C+

Risk Architecture & Market Regimes

Trend Following (Time-Series Momentum)

Method dossier — documented in AQR and Man AHL research · documented since 2012

The most openly documented hedge-fund method there is: go with the sign of the trailing trend, size by inverse volatility, rebalance monthly. A century of public evidence, and simple enough to run as software on your own data.

The big picture

Managed-futures funds have run one core engine for decades, and unusually, they published it: if an asset's trailing return over roughly the last year is positive, hold it long; if negative, stand aside or go short — with each position scaled inversely to its volatility. The canonical academic write-up (Moskowitz, Ooi and Pedersen, 2012) and the long-horizon replication (AQR's century study) put the rules, the evidence and the caveats in the open. The method's core bet: markets underreact first and herd later, so trends persist longer than efficient prices should allow.

Why it matters now: trend is the classic complement to a concentrated long book — it tends to earn its keep exactly in the extended sell-offs that hurt buy-and-hold most, which is why the industry calls the payoff profile crisis alpha. For an AI-heavy portfolio, that convexity is worth understanding even if you never run the system.

The trend-following smile — why the method pays at the extremes EQUITY MARKET RETURN → TREND SYSTEM RETURN CRASHES: SHORT THE TREND DOWN MELT-UPS: LONG THE TREND UP THE CHOP: SMALL, FREQUENT WHIPSAW LOSSES the entry fee for both tails sign(12-month return) decides direction · target vol / realized vol decides size · monthly · no discretion WHOEVER CANNOT SIT THROUGH THE MIDDLE DOES NOT COLLECT THE ENDS
The documented payoff of time-series momentum: a smile against the equity market — earned at both extremes, paid for in the range-bound middle.

The 3 strategic pillars

  1. The signal

    One number decides: the sign of the trailing 12-month return (measured to the prior month-end, skipping nothing fancy).

    Positive → long, negative → flat (long-only variant) or short (full variant). Re-evaluated once a month; no discretion, no exceptions — the discipline is the method.

  2. Volatility-scaled sizing

    Every position is scaled so it contributes similar risk — the trend decides direction, volatility decides size.

    Weight = target volatility / realized volatility of the asset. Calm assets get more capital, wild ones less; portfolio risk stays roughly constant across regimes.

  3. The cost and the convexity

    Trend loses small and often in range-bound markets (whipsaw) and wins large and rarely in sustained moves — in both directions.

    The payoff plotted against equity returns forms a smile: profits in crashes AND melt-ups, bleed in the chop between. Whoever cannot sit through the bleed does not collect the smile.

What a Closelook reader does with it

The working use is twofold. As a lens: the 12-month sign on your own holdings is a one-line regime check — fighting a negative trend is a choice you should at least make consciously. As a system: the pack's backtester runs the full mechanical rulebook on any price series you feed it, so you can see the smile, the whipsaw years and the drawdown profile on your own data before believing anything. The mistake the method prevents is structural: riding a broken trend all the way down because conviction outvoted the tape.

The bridge to the Closelooknet approach

Closelooknet already runs momentum machinery in production: the CL-*-AM index line is a monthly momentum overlay on our own indices — a cousin of this method applied cross-sectionally. The structure block's regime score on every asset page (trending vs. mean-reverting) answers the method's operating question — trend systems earn in exactly the regimes that score high there — and the Market Regime framework is the discretionary version of the same read. Run the pack on the tickers you follow and compare its signal flips with the regime score: the two should disagree exactly where the tape is transitioning.

Action-Kit — from theory to practice

Tooling & data

What you needWhere to get itCost
Daily or monthly price history for YOUR instruments The only input the system needs — this pack ships no data and no signals; you connect your own source Stooq (free CSV downloads), your broker's export, or any data subscription you already hold A total-return series is cleaner than price-only; for indices, use the TR variant where available. Free
The published evidence Read the caveats from the source before trusting any backtest, including ours AQR research library — Time Series Momentum (2012), A Century of Evidence on Trend-Following (2017) Free

The formulas

  • Signal

    S_t = sign( P_{t-1} / P_{t-13} − 1 ) (monthly data)
    • P — month-end prices; the window ends at the PRIOR month-end

    The 12-month lookback is the canonical choice; the published result is robust across 3–12 month windows.

  • Position size

    w_t = S_t × (σ_target / σ_t)
    • σ_target — your annualized risk budget per asset (the papers use 40% for futures; equities need far less)
    • σ_t — trailing realized volatility, e.g. 60-day annualized

    Cap w at 1 (no leverage) in the cash-account variant the pack defaults to.

  • Whipsaw diagnostic

    Flips per year = count(S_t ≠ S_{t−1}) / years
    • The signal series

    Above ~4 flips/year on an instrument, trend is paying the market, not you — the regime score tells the same story from the other side.

Applied Pack · free members

Trend System Pack

A complete mechanical trend system as software: your data in, the full backtest out — signals, vol-scaled sizes, equity curve, drawdowns, whipsaw count. We ship rules and code, never signals.

  • Trend_System.xlsx — monthly close series in, 12-month signal, vol-scaled weight and system-vs-buy-hold comparison out, all as live formulas
  • trend_backtest.py — stdlib-only backtester: feed any daily-price CSV (your own source), get the full mechanical run — CAGR, volatility, max drawdown, flips/year, yearly table — for the long-flat and long-short variants
  • prices_sample.csv — synthetic example series showing the expected format (date,close per instrument)
  • README.txt — the rulebook in one page, parameter guidance from the published research, and the software-not-signals disclaimer

Pack security

Macro-free Excel · plain-text Python you can read before you run it · no installers, no network access — the code works only on files you provide. Served only from closelook.net; we never distribute through download portals or email attachments. How to verify in 30 seconds →

SHA-256 55d1c40c8eb511209c82dcb0cd500dff63503651c9153f5d52f258024772e7f5

The pack is a free-membership download — no card, free forever. Membership also delivers the Daily Pulse and Weekly Signal to your inbox.

Get free access →

Closelook publishes a market diary, not investment advice. This condensed read restates the book's ideas in our own words for education — for the author's full argument, go to the source.