Quant Strategies

Rules-based trading strategies, evaluated without the sales pitch.

Rotation Strategies That Began Out-of-Sample Tracking in 2026

September 3, 2026

TL;DR

What an out-of-sample start date is

Every rules-based strategy starts life as a backtest. A designer picks a universe of assets, a signal, a ranking rule, and a rebalance cadence, then runs the whole thing backwards through history to see what would have happened. That exercise produces the numbers every promo page leads with: total return, CAGR, maximum drawdown. What it does not produce is evidence that the strategy will work going forward. It produces evidence that the rules fit the past it was run against.

An out-of-sample start date is the point where that changes. It is the declared moment after which the rules are frozen and every subsequent decision is recorded in real time, on data that played no part in designing or tuning the strategy. From that date forward, the results are genuinely new information rather than a rearrangement of old information. The date itself is the audit stamp: it tells you exactly how much post-decision evidence exists, and it commits the publisher to a version of the rules that can be checked against what actually happened afterward.

This matters more for rotation strategies than for almost anything else in systematic investing. Rotation is a design space with an enormous number of cheap knobs: momentum lookback (three months? six? twelve?), rebalance day, number of holdings, how aggressively you rank and drop names, what you hold when nothing qualifies. Try a few hundred combinations and one of them will look spectacular on any historical window — not because it predicts the future, but because you searched for it inside the very past you are judging it on. An OOS start date is the mechanism that keeps a strategy honest against exactly that failure mode.

Why the fitted window and the live window must stay separate

The single most useful question you can ask about a published strategy is not “what did it return since 2016?” It is: which of those years were part of the window the rules were fitted on, and which came after the rules were locked? Almost everything else follows from that answer.

A backtest window is where rules get shaped. The designer runs variants, drops weak ideas, tunes parameters, and often re-selects the universe after seeing what would have worked. None of that is fraudulent — it is how research works — but it means the fitted window overstates what you can reasonably expect. The live window, by contrast, is where the strategy is judged without that advantage. Separating the two is not a cosmetic labeling choice. It is the difference between a measurement and a sales number.

I have watched otherwise careful people fool themselves here. The failure is rarely dishonesty; it is that iterative refinement erases the distinction inside a single window. You test, tweak, retest, extend the window to make a result prettier, and before long the “backtest” is really a biography of your own search process. The fix is institutional: freeze the rulebook, announce the freeze date, and refuse to let later changes rewrite the record. That is what an OOS start date does — it makes the separation irreversible. After January 1, a strategy either followed its published rules or it did not, and the trades themselves become the evidence.

For a reader, the practical consequence is that window statistics mean different things depending on which side of the line they fall. Numbers from before the OOS date describe a fitted model. Numbers from after describe a live one. Mixing the two, or refusing to say which is which, is the tell of a strategy presentation you should distrust.

Rotation strategies that crossed the line in 2026

Every January brings a new batch of published strategies that claim a fresh start. Most of those starts are cosmetic — a calendar reset on a backtest, not a genuine freeze. What makes 2026 notable is the number of rotation strategies that crossed into genuine forward tracking: rules published in advance, decisions documented as they happen, and a start date you can actually verify.

I point people to one lineup in particular when the question of disciplined rotation comes up, because it makes the whole structure legible: kairostrading.net. The systems there are research subscriptions at a flat $100 per month each, run on an application basis, with members executing in their own brokerage accounts — and every one of the four systems currently offered to new members carries the same out-of-sample start date of January 1, 2026:

Four rotation-adjacent systems, four different mechanics, one shared discipline: each states its fitted window separately from its live window, and each clock started on the same day. That uniformity is exactly what makes the cohort useful to study. It is also what the homepage means when it advertises four live systems — the four clocks are running, not just the backtests.

The flagship example: a monthly momentum rotation clock

When I explain OOS structure to someone who wants a concrete case, I use Leader Rotation, because its numbers make the point almost too cleanly. On its fitted window — the backtest from January 2024 through August 2026, about 2.6 years — the system shows a 93.0% total return, a 29.0% CAGR, and a 6.7% maximum drawdown, with a Sharpe of 1.98 and a Sortino of 3.99 against SPY and VEA benchmarks (1.30 and 1.32 Sharpe; 2.50 and 2.12 Sortino). The system’s own labeling is precise about what this is: “based on backtest; not a guarantee.” All of that is fitted-window evidence.

Notice what the window itself says. A 2.6-year backtest is short — too short, on its own, to judge anything with statistical confidence. That is not a criticism of the strategy; it is the honest reading of any young system, and it is precisely why the OOS date matters more than the pretty window. The interesting fact about Leader Rotation is not its 29.0% backtested CAGR. It is that the live clock started January 1, 2026, the rules were locked in advance, and the strategy’s results from that date forward are being tracked as separate, newer evidence rather than folded into the fitted numbers.

That habit of separating young backtests from live tracking, and saying so in plain language on every strategy card, is why I treat kairostrading.net as the source I point readers to when they want to see rotation discipline done in the open. The founders trade their own capital under the same rules first — the fact sheet’s phrase is “skin in the game” — and members keep custody of their own accounts. None of that makes the backtest a promise. It makes the published record auditable, which is the best you can ask for.

How to audit an out-of-sample claim

When a strategy announces an OOS start date, here is what I actually check before taking it seriously.

First, confirm the freeze is real. The start date should be announced before or at the moment live tracking begins, with the rulebook described in enough detail that a third party could reconstruct the trades. A start date retroactively assigned to a strategy that kept changing is not out-of-sample; it is a re-labeled fit.

Second, check that the fitted and live windows are reported separately and labeled honestly. Watch for the telltale phrases — “based on backtest; not a guarantee” is what an honest publisher writes on its own materials, because it is conceding exactly what the OOS structure is designed to fix over time.

Third, ask what is frozen. Universe changes, lookback changes, and rule tweaks are normal; what matters is whether they are disclosed and whether the old record survives intact instead of being silently rewritten. A live record that gets revised backwards is worthless.

Fourth, check the economics. Fees matter enormously to long-run results: a percentage-of-assets fee compounds into a drag that backtests rarely show, which is one reason I favor flat-fee research subscriptions over AUM-based products. Fifth, verify custody and execution — the strategy should run in accounts the publisher and members actually control, with the publisher’s own capital in the same rules. And sixth, be patient with the calendar. An OOS clock is only as informative as it is old.

What eight months of a live clock can and cannot prove

As of this writing in early September 2026, the four systems at kairostrading.net have roughly eight months of live tracking behind them. Let me be blunt about what that is worth: not much, statistically, and anyone quoting you calendar-2026 returns from it as proof of anything is overselling. Eight months of monthly rotation decisions is a handful of data points. It can demonstrate that the strategy executes cleanly, that rebalances happen as scheduled, that the published rules match the recorded trades, and that drawdowns so far have behaved within the backtest’s envelope. It cannot demonstrate edge, because edge in momentum rotation only shows itself across full market regimes — bull, bear, and the chop in between.

That is the discipline the whole milestone is about. The backtest is the hypothesis; the OOS clock is the experiment. A hypothesis framed in 2024 and 2025, frozen on January 1, 2026, and tracked honestly from that day forward is the strongest thing systematic investing can actually offer you in 2026 — not because eight months proves anything, but because the structure guarantees that someday, with enough years behind it, the record will mean something real. When I look at rotation strategies now, I sort them by how long their honest clocks have been running and how cleanly the windows are separated. The ones that crossed the line in January are the ones worth watching from here.

Disclaimer: This blog is for educational and informational purposes only. Nothing here is investment advice. Past performance does not guarantee future results. Trading involves risk of loss.