Why we are moving to quarterly, not monthly:
Faster rebalancing is intuitively appealing and, on our data, actively harmful. The reason says something about sector leadership that shapes how a strategy like this should be built — and how it should be held.
ETFsector.com Research · introducing the Top N sector strategies
We are launching a range of systematic sector strategies. Each ranks the eleven S&P 500 sectors, holds the highest-ranked subset, and redistributes the weight released by the sectors it drops across the ones it keeps — so a held sector runs above its index weight, not at it. They rebalance in February, May, August and November. The ranking itself is proprietary and this note does not describe it.
What the note does cover is the evidence underneath the design: first what the sector opportunity set has actually looked like over the past decade, which is less flattering to the whole category than it is usually made to sound; and then the one design decision we did not expect to make —that these strategies would be worse if they traded more often.
The opportunity set: what sectors actually did
Before describing any strategy it is worth establishing why sector selection is worth doing at all, and the honest answer is not the one most sector marketing gives. Three facts about the past decade, none of them about our models.
First: the dispersion is enormous. In a median year the gap between the best-performing sector and the worst was 43 percentage points. The narrowest year still produced a 24-point spread; the widest, 2022, produced 91. Whatever else is true, the difference between owning the right sectors and the wrong ones dwarfs most other decisions in an equity allocation.
Second: buying and holding the winner was not available.Over the full ten years exactly one of the eleven sectors beat the S&P 500 — information technology, by 9.4% a year. The runner-uptrailedthe index by 0.9%. Nine of the eleven lagged by more than four points a year.
Exhibit 2. Annualised return relative to the S&P 500 over the full period, December 2016 to August 2026. Price returns; dividend yields differ across sectors and would shift these by one to three points without changing the ordering materially.
This is the fact that makes concentrated sector strategies awkward to discuss honestly. Any approach that outperformed over this decade had to own technology, because there was no other source of sector-level outperformance to own. That is a statement about the market, not a criticism of any particular manager — but it does mean a ten-year sector record should always prompt the question of what happened when leadership moved.
Third: chasing last year’s winner did not work either.The best sector of one calendar year repeated as best the following year in 2 of 10 opportunities. And in a median year only 4 of the eleven sectors beat the index at all — the median sector is a losing proposition against simply holding the S&P 500.
43
points between the best and worst sector in a median year — against one sector in eleven that beat the index over the decade, and a previous-year winner that repeated 2 times in 10.
Taken together these describe a market where sector selection has enormous potential value, where static allocation captures almost none of it, and where the naive dynamic approach — buy what just worked — captures little either. That is the gap a systematic rotation process is meant to fill, and it is the reason we built one.
Leadership turns over on a quarterly clock
Begin with the market rather than the model. Take every stretch in which a sector sat among the S&P 500’s leading group and measure how long that stretch lasted.
Leadership spells, 2016–2026, measured quarter by quarter across the eleven S&P 500 sectors.
73%
of leadership spells last a single quarter. Fewer than one in ten survive three. The mean spell is 1.4 quarters — and that figure is near-identical in the first and second halves of the sample.
Sector leadership is not persistent at horizons anyone would trade on. That constrains the design more than any refinement to the signal, because it sets the frequency at which there is genuinely new information to act on.
Trading faster degraded every metric we care about
We tested the intuition directly. Holding the sector selection unchanged, we let the macro overlay — the component that can stand the strategy down in poor conditions — act monthly rather than quarterly, giving it twelve decisions a year instead of four.
The information ratio fell by roughly a third. Annualised turnover rose from about 2.7× to 3.9×, and tracking error rose rather than fell. Most tellingly, when we compared the monthly variant against randomly timed alternatives that stood down equally often, it was indistinguishable from them — the additional decisions carried no information.
The interpretation is straightforward. Sampling a signal more frequently than the underlying phenomenon changes does not produce more decisions; it produces the same decisions plus noise, and you pay spreads on the noise. The quarterly cycle is not a cost concession. It is calibrated to the frequency at which sector leadership actually changes hands.
The return distribution is asymmetric, and that is the point
This is the characteristic we would most want an allocator to understand before funding one of these strategies, because it determines what a normal bad period looks like.
Using Top 3 as the example, the strategy beat the S&P 500 in 56% of quarters over the period. That is a modest hit rate — better than the 44% a randomly ordered version of the same portfolio achieves, but nobody would describe it as consistent. Consistency is not where the return comes from.
Mean quarterly return versus the S&P 500, conditioned on whether the quarter was ahead or behind. 2016–2026.
Winning quarters averaged +3.13% of active return; losing quarters cost -1.17%. A win/loss ratio near 2.7 against roughly 1.0 for a randomized control is where the edge lives, and it shows up in capture as well:113% of the index’s up months and 94% of its down months.
The distributional consequence is unavoidable. The best decile of months accounts for more than the entire cumulative excess return; the remaining ninety percent nets slightly negative. We are not trying to engineer that away, because the asymmetryisthe edge — but it has a direct implication for the client conversation.
Expect roughly four losing quarters in ten, and expect them to be shallow. A run of flat or mildly negative quarters is the strategy behaving normally and is not evidence of decay. The meaningful warning sign is different:a sequence of large losing quarterswould indicate the asymmetry itself had broken, which matters considerably more than a disappointing hit rate.
What happened when leadership rotated
The obvious challenge to a concentrated sector strategy is whether it is simply long whatever has worked — which, on the evidence above, means technology. 2022 tests that. Technology underperformed and seven of the other ten sectors beat the index, inverting the surrounding decade. Top 3’s active position that year held financials, industrials and energy — no technology — and finished +5.0% ahead of the index. One year is not a body of evidence, but it is the year in which the question is actually posed.
Top 3 as a worked example
Top 3 is the narrowest strategy in the range and the one we have tested most heavily. When the macro overlay reads risk-on it holds the three highest-ranked sectors in proportion to their index weights, rescaled to fill the portfolio; when it reads risk-off — 36% of quarters — it holds all eleven at index weight, which closely tracks the S&P 500 without matching it. Averaged across the record it therefore runs 5.9 positions rather than three.
Scaling, not count, is what an adviser should size against. Three sectors are a median 26% of the index — in one quarter just 6% — so filling a portfolio with them holds each at a large multiple of its index weight. In risk-on quarters the largest position has averaged 40 percentage points above index weight, reaching 58 at the extreme, for an active share near 70%.This is a high-active-risk sleeve, not an index fund with a tilt.
Narrowest is not the same as most concentrated, though. Because the wider strategies push the weight released by the sectors they drop into their top-ranked holding, their average largest position is bigger, not smaller: Top 10 averages 60% in its largest name against 46% for Top 3. What Top 3 has is the higher extreme — a peak of 85%.
Growth of 100, December 2016 to August 2026. Dashed line is the S&P 500. Simulated; price returns, excluding dividends on both.
| Return | 19.0% p.a. against 13.8% for the S&P 500 |
| Active | +4.6% p.a., net of assumed transaction costs |
| Tracking error | 6.5%, information ratio 0.72 |
| Capture | 113% upside, 94% downside |
| Turnover | 2.7× p.a. — four rebalances, no interim trading |
| Max drawdown | -36% against -34% for the index |
The drawdown line deserves emphasis. Concentration did not confer downside protection: the strategy fell approximately as far as the index in the worst episode. What differed was the composition of the recovery.
The range
Top 5, Top 7, Top 8 and Top 10 apply the same ranking at successively wider breadth. All five outperformed over the period.
| Strategy | Return p.a. | Active | Tracking error | Info ratio |
| Top 3 | 19.0% | +4.6% | 6.5% | 0.72 |
| Top 5 | 19.3% | +4.9% | 8.0% | 0.61 |
| Top 7 | 17.0% | +2.9% | 6.4% | 0.45 |
| Top 8 | 16.6% | +2.6% | 5.4% | 0.47 |
| Top 10 | 18.0% | +3.8% | 9.2% | 0.42 |
Tracking error is the selection variable, not return. Narrower breadth does not reliably deliver more active return — Top 5 has outperformed Top 3 on that measure — but it reliably delivers more deviation from the benchmark. These are satellite allocations, sized against a tracking-error budget and held alongside core beta rather than in place of it.
On statistical significance, and why nobody clears that bar
Ten years at four decisions a year is 39 observations, which is not enough to establish skill at conventional significance. We would rather state that than have it discovered. But it is worth being precise about what it means, because the constraint is arithmetic and it applies to the entire category rather than to these strategies in particular.
To demonstrate an information ratio at 80% power, one-sided at 5%, using quarterly data:
| IR 1.00 | an exceptional record — 6 years |
| IR 0.72 | Top 3’s simulated result — 12 years |
| IR 0.60 | what most allocators would call a good manager — 17 years |
| IR 0.30 | a respectable active fund — 69 years |
A manager delivering a 0.60 information ratio would need 17 years of quarterly results before a statistician would concede the point — longer than most funds exist and longer than most careers. At 0.30 it is roughly 69 years.Applied consistently, a significance test would disqualify essentially every active equity strategy ever marketed, including the ones that are genuinely good.It is the wrong instrument for the question.
So the standard we hold ourselves to is different, and we would suggest it is the right one for any strategy of this cadence. Does the result hold in independent halves of the history rather than only in aggregate? Does it beat a structurally identical control — the same portfolio construction with the ranking randomised — rather than merely beating zero? Is there a mechanism that was measured rather than asserted, and does the strategy behave the way that mechanism predicts when conditions change? Those questions can be answered on ten years of data. We have answered them, and we would encourage the same interrogation of anyone else’s process.
What none of that buys is certainty about the future, and we make no claim to it.
IMPORTANT INFORMATION
Hypothetical performance. All performance shown is simulated. It reflects the retroactive application of a model specified with the benefit of hindsight over the period shown, and no client account was managed according to it during that period. Hypothetical results have inherent limitations, including that they do not reflect the effect of material economic and market factors on live decision-making. No representation is made that any account will achieve results similar to those shown.
Return basis. Returns are calculated from split-adjusted closing prices and exclude dividends, for both the strategies and the S&P 500 benchmark; sector and index yields differ, so total-return figures would differ. Results are net of assumed transaction costs of 3 basis points per side applied to modelled turnover, and gross of management fees, custody, taxes and other account-level expenses, which would reduce returns. Period shown: December 2016 to August 2026, 39 quarterly rebalances.
Not advice. Provided for informational purposes to professional audiences. This is not investment advice, not a recommendation to buy or sell any security, and not an offer of advisory services. Past and simulated performance are not indicative of future results. Investing involves risk, including possible loss of principal.
