TL;DR: the answer box
- "Tactical" is two different bets. A trend filter and a momentum rotation used the same five assets and behaved like opposites.
- The trend filter bought safety, not return. From 2017 to 2026 it grew 6.31% a year with a 7.35% worst drop. The fixed mix grew 8.30% with a 15.98% worst drop.
- On worst drop, it beat random timing and a same-risk fixed mix. Its worst drop was smaller than 99.6% of 500 random switchers. On return a year it beat only 36.8% of them. A fixed mix holding the same average amount of risk still fell 11.60%.
- The momentum rotation did not stand out from random picking. On return per unit of drop it beat 55.8% of 500 random rotations from 2017 to 2026. That is the middle of the pack.
- Low costs barely mattered. High costs flipped a ranking. At $0.015 a share each side the rotation gave up 0.06 points a year. At $0.25 it gave up 0.96 and fell below the fixed mix.
- Safety has a waiting cost. From 2008 to 2026 the trend filter beat the fixed mix in only 4 of 17 full years.
What is the difference between strategic and tactical asset allocation?
Strategic asset allocation picks a mix and holds it. Say 20% in each of five assets. Once a month you trade back to those weights. The mix never reacts to the market.
Tactical asset allocation lets a rule move the weights. Each month the rule looks at recent prices and decides what to hold next.
You have heard the sales pitch. Tactical gives you higher returns with lower risk, because it steps into safer assets before a crash does its damage.
That pitch makes two claims. Higher return. Lower risk. I tested them one at a time, because a rule can win one and lose the other.
Think of the fixed mix as a ship on autopilot. A tactical rule is a navigator who changes course. The navigator earns a place on the ship only if it steers around more storms than a coin flip would, without burning more fuel on turns than the storms would cost.
Three terms carry the numbers below:
- Worst drop (also called drawdown): the biggest fall from a high point to a later low.
- Return a year: the average yearly growth, after costs.
- Return per unit of drop: return a year divided by the worst drop. Higher means you earned more for each point of pain. Traders call it MAR.
How I tested tactical against strategic
The five assets. SPY (US stocks), EFA (international stocks), IEF (7 to 10 year Treasuries), GLD (gold) and BIL (T-bills, the cash). Prices include dividends.
The two tactical rules. Both were fixed before the test, at the settings their authors published.
- Trend filter. Each asset gets a 20% slot. Stocks, bonds and gold keep their slot only while the month-end price is above its 10-month average. If not, that slot moves to T-bills. This is the model in Mebane Faber's A Quantitative Approach to Tactical Asset Allocation.
- Momentum rotation. Rank the four risky assets by how much they rose over 12 months. Hold the top 2, half each. If a pick did not beat T-bills, hold T-bills instead. Gary Antonacci calls this structure dual momentum. Moskowitz, Ooi and Pedersen show the trend side of momentum across markets in Time Series Momentum.
The fixed mixes. 20% in each of the five, and 60% US stocks with 40% Treasuries (60/40). Both trade back to their weights every month. US stocks bought and held is the do-nothing baseline.
Two tests. The main test runs from 2017 to 2026 on daily bars with real opening prices. Each rule decides at the last close of the month and trades at the next open. The long test runs from 2008 to 2026, so it includes the 2008 crash. Its data has closing prices only, so trades fill at the next day's close. Every number says which test it comes from.
Costs. $0.015 a share on every buy and sell, swept up to $0.25. Every portfolio starts with $100,000.
Here is one rule at work on real prices.
On April 29, 2022, SPY closed at 388.80. Its 10-month average was 418.76. The close was below the average, so the rule sold SPY's slot at the next open, 388.87, and parked it in T-bills.
It stayed out until November. On November 30, SPY closed at 387.98, above its average of 383.35. The rule bought back at the next open, 389.02.
So it sold and bought at almost the same price, and sat out the months in between. In 2022 the SPY slot was out at 9 of 12 month-end checks. It was not perfect, though. It also whipsawed: out in February, back in March, out again in April. A whipsaw is a quick out-and-back-in that costs you a little each time.
Does tactical asset allocation beat a fixed strategic mix?
On return, the trend filter did not and the rotation did by a little. On safety, only the trend filter did, by a wide margin.
The trend filter cut the worst drop by more than half. From 2008 to 2026 its worst drop was 64.8% smaller than the fixed mix's. From 2017 to 2026, 54.0% smaller. The price was 1.81 points of return a year on the long test and 1.99 on the main one.
The momentum rotation earned a bit more and fell at least as far. 0.89 points a year more than the fixed mix from 2008 to 2026. Its worst drop was 23.53%, against 23.28%.
Nobody beat US stocks on return. They grew 11.69% a year from 2008 to 2026. $100,000 became $743,835. You paid for it with a 50.65% fall along the way.
Is the trend filter really timing, or does it just own less stock?
Owning less explains part of it. Not all.
The trend filter spends a lot of time in cash. From 2017 to 2026 it held on average 57.8% in the four risky assets and 42.2% in T-bills. Less risk on average would give a smaller drop with no timing skill at all. Most tactical comparisons skip this test.
So I built a fixed mix with the same average risk: 14.45% in each risky asset and 42.18% in T-bills, traded back every month.
It still fell further. Its worst drop was 11.60% from 2017 to 2026. The long test gets its own same-risk mix, set from that test's average: 14.00% in each risky asset and 44.01% in T-bills. Its worst drop was 16.53% from 2008 to 2026. The trend filter fell 7.35% and 8.19%. So holding less risk on average does not explain the whole gap. The fixed mix splits its money evenly across the four risky assets, though. The trend rule held more US stocks (16.3%) and gold (15.2%) than Treasuries (11.9%). So this test does not split timing from which assets it favoured.
Then the luck test. I built 500 random switchers for the main test and 250 for the long one. Each month, each one moves its slots in or out with the same odds of switching that the trend rule had. Only the timing is random.
The trend filter's worst drop was smaller than 99.6% of them, in both tests. It also held at every average length I tried: 6, 8, 10 and 12 months all cut the worst drop below the fixed mix's.
Does momentum rotation beat random picking?
Not clearly. The random rotations each held two of the five assets, chosen at random. At each month end, a random rotation redrew its pair with the same chance the real rule had of changing its holdings in that test. A redraw can land on the same pair again.
On return per unit of drop, the rotation beat 55.8% of them from 2017 to 2026. That is close to a coin flip. On the long test it did better, 82.0%, but still short of the top 5% of random runs.
The verdict also flips with the setting. I ran the rotation at five look-back lengths and three portfolio sizes, 15 settings in all. From 2017 to 2026, 5 of 15 beat the fixed mix on return per unit of drop. From 2008 to 2026, 8 of 15. The best setting in one test was not the best in the other.
What did tactical allocation do in 2008, 2020 and 2022?
In all four sell-offs I measured, the trend filter lost less than the fixed mix and US stocks. The rotation was great in 2008 and worse than the fixed mix in two of the other three.
2008 is the crash the sales pitch describes. From May 2008 to March 2009, US stocks fell 50.00%. The trend filter made 1.83%. The rotation made 3.38%.
Late 2018 went the other way. The rotation lost 12.93% in three months. The fixed mix lost 5.09%.
In the COVID crash the trend filter lost 3.80%, against 13.47% for the fixed mix and 33.70% for US stocks.
2022 was the hard test. Stocks and bonds fell together. The 60/40 mix lost 16.61%, close to the 18.17% for stocks alone. Treasuries (IEF) on their own lost 15.15% that year, so the 40% in bonds barely cushioned the fall. The trend filter lost 5.98%. The rotation lost 11.01%, more than the fixed mix's 9.17%.
How much do trading costs take out of a tactical portfolio?
Very little at the low rate. At the high rate, enough to flip a ranking.
The two tactical portfolios here trade a lot more. The rotation traded 195% of the portfolio a year. The trend filter traded 151%. The fixed mix traded 13%. So costs hit tactical harder.
At $0.015 a share, harder is still small. At $0.25 a share, the rotation gave up 0.96 points a year and slipped below the fixed mix. The fixed mix gave up 0.10.
How long can a tactical portfolio trail a fixed mix?
For years.
The trend filter beat the fixed mix in only 4 of 17 full years from 2008 to 2026. It beat it in all four sell-offs above. It trailed the fixed mix in every other full year.
Its lead peaked in March 2009. In the 208.6 months after that, it never got back to that high point. It finished with 0.73 of the fixed mix's growth.
I entered the markets in 1998, at the dot-com peak. That is why I read the worst-drop column before the return column. But be honest with yourself. Would you hold a rule that beat the fixed mix in only 4 of 17 years?
The verdict
Where the tactical pitch is right. A simple trend rule cut the worst drop by more than half, in both tests, at every average length tried. It fell less than a fixed mix with the same average risk. It beat 99.6% of random switchers on worst drop. In the 2008 crash it made money.
Where it is wrong.
- Higher return and lower risk did not come together. The trend filter earned 1.81 points a year less than the fixed mix from 2008 to 2026.
- Lower risk belongs to the rule, not to "tactical". The rotation's worst drop was 21.47% from 2017 to 2026, deeper than the fixed mix's 15.98%.
- Momentum picking did not stand out from random picks. On return per unit of drop, the rotation beat 55.8% of random rotations on the main test and 82.0% on the long one. It never reached the top 5%.
- "Safer assets in a bear market" depends on the bear. In 2022, Treasuries (IEF) lost 15.15% while US stocks lost 18.17%.
- Neither rule beat buy and hold on return. US stocks grew 11.69% a year from 2008 to 2026. The better tactical rule grew 7.28%.
The limits, in plain words.
- Five ETFs only. Other asset lists, sectors or single stocks can behave differently.
- The main test is 8.99 years long. The long test adds 2008 but trades at the next day's close, not the next open.
- Two rules at their published settings, plus the 15 rotation settings and 4 trend lengths around them. Many other tactical rules exist.
- Taxes on monthly trading, fund fees and wide price gaps in a fast market are not modelled.
- The same-risk fixed mix matches the trend filter's average risk only, not which assets it held or when.
- These are backtests of the past. The random tests and the settings sweep exist to show how much could be luck or choice.
What this means for you
- Decide what you are buying. If you want a shallower drop, the trend filter delivered it: 8.19% at worst from 2008 to 2026, against 23.28% for the same assets held fixed. If you want more return, neither rule delivered it reliably.
- Price the waiting before you start. The trend filter won only 4 of 17 full years. Decide now whether you would still hold it through a run like that.
- Test your rule against random timing. The rotation looked fine against the fixed mix. From 2017 to 2026 it sat in the middle of the random runs. If your own rule has a trade history, AlgoChef runs the Monte Carlo and out-of-sample checks that show whether its record could be luck.
- Watch how much you trade. At $0.25 a share the rotation gave up 0.96 points a year. The fixed mix gave up 0.10.
Tactical allocation is not magic. One rule kept you out of deep holes and earned less. The other earned a little more, traded far more and fell at least as far.
Reference tables
For readers who want every headline number in one place.
| 2017 to 2026, next-open fills | Return a year | Worst drop | Return per unit of drop | Traded a year | 2022 |
|---|---|---|---|---|---|
| Trend filter | 6.31% | 7.35% | 0.86 | 151% | -5.98% |
| Momentum rotation | 8.84% | 21.47% | 0.41 | 195% | -11.01% |
| Equal weight, 20% each | 8.30% | 15.98% | 0.52 | 13% | -9.17% |
| Fixed mix, same average risk | 6.71% | 11.60% | 0.58 | 11% | -6.22% |
| 60/40 | 9.37% | 21.25% | 0.44 | 10% | -16.61% |
| US stocks only | 14.73% | 33.70% | 0.44 | 0% | -18.17% |
All six use the same five ETFs (60/40 holds two, US stocks one), the same month-end calendar and $0.015 a share each side.
| 2017 to 2026, return a year | $0 | $0.015 | $0.05 | $0.10 | $0.25 |
|---|---|---|---|---|---|
| Momentum rotation | 8.89% | 8.84% | 8.70% | 8.51% | 7.94% |
| Trend filter | 6.37% | 6.31% | 6.19% | 6.02% | 5.49% |
| Equal weight | 8.31% | 8.30% | 8.29% | 8.27% | 8.21% |
Cost per share on each buy and each sell. At $0.25 the rotation paid 0.214% of each dollar it traded.
Everything here is research, not investment advice. It is a backtest on historical data with the assumptions stated above. It is not a recommendation to buy or sell anything, and past results do not predict future ones.
The next portfolio test goes to the Overfit newsletter first. StatOasis.com/Overfit
