have used myself:

Size for the chop, not the crash. The right question is

"how big a position can I hold in 2015 when this returns 0% for the year while my friends are up 20% in stocks?" If your answer is 5%, hold 5%. If you size for the 2008-style payoff and then redeem in the chop, you get the worst of both worlds.

Pick one ETF and hold it. Switching from DBMF to KMLM to

AHLT based on YTD performance is a disaster — you are buying the manager who just had the lucky year. The dispersion across CTA indices in any given year is tiny (most are within 3% of each other) so the choice does not matter as much as the stickiness.

Treat the carry as a permanent cost. The strategy can have

3-year stretches of zero return. If you are not prepared to wear that, the position is not the right one for you. Better to size smaller and never sell than to size big and capitulate.

The behavioural trap is identical to the one with tail hedges (Week 47): the strategy looks broken right when it is closest to paying off. Markets stay irrational longer than you stay solvent — that applies to the holder of the hedge as much as it does to the trader on the other side of one.

---

3. Common Misconceptions

"CTAs don't work anymore — the trend factor is dead." The

2011–2019 stretch made this a popular thesis. Then 2022 happened and the index returned 20%+ for the year, on the same model the "dead" critics were criticising. The trend signal is structurally tied to drawdown duration, and drawdowns have not stopped.

"It's the same as just buying VIX or puts." No. VIX/UVXY are

long implied vol; trend is long realised vol. They behave differently in a 2018-style implied-vol shock without follow- through (UVXY +200% in February, trend −2%). Tail puts pay on speed; trend pays on duration.

"It's a hedge — so it should always make money in a bad year."

No. It needs the bad year to be a trend, not a single-day or single-week shock. February 2018, August 2015, August 2024 are all years CTAs lost money in equity-stress months because the move was too fast for the signal to engage.

"DBMF / KMLM should track AQR's mutual fund tightly." They

target the index (SocGen CTA), not any one manager, so manager dispersion shows up as tracking error. The right benchmark is the index, not another product.

"You can replace your bond allocation with CTAs." Bonds and

CTAs are diversifiers in different regimes — bonds in deflationary/growth shocks (2008, 2020), CTAs in stagflationary/ trending shocks (2022, 1970s). A mature portfolio carries some of each, not just one.

"The fees aren't worth it — I should run trend myself." You

can, on the micro futures (Week 39 — /MES, /MNQ, /MCL, /MGC). The capital required to run 30 contracts at the right vol-target is roughly $300k+, the bookkeeping is daily mark-to-market, and the operational risk is real. For most retail allocations DBMF at 85bp is the right call.

**"It worked in 2008 because of the GFC — that won't happen

again."** It worked in 1973–74, 2000–02, 2008, 2014 dollar shock, 2020 COVID, 2022 inflation regime change. That is six distinct macro environments over 50 years. The mechanism — long- duration trends across uncorrelated markets — has not been regime-specific.

"Trend is 'short gamma' because it sells into rips." No. Trend

buys strength and sells weakness — that is long gamma in the options-Greek sense. The whipsaw cost in chop is the gamma cost you pay every period for the convex payoff in a sustained move.

"I can just hold managed futures inside a 60/40 fund." Mutual

funds with "managed futures" sleeves often run 5–10% allocations that are too small to move the line. To get the 2008/2022-style convexity you need 15–20% in a pure CTA, not a slice of a target-date fund.

---

4. Q&A Section

Q: How big should the CTA sleeve be? A: For a default L5 portfolio, 10–15% is the right range. The lower end is a hedge, the upper end starts to materially change the drawdown profile. Above 20% you are actively expressing a trend- follow-as-a-return-source view, which is fine but requires more conviction in chop years.

Q: DBMF, KMLM, FMF, or AHLT? A: Default DBMF for the cheapest index replication, KMLM if you want a slightly longer-only-commodity tilt, AHLT if you want direct AHL exposure. The dispersion among them is small in most years.

Q: What about AQR's QMHIX mutual fund? A: It is the most "pure" version (full-strength AQR trend, no ETF replication overlay), but ER is 1.18% and minimums apply at some brokerages. For tax-deferred accounts (IRA, 401k where available), QMHIX is excellent.

Q: Section 1256 tax treatment — does it pass through DBMF? A: Yes. DBMF holds futures directly; gains are 60% LTCG / 40% STCG regardless of holding period at the fund level, and that flows through to your 1099 (no K-1). This makes CTAs tax-efficient even in taxable accounts.

Q: Does CTA pair well with my SPX-puts tail hedge from Week 47? A: Yes — they are complements, not substitutes. Puts pay on a fast shock (March 2020); CTA pays on a sustained trend (2022). A 5-15% CTA + 0.5-1% puts-rolled-quarterly is a robust crisis-convex sleeve.

Q: What's the worst-case scenario for trend-following? A: Sharp reversal at the end of a long trend — September-October 2008 caught some funds wrong-side as commodities collapsed faster than positions could flip; February 2018 caught funds long after years of low-vol uptrend. Three- to nine-month drawdowns of 10-15% are normal. Drawdowns of 20%+ have happened roughly once a decade.

Q: Should I rebalance the CTA sleeve? A: Yes — annually or at 5%-band thresholds, like any other sleeve. Rebalancing into a CTA in a chop year (say end-2015 or end-2019) is exactly when it has historically been most additive going forward.

Q: What about long-only commodities — same thing as CTA? A: No. Long-only commodities (DBC, PDBC) are passive futures exposure with roll cost (Week 39). CTA can go short commodities, which is the entire point. In 2014 commodities fell 30%+ and CTAs made money on the short leg; long-only got crushed.

Q: Is the strategy "crowded"? A: Total CTA AUM is about $400B globally, vs $50T+ in liquid global futures markets. The market-impact cost of the strategy is real but manageable, and academic studies (Hurst-Ooi-Pedersen 2017, AQR 2022 update) find no decay in the trend factor through 2024. Still working, even after the audit.

**Q: How does CTA fit with private alts / hedge funds in a fancy portfolio?** A: Most institutional portfolios already have CTA via their hedge- fund sleeve, and may double-count. The Yale endowment, for example, holds ~25% in absolute-return / hedge funds, of which a meaningful fraction is trend-style. For retail, DBMF at 10-15% replicates that exposure cleanly.

**Q: Why does the SocGen CTA Index look so much better than the HFRI Macro Index?** A: HFRI Macro mixes discretionary global macro (which has underperformed for a decade) with systematic trend. SocGen CTA is the trend-only sub-index. Same reason the SP500 looks better than "all stocks" — composition matters.

Q: What's the interactive for this week? A: A blender. Slide your % to S&P, bonds, and CTA; it backtests against real annual returns 1990-2024 plus the synthetic CTA proxy and shows you CAGR, vol, Sharpe, max drawdown, the 2008 and 2022 calendar returns, and the correlation matrix. The point is to see what the −22% 60/40 of 2022 turns into when you swap 20% of it into a CTA proxy. Spoiler: roughly −5%.