Methodology

The problem

Financial markets are noisy. In any given quarter, a great company can fall 30% for reasons that have nothing to do with its fundamentals, and a terrible one can rise 40% on a short squeeze. Luck dominates short-term outcomes. That's why most stock-picking systems are built on hope.

Across thousands of companies and over multi-year horizons, something else shows up. Certain combinations of fundamental characteristics — things you can measure in the quarterly filings — have preceded sustained outperformance more often than not. Not every time, and not in every period. But consistently enough across decades and across markets to take seriously.

This isn't new. Graham and Dodd were writing about it in 1934. Fama and French formalised parts of it in 1992. James O'Shaughnessy published What Works on Wall Street in 1996, testing dozens of factor strategies against 43 years of data. Novy-Marx extended the profitability literature in 2013. Asness, Frazzini and Pedersen refined the quality framework in 2019. What's new is what technology now makes possible: doing this for every public company, every quarter, in close to real time.

MonkScore™ is our answer to a single question: after seven decades of peer-reviewed research on what makes stocks outperform, what does the evidence look like when it's distilled into one number per company?

How we built it

Before writing a single line of code, we spent months reading academic research. More than 300 papers, from the early Fama-French work of the 1990s through to the recent quality-minus-junk and intangibles literature. Papers from the Journal of Finance, the Journal of Financial Economics, the Review of Financial Studies. Work by researchers who have spent their careers on one question: which company characteristics actually predict forward stock returns, and which only look like they do.

Out of that we pulled a large set of candidate metrics — every ratio and signal that anyone in the peer-reviewed literature had shown could predict returns. Then we started testing.

We tested each candidate on its own and in combinations. Across decades, across regions, across sectors, through different market regimes. We were ruthless about what we kept. A factor had to be statistically significant on more than one dimension, work in more than one decade, work in more than one geography, and have a plausible economic mechanism behind it — a real reason why it should drive business value, not just a pattern in the data. Anything that failed any of those tests got cut.

What survived became MonkScore: five pillars, each built from a small set of factors with academic backing and empirical validation, tested against the biggest risk in quant investing — that you've fit the past instead of finding something real.

Every component of MonkScore is in there because the academic research said it should work, and our own testing showed it does.

The five pillars

Every company is scored across five dimensions. Each pillar measures something economically distinct, and each one rests on academic research that identified it as a real predictor of long-term returns.

Growth — is the company genuinely expanding?

Growth is the trickiest of the five pillars in factor research, and it's worth being clear about what MonkScore does with it. The academic literature (Lakonishok, Shleifer and Vishny, 1994; La Porta, 1996; Chan, Karceski and Lakonishok, 2003) has established that past growth on its own is a poor predictor of future returns. The "glamour stocks" identified by trailing growth tend to underperform.

The Growth pillar is built around that finding, not against it. What actually predicts returns is the quality, persistence, and breadth of growth: expansion that shows up across multiple parts of the business, holds up over time, and is confirmed by forward expectations. A single strong trailing metric on its own triggers nothing.

That's how Growth ends up being a positive predictor in MonkScore despite the literature's warnings about glamour stocks. It measures the shape of growth across the business, not its magnitude at any one point.

Profitability — does the business model generate real economic returns?

Profitability is the most academically settled of the five pillars. Novy-Marx (2013), Ball, Gerakos, Linnainmaa and Nikolaev (2016), and Fama and French (2015) all show that companies earning higher returns on the capital they deploy outperform over long horizons. The mechanism is compounding. A business that earns 20% on its capital compounds wealth far faster than one earning 5%, and over a decade the gap is enormous.

The pillar evaluates profitability at several layers, from gross economics down to returns on total invested capital, and cross-checks accounting profits against cash generation. The point of the cross-check is to confirm that reported earnings actually convert to cash, rather than just sitting on the income statement.

Quality — how well is the company managed?

Quality captures what academics call "capital allocation discipline": the management decisions that determine whether a profitable business stays profitable, or whether returns erode over time through poor reinvestment, weakening balance sheets, or earnings manipulation.

The research base is unusually deep here. Sloan (1996) on earnings quality. Piotroski (2000) on fundamental strength. Soliman (2008) on the DuPont decomposition. Lev and Thiagarajan (1993) on capital expenditure efficiency. Together they describe a company that's profitable today and likely to stay that way.

Conviction — has the market priced in confidence?

This pillar starts from a principle that's easy to forget: the market is not stupid. Stock prices reflect the collective judgment of thousands of analysts, portfolio managers, and investors, all of them processing information and paid to get it right. When a company trades at a sustained premium to its peers, not for a quarter but consistently, that premium is the market saying it keeps seeing evidence of quality that justifies the price.

David Gardner of The Motley Fool was making this argument decades before the academic literature caught up. His Rule Breakers framework rests on the observation that the stocks most investors dismiss as "overvalued" often go on to generate the greatest long-term returns, because the crowd repricing them upward is processing real information about the business. Selling a winner just because it's become expensive is, in his view, a systematic mistake.

The academic evidence supports him. Jegadeesh and Titman (1993) showed that recent winners continue to outperform. Asness, Frazzini and Pedersen (2019) showed that high-quality companies reliably trade at premium valuations and continue to outperform despite those premiums. Together, the literature points at something that feels counterintuitive to traditional value investors but is empirically clear: winners keep winning, and a sustained premium on a quality business is a confirmation signal, not a warning.

The Conviction pillar is evaluated across several independent valuation measures, requiring agreement among them so the result isn't driven by a quirk in any one metric.

Safety — are there warning signs?

Safety is different from the other four pillars. It flags warning signs rather than rewarding strengths: specific patterns in a company's fundamentals that the research has identified as leading indicators of future deterioration.

Three failure modes do most of the work, each independent of the others and each grounded in peer-reviewed research: valuation-based distress dynamics (Campbell, Hilscher and Szilagyi, 2008), fundamental deceleration patterns, and accruals-based earnings quality risk (Sloan, 1996). Other metrics contribute, but these three mostly determine whether a company gets a high or low Safety score.

A high Safety score means none of the patterns historically associated with sharp forward deterioration are showing up in the current data. It does not mean the company is risk-free. Nothing is.

How the pillars combine

The five pillars combine using a geometric mean. This was a deliberate choice. A geometric mean penalises imbalance in ways a simple average can't. A company that excels on one dimension but fails on another scores lower than one that does well on all five. The highest-scoring companies are the ones strong across all five at the same time — a profile that's rare and historically predictive.

How we know it works

Every quantitative signal has to be tested against data. A signal that can't show statistical significance across different time periods, market regimes, and regions isn't a signal. It's noise.

We've validated MonkScore using the same methods peer-reviewed factor research uses, across more than 330,000 company-quarter observations: 25 years of North American data and a decade of international data. Here's what came out.

25 years of North American data

Metric	Result
Quarterly cohorts tested	99 (2000 Q1 – 2024 Q3)
Company-quarter observations (North America)	141,870
Average top-vs-bottom quintile spread	+13.8 percentage points per year
Positive-signal quarters	94 out of 99 (94.9%)
Newey-West t-statistic (3 lags)	9.20
Information Coefficient (5-year)	0.248
Information Ratio (5-year)	3.50

The t-statistic of 9.20 exceeds the Harvey, Liu and Zhu (2016) threshold for multiple-testing-adjusted statistical significance — the most demanding bar in factor research — by a factor of three.

International confirmation

Across 40 quarterly cohorts covering Europe, Asia-Pacific, and Japan, the signal was positive in 40 of 40 quarters. The t-statistic was 8.36. The result extends cleanly outside North America.

The signal is strengthening over time

One of the most telling tests of whether a factor model is capturing real alpha or just overfit to one historical period is what happens to its signal over time. Overfit models decay. MonkScore has done the opposite:

Decade	Average quarterly spread
2000s	+12.1 pp
2010s	+13.5 pp
2020s	+18.1 pp

Overfit signals decay as the market learns. Signals built on real economic mechanisms tend to persist. MonkScore's decade-over-decade strengthening is consistent with the second.

Regime robustness

The signal held across 6 of 8 major market regimes tested: the dot-com crash, the 2003–2007 value rally, the post-2008 recovery, the ZIRP growth era, and the 2022 rate hike cycle were all positive. The only negative quarters across 25 years came during the acute Global Financial Crisis shock (three quarters in 2008–2009) and the initial COVID crash (two quarters in 2020). Five quarters in total across two and a half decades.

What the top 1% looks like

The rarest MonkScore — a perfect 100, hit by roughly 1% of the universe at any given time — gives the cleanest picture of what the model is picking up:

Metric	Score = 100
Observations (25 years)	1,599
Average 5-year annualised return	+25.9%
Median 5-year annualised return	+18.3%
Positive 5-year outcomes	91.4%
5-year entry periods with negative median returns	zero (0 of 21)
Median performance in 2008 (S&P 500: –37%)	+1.6%

None of this is a guarantee. It's the historical behaviour of a single carefully built signal, tested across 25 years and six major market regimes.

The regime risk you should know about

It's worth being upfront about where this model could fail.

Two of the five pillars, Conviction and Growth, are regime-dependent. Both have delivered strong forward-return predictions across the full 25 years of North American data and the 15 years of international data. But both depend on market conditions that favour quality-momentum over pure mean reversion.

Conviction rewards companies trading at premium valuations. Growth rewards companies whose multi-dimensional expansion is confirmed by forward expectations. In a sustained value rotation, the kind that dominated 2000–2007 before reversing, both signals could weaken at the same time, compressing MonkScore's forward returns.

The geometric mean gives partial protection. A company can't reach the top of MonkScore on Conviction or Growth alone. It also has to score well on Profitability, Quality, and Safety, which together account for 60% of the composite and rest on findings that have held across more than 60 years of academic evidence. The highest-scoring companies are expensive and growing because the underlying fundamentals justify it.

Partial is not full, though. The 25-year backtest includes the most severe value regime of the modern era (2000–2007), and MonkScore stayed positive throughout. That's reassuring. It is not a guarantee. We publish this risk in plain sight so every subscriber can decide for themselves how to weight MonkScore in their process.

What MonkScore isn't designed to do

• Not a market timing signal. MonkScore ranks companies within the current universe; it doesn't tell you when to be in or out of the market.
• Not a short-term trading signal. The statistical evidence is strongest at 3-year and 5-year horizons.
• Not a substitute for diversification. Even the highest-scoring stocks can underperform individually, and they do. The signal describes the behaviour of groups of high-scoring stocks on average.
• Not a substitute for professional advice. Significant financial decisions warrant a licensed investment advisor who can consider your full situation.
• Not static. Scores are recomputed several times a day as new market data arrives and peer rankings shift.

Ongoing validation

A signal that worked yesterday may not work tomorrow. We monitor MonkScore continuously. Every quarter we revalidate the quintile spread and hit rate as new cohorts come in. Every year we run a regime analysis to check whether the signal is weakening in specific market environments. Any factor we want to add or remove has to clear a pre-registered validation protocol before it goes into the composite.

When we make changes, we document them. When we find something that contradicts a previous claim, we publish the correction. This page gets updated as the research evolves.

What MonkScore is, and what it isn't

MonkScore™ is a quantitative research tool. An input into your investment process, not a substitute for it. It is not a solicitation, a recommendation, or advice to transact in any security. It doesn't know your personal financial situation, your risk tolerance, your tax position, or your time horizon. Past performance, however rigorously measured, doesn't guarantee future results.

We built MonkScore because we think independent advisors and serious retail investors deserve the same quality of factor analysis that institutional quant shops have had for decades. The difference is we're showing our work — publicly, in plain language, with the caveats and the limitations included.

Important disclosures

MonkScore™ is a quantitative research tool. It is not investment advice, a recommendation to buy or sell any security, or a solicitation to engage in any investment strategy. MonkStreet is not a registered investment advisor.

All statistics cited on this page are derived from historical backtesting. Past performance is not indicative of future results. No investor should rely on historical simulations as a guarantee of future returns. The statistical significance of a historical signal does not guarantee that the signal will persist in future market environments.

MonkScore is computed from fundamental financial data provided by third-party vendors. While we take reasonable care to ensure the accuracy of the underlying data, we cannot guarantee that all inputs are error-free. Errors in source data will propagate to scores.

Investors should consult a licensed investment advisor before making investment decisions, particularly for significant allocations of capital. MonkStreet and its affiliates accept no liability for investment decisions made on the basis of MonkScore or any information presented on this page.

For questions about methodology or to report an error in the documentation, contact [email protected].