Generalized Protective Momentum (GPM) Overview

Generalized Protective Momentum (GPM) is a monthly tactical strategy from Wouter Keller and Jan Willem Keuning. It scores twelve global assets on two things at once: how well each has been performing, and how much it has been moving in lockstep with everything else. The three best scorers share the portfolio's risk allocation. Meanwhile the strategy counts how many of the twelve are scoring positive at all, and the fewer there are, the more of the portfolio moves into Treasuries.

The result is a strategy that avoids the classic failure of momentum ranking, which is loading up on three assets that turn out to be three versions of the same bet. Ranking on performance alone will happily hand you U.S. large caps, U.S. small caps and European equities in the same month and call it a portfolio.

Dual Momentum Systems tracks GPM as the reference line for GPMv, a variant built here that follows GPM's structure with several changes.

The Backstory

The most productive pair in tactical allocation

Keller and Keuning have published more of the tactical allocation strategies people actually follow than anyone else working today. Their catalog runs roughly in this order:

• PAA, Protective Asset Allocation, 2016. Introduced the breadth idea: use the health of the whole universe, not just the assets you happen to hold, to decide how defensive to be.
• GPM, Generalized Protective Momentum, also 2016. PAA's breadth machinery with a correlation-aware scoring method bolted on.
• VAA, Vigilant Asset Allocation, 2017. Faster, harsher breadth. One bad asset in a small universe sends the whole portfolio defensive.
• DAA, Defensive Asset Allocation, 2018. Introduced the canary universe: a couple of assets watched purely as an early warning, never held.
• BAA, Bold Asset Allocation, 2022. Canary signals with a more aggressive risk side.
• HAA, Hybrid Asset Allocation, 2023. A deliberately simpler model aimed at retail investors, and probably their most widely followed.

Keller has also co-authored work outside that line, including Elastic Asset Allocation with Adam Butler, which matters here because GPM borrows its scoring idea from it.

GPM is the odd one out

Most of their strategies arrived as formal papers on SSRN, numbered and citable. GPM did not. It appeared in mid-2016 as a Seeking Alpha article and a companion post on Keuning's blog, building directly on the PAA paper published a couple of months earlier.

That difference in packaging is probably why GPM is the least famous of the set despite being one of the more interesting ideas in it. It is a working note rather than a paper, and working notes do not get cited.

What "generalized" means

The name is not marketing. PAA scored assets with one simple measure of trend. GPM generalizes that scoring step: instead of asking only how an asset has performed, it asks how an asset has performed relative to how much it duplicates the rest of the portfolio.

The mechanism comes from Elastic Asset Allocation, where Keller and Butler combined momentum, volatility and correlation into a single number. GPM keeps only the return and correlation parts, on the reasoning that those two carry most of the benefit. Keuning proposed two ways to combine them, one multiplying and one dividing; the version implemented here is the multiplying one.

Why this site carries it

GPM is on this site as a strategy to compare to DMS strategies as well as because it is the control for GPMv. When a strategy here modifies a published model, the published model needs to be running alongside it on the same data, with the same trading costs and over the same period, or the comparison is a claim rather than a measurement. Every number in the table below is produced by the same engine on the same day.

Core Strategy Logic

GPM is evaluated once a month. Signals come from month-end data and set the holdings for the following month. There are three moving parts.

Part 1 - Score every asset

Each asset gets a momentum figure first: take its returns over the past 1, 3, 6 and 12 months and average them, with each period counting equally. Nothing exotic, just a blend that keeps one unusual month from dominating.

Then comes the correlation adjustment. For each asset, measure how closely its monthly returns over the past year tracked the average of the whole risk universe. An asset that moved almost identically to the pack gets most of its momentum figure taken away. An asset that moved on its own schedule keeps nearly all of it, and one that moved opposite the pack can score higher than its raw momentum.

The practical effect: a strong performer that is just another slice of the same trade gets marked down, and a strong performer that is doing something genuinely different gets rewarded. An asset with negative momentum scores badly no matter how independent it is, so the adjustment can never talk the strategy into buying something that is falling.

Part 2 - Let breadth set the risk dial

Count how many of the twelve risk assets have a positive score.

• Six or fewer positive: the portfolio goes fully defensive. Nothing in risk assets.
• Seven through twelve positive: the defensive share is the number of non-positive assets divided by six. Nine positive means three are not, so half a sixth times three, which is 50% defensive. Twelve positive means nothing defensive.

The important idea is that the dial is set by the whole universe rather than by what the strategy happens to own. Broad deterioration pulls money out of the market even when the three assets being held still look fine, which is usually earlier than a strategy watching only its own positions would react.

Part 3 - Fill the two halves

The risk portion goes into the three highest-scoring assets, split equally.

The defensive portion goes into either short-term Treasuries or intermediate-term Treasuries, whichever of the two scores better. That choice matters more than it sounds: in a falling-rate environment the longer duration usually wins and adds return, while in a rising-rate environment the strategy can retreat to the short end instead of being forced to hold bonds that are losing money.

The portfolio is rebalanced monthly.

The universe

Category	Assets
U.S. equities	large cap (SPY), Nasdaq 100 (QQQ), small cap (IWM)
International equities	Japan (EWJ), Europe (VGK), emerging markets (EEM)
Real assets	gold (SGOL), commodities (PDBC), real estate (VNQ)
Credit and duration	high yield (HYG), corporate bonds (LQD), long Treasuries (TLT)

Defensive: short-term Treasuries (SHY) or intermediate Treasuries (IEF).

Note that long Treasuries appear as a risk asset while shorter Treasuries serve as the defensive holding. That is intentional. Long Treasuries are volatile enough to be worth owning for return when they are trending, and their low correlation to equities means the scoring method tends to favor them at exactly the moments equities are struggling.

What GPMv Changes

GPMv is the variant built here, and it keeps GPM's three-part structure while changing four things: a modernized fund universe with emerging markets dropped, and Japan swapped out for Asia, a different defensive instrument, a defensive bias that shifts the risk dial one step earlier, and an override that forces the shortest-duration holding when intermediate Treasuries are themselves falling. The full treatment is on its own page.

Performance Highlights

Over the full published history, alongside its parent strategy PAA and the variant built here:

January 1980 through August 2026, net of trading friction:

	CAGR	Max Drawdown	MAR
Generalized Protective Momentum	+9.9%	-11.7%	0.84
GPMv	+12.7%	-11.8%	1.08
Protective Asset Allocation	+9.8%	-13.9%	0.70
S&P 500	+12.0%	-51.0%	0.24

Three things stand out.

Against PAA, the model it was built from, GPM earns a similar return with a shallower worst decline. The correlation adjustment did what it was supposed to do: the portfolio is less prone to holding three flavors of the same risk, so its bad months are less bad. That is a modest improvement rather than a dramatic one, which is honest for a change to the scoring step alone.

Against the index, GPM gives up return and cuts the worst decline to a small fraction. This is a defensive strategy and the table says so plainly.

Against GPMv, GPM trails by a wide margin on return at almost exactly the same drawdown. The variant's changes are aimed at the scoring inputs and the defensive side rather than at the core idea, which is intact in both.

The same four, measured from January 2000:

January 2000 through August 2026, net of trading friction:

	CAGR	Max Drawdown	MAR
Generalized Protective Momentum	+7.2%	-11.7%	0.62
GPMv	+10.4%	-11.8%	0.89
Protective Asset Allocation	+7.9%	-11.0%	0.72
S&P 500	+8.0%	-51.0%	0.16

The recent window is less kind. GPM roughly matches the index's return over a period that included two declines of about half, which is a real achievement in drawdown terms, but its edge on raw return is gone and PAA slightly outperforms it here. Note also that GPM's worst decline is identical in both tables, meaning its deepest drawdown across 46 years happened after 2000.

Two cautions apply as always. A window starting in January 2000 begins shortly before a major bear market, which flatters strategies that step aside during one. And a single figure covering decades says nothing about the order the returns arrived in, which is most of what an investor lives through.

Portfolio Characteristics

• Correlation-aware selection. The defining feature. Three holdings chosen to be genuinely different from each other rather than three names that behave alike.
• Breadth-driven defense. How defensive the portfolio gets depends on the health of all twelve assets, not on the three being held.
• Gradual, not binary. The defensive share moves in sixths, so the portfolio can be anywhere from fully invested to fully in Treasuries.
• Two defensive speeds. Short or intermediate Treasuries, whichever is scoring better, which keeps the defensive position from becoming a losing trade when rates rise.
• Concentrated risk side. Only three risk holdings at a time out of twelve candidates.
• Broad opportunity set. Equities, credit, real assets and duration are all eligible.
• Monthly rebalancing. More trading than a hold-until-the-signal-changes design.
• No leverage. GPM never holds a leveraged fund.
• Best suited to tax-deferred accounts. Monthly rebalancing generates short-term gains. In a taxable account, plan accordingly.
• Fully mechanical. Every allocation follows from published rules with no discretionary override.

Who It's For

GPM is designed for investors who want:

• Momentum selection that accounts for how much the picks overlap.
• Risk that comes off gradually as market breadth deteriorates, rather than all at once.
• A defensive position that can shift duration instead of sitting in whatever bonds are available.
• Shallow drawdowns as the primary objective, with return as the secondary one.
• A rules-based process with no market-timing judgment calls.

It is a poor fit for investors seeking maximum growth, those uncomfortable holding only three positions on the risk side, or those trading in a taxable account. Investors who like this structure should compare it with GPMv, which follows the same design with modifications, and with Protective Asset Allocation, the model GPM was built from. Those wanting the authors' simpler and more recent thinking should look at Hybrid Asset Allocation.

---

Generalized Protective Momentum was created by Wouter Keller and Jan Willem Keuning and is presented here as an independent implementation. Dual Momentum Systems is not affiliated with or endorsed by either author. The original description is on Keuning's blog.

Past performance, including backtested results, is not indicative of future results. Backtested performance is hypothetical, does not reflect actual trading, and benefits from hindsight in the selection of strategy rules. Portions of the long-term history rely on reconstructed proxy data for funds that did not exist for the full period. Investors should carefully consider their risk tolerance and consult with a financial advisor.

For the latest details, visit www.DualMomentumSystems.com