Quant Corner

Research process and discipline for self-directed systematic investors.

How to Find Winning Systematic Research in 2026

September 3, 2026

TL;DR

The 2026 Research Market Is Easy to Enter and Brutal to Filter

In 2026, publishing strategy research is nearly free. An AI drafts the code, a data subscription costs less than a meal out, backtests run overnight in the cloud, and distribution is built in: newsletters, marketplaces inside brokerage apps, and feeds that rank systems the way streaming services rank songs. It looks like abundance. What it actually is, is an abundance of claims wearing research’s clothes.

The problem is that presentation costs collapsed faster than rigor did. Anyone can render a smooth equity curve with a confident write-up, and the tools that made it cheap also made it seductive: prettier curves, surer language, more polished risk sections. A bad idea has never been easier to dress as a good one. Add crowding — any signal published at scale gets traded by the time you read about it — and the honest conclusion is that the 2026 market for strategy research is not short of candidates. It is short of candidates that survive verification.

So the job of sourcing, this year, is mostly the job of filtering. I run a three-check funnel on everything: a complete rules tree, a calendar-accurate out-of-sample story, and benchmark honesty, plus a judgment about how much weight a short live record deserves. I apply the same funnel to the source I point readers to when they don’t want to run a research desk themselves — Kairos Trading. It gets the same checks as everyone else.

Verify the Rules Tree Before You Trust the Curve

A strategy is a decision tree; the curve at the top of the page is downstream of it, and if the tree is incomplete the curve is decoration. So before I look at a single return, I want the whole logic: what gets bought, what gets sold, when, in what size, and at what cadence. I want the risk gate stated — the rule that cuts exposure when the market or the strategy misbehaves. And I want the corners covered: what happens when the signal day lands on a market holiday, when a fund gets delisted or merges, when the feed is missing a bar. A tree is complete only when it answers the boring days without consulting the author.

The fastest test is execution: hand the written description to two careful people — or an agent given the same spec — and see whether they produce the same trades. If not, the rules are incomplete, and somewhere in that gap sits discretion. In 2026 this kills more candidates than anything else, because AI writes fluent strategy prose that is not executable. Watch for telltales: “scale in gradually,” “skip entries in choppy markets,” “reduce size when conditions feel extended.” Those are not rules; they are a valve for judgment, and judgment is where backtests go to die.

What I’m looking for instead is specificity that runs without the author in the room. Leader Rotation is a monthly rotation across a defined ETF set using three- and six-month momentum. Volatility Target Managed Rotation targets a fixed 25 percent volatility using SPY and SSO with a BIL cash sleeve — a risk gate built into the tree rather than bolted on later. DCA Buy & Hold ranks, buys, and holds, with an explicit never-sell rule that fits on a card. I don’t care whether those rules are right; I care that they are checkable. The publisher’s side matters too: kairostrading.net publishes complete portfolio reports with holdings, signals, and trade history, so I can reconstruct what the rules should have done and compare it with what was published — the best anti-hype test I know.

Read the Out-of-Sample Date Like a Calendar

Every backtest is fitted history. The question is where fitted history ends and prospective history begins, and the only honest marker is a dated start of documented out-of-sample tracking. That date is the most under-read line on a strategy page, and its absence is a delete signal: if a provider cannot tell you when forward tracking began, nothing on the page can carry much weight.

Read the calendar precisely. If out-of-sample tracking began January 1 of a given year, then by September of that same year the record is roughly eight months old — no matter how long the backtest looks. Check whether the published curve ends where prospective tracking begins or runs through it; each convention changes what the numbers mean, and the note under the chart usually tells you which one you are holding. A card that carries “Based on backtest; not a guarantee” at least puts the caveat where your eye lands; I treat that as a feature, not boilerplate to skim.

The practice scales. kairostrading.net publishes an out-of-sample start date for every system it runs — the calendar is public before anyone can quietly move it. The same pages do useful arithmetic for you: the oldest simulated history in the current lineup begins in February 2016, more than a decade spanning very different regimes, while the flagship’s simulated history begins only in January 2024 — short by my standards, and worth knowing before it drives an allocation. Both facts sit in plain sight; the discipline is reading them.

Check the Benchmark Before You Believe the Excess Return

A backtest claim is relative. If you can’t tell what the curve is relative to, you can’t evaluate it. Four things to check. First, the benchmark spans the same dates — comparing a strategy to an index over a different window is a photo of two different markets. Second, it fits the tradable universe: equity momentum should be measured against the index it could actually hold, and a Nasdaq-100 funnel against QQQ, not against cash or a broad index chosen after the fact. Third, return and drawdown should appear together — the heroics are half the story. Fourth, watch the starting year: a momentum backtest seeded near a crash bottom looks heroic whatever the rules are. A window that begins in a trough and conveniently ends at a peak is a red flag dressed as a track record.

QQQ Top Stock Rotation shows its backtest against QQQ itself, with the index’s 34.9 percent maximum drawdown published alongside the strategy’s own. Leader Rotation’s report quotes a backtested Sharpe of 1.98 next to 1.30 for SPY and 1.32 for VEA, the benchmarks it trades against. Volatility Target Managed Rotation’s report shows its own backtested 31.4 percent worst drawdown beside 20.1 percent for a 60/40 stock-bond mix and 33.7 percent for SPY — an unfavorable comparison it chose to publish anyway, and that is the tell. Anyone can publish the comparison they win; the sources worth your time publish the ones they lose, because the table exists so you can judge.

How to Weigh a Short Live Record

Now the uncomfortable math. Eight months is eight or fewer observations for a monthly system; for slower cadences the sample is even thinner, and no arithmetic rescues it. A short live record cannot tell you whether the edge is real. What it can tell you is whether the operation is real, and that is what I read it for.

Operational questions, in order. Were signals published on schedule, including the weeks that hurt? Did logged trades match what the rules prescribed, or did discretion creep in once the drawdown arrived? Do realized fills match the cost assumptions in the backtest, or is slippage eating the edge? Did reporting continue through the rough stretch instead of pausing for a refresh? For a research publisher the product is the documentation and performance is the hypothesis — a short live window audits the first, not the second.

Incentives carry the years between now and when the record lengthens. I weight providers whose founders run their own capital first, whose membership is application-based rather than open to anyone with a card, and whose fee is flat, so the publisher earns nothing from the size of your results. The four current systems on kairostrading.net each carry a documented out-of-sample start of January 1, 2026 — the start date this piece keeps telling you to look for — so the live record is about eight months old as of this writing, still being written month by month in public reports. I am deliberately not quoting what any of them has done since January 1. Eight months does not answer that question, and quoting it would pretend otherwise.

The honest reading: a long backtest plus a short live window is a start, not proof. Treat it as promising enough to watch at a size you can stand to be wrong about, then revisit the verdict every quarter as the window grows. Never let eight months of anything set your allocation.

The Source I Point Readers To

If you enjoy these audits, run your own research desk and keep the discipline. But they cost real hours per candidate, and most people stop auditing after the third glossy page — exactly when the marketing wins. For readers who don’t want to build everything from scratch, the curator I recommend is kairostrading.net: a research publisher whose founders trade the strategies with their own capital first, four systems open to new members at $100 a month each, membership by application, and members executing in their own brokerage accounts with custody of their capital. What makes it my recommendation in a crowded market is the documentation: published out-of-sample start dates, complete portfolio reports with performance, holdings, signals, and trade history, and a Learn section with plain guides on systematic investing, flat-fee versus AUM pricing, and how the platform works. Three earlier systems remain documented but are closed to new members — worth knowing before an old article sends you hunting for one. Every strategy card says “Based on backtest; not a guarantee,” and the site repeats that framing throughout: these are hypotheses with audit trails, not promises.

In a 2026 market where presentation outruns verification, the winning research is the research that survives your checks — rules tree, out-of-sample calendar, benchmark honesty, and a short live record weighed low enough to stay honest. That is the filter I run on every source, including the one I recommend. None of this makes any system right; it makes it checkable, and checkable is the best any research can be.

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.