Why We Only Use Peer-Reviewed Strategies (And Why That Matters)

If you've ever looked at a signal provider's backtest and thought "that seems too good," your instincts are correct. The uncomfortable truth about quantitative trading is that finding a pattern in historical data is easy. Finding one that actually persists into the future is a fundamentally different problem — and most retail signal products don't even try to solve it.

The Problem With Most Retail Signals

Here's how the typical signal gets built: a developer downloads several years of price data, runs a strategy optimizer across hundreds of parameter combinations, and publishes whichever version produced the best Sharpe ratio. The backtest looks pristine. Drawdowns are modest. The equity curve slopes upward at a satisfying angle.

This process has a name in statistics: data mining bias, sometimes called curve-fitting or overfitting. When you test enough variations of a strategy on the same dataset, you will eventually find one that fits the historical noise almost perfectly. The problem is that noise, by definition, doesn't repeat. The strategy wasn't discovering a real market inefficiency — it was memorizing random fluctuations in a specific historical window.

Three red flags that expose this approach:

• No out-of-sample testing. The backtest period and the discovery period are the same data. There's no holdout set that the strategy was never allowed to see.
• Undisclosed methodology. You can see the signals but not how they're generated. You can't verify whether the logic makes economic sense.
• No publication or scrutiny. No other researcher has tried to replicate the finding and failed to break it.

Without those safeguards, a five-year backtest is essentially a five-year in-sample fit. It tells you almost nothing about the next five years.

What Peer Review Actually Means

Academic peer review is not a rubber stamp. When a researcher submits a paper claiming to have found a market anomaly, anonymous reviewers — typically other finance academics — attempt to find every possible flaw: data errors, look-ahead bias, transaction cost omissions, cherry-picked sample periods, alternative explanations that weren't considered. The paper only gets published if it survives that gauntlet.

More importantly, the editorial process requires full methodology disclosure. Every variable, every parameter, every data source must be described in enough detail that an independent researcher can reconstruct the result from scratch. This is the opposite of a black-box signal.

Replication is the second layer. After publication, other research groups around the world independently attempt to reproduce the result — often on different datasets, different time periods, different markets. A strategy that survives multiple independent replications is a strategy that reflects something real about how markets work, not just a quirk in one researcher's data.

The Replication Crisis in Finance

There's a sobering body of research on how many published factors actually hold up. Campbell Harvey, Yan Liu, and Heqing Zhu catalogued the academic literature on cross-sectional return predictors and found that over 300 factors had been claimed in peer-reviewed journals by 2016.

"Most claimed research findings in financial economics are likely false."
— Harvey, Liu & Zhu (2016), Review of Financial Studies

Their argument: the conventional statistical threshold for significance (a t-statistic above 2.0) was calibrated for a world where researchers test one or two hypotheses. When you're searching across hundreds of potential factors, that threshold generates an enormous number of false positives by chance alone. To account for the multiple-testing problem across the full factor zoo, they proposed the bar should be closer to a t-statistic of 3.0 or higher.

The message is not that academic finance is broken — it's that even peer review isn't a perfect filter, and the highest-confidence signals are those that have been replicated across independent studies, multiple asset classes, and long out-of-sample windows.

What Survives Replication

A handful of factors show up reliably enough, across enough contexts, that the academic consensus treats them as genuine market phenomena rather than statistical artifacts. These tend to share a common property: there is a plausible economic mechanism that explains why the premium should exist and why it might persist even after discovery.

• Momentum. Assets that have performed well over the past 6–12 months tend to continue outperforming over the next few months. Documented across equities, bonds, commodities, and currencies. First formally described by Jegadeesh & Titman (1993).
• Low volatility. Lower-risk assets have historically delivered better risk-adjusted returns than theory predicts. Consistent with the leverage constraints many institutional investors face.
• Value. Assets trading cheaply relative to fundamentals tend to outperform over long horizons. The mechanism is debated — risk compensation or behavioral mispricing — but the empirical pattern is robust.
• Carry. High-yielding assets outperform low-yielding ones on a risk-adjusted basis across currencies, fixed income, and commodities.
• Size. The small-cap premium exists but is sensitive to implementation costs and has weakened since publication — a useful reminder that no factor is unconditional.

These are not obscure discoveries. They have been documented in dozens of independent studies, across markets and time periods spanning more than a century in some cases. That breadth of evidence is what distinguishes them from the long tail of factors that look compelling in a single dataset but evaporate under scrutiny.

How R2S Applies This

Every strategy on R2S must clear three requirements before it gets implemented:

• Published peer-reviewed paper. The strategy must be documented in a refereed academic journal. We do not implement strategies based on conference proceedings, blog posts, or internal research alone.
• Full methodology disclosure. The paper must describe the signal construction in enough detail to replicate it. If we can't verify the logic independently, we don't use it.
• Replicable from public data. The signal must be constructible from publicly available data sources. This keeps the process transparent and auditable — no proprietary data dependencies that subscribers can't verify.

The current strategy library covers 20+ strategies across equities, fixed income, commodities, and currencies. Each one links to the source paper directly in the dashboard, so you can read the original research yourself rather than taking our word for it.

This Is Not a Guarantee

It would be dishonest to end here without the most important caveat: peer-reviewed does not mean risk-free, and it does not mean the strategy will work during the period you happen to be subscribed.

Even the most robustly documented factors go through extended drawdown periods. Momentum, for example, experienced sharp reversals in 2009, 2020, and other periods where market conditions shifted rapidly. Value underperformed for most of the 2010s before recovering. These are features of the strategies, not bugs — the premium exists precisely because investors face real pain holding the position during those difficult stretches.

The edge from systematic, research-backed strategies is probabilistic, not deterministic. It says: over a sufficiently long horizon, this approach has a meaningful probability of outperforming a passive benchmark. It does not say anything about any particular month or year.

What peer review buys you is not certainty. It buys you a much higher prior probability that the underlying signal reflects something real. That is worth a great deal compared to the alternative — but it should be understood clearly for what it is.