SharpeMetrix Research · November 16, 2026
Each month SharpeMetrix scores the Russell 3000 on seven signal families. Those scores feed the Power List and the 15 model portfolios. This paper explains what each family measures, why the underlying effect has been studied for decades, and how the seven are combined into one signal that portfolio construction can use.
The short version: no single signal is reliable every month. Combining several signals that measure different things, each grounded in published research, gives a steadier ranking than any one of them alone.
The seven families at a glance
| Family | What it measures | Data type |
|---|---|---|
| Value | Price relative to forward-looking earnings, book value and cash flow, compared within sector | Fundamental |
| Momentum | Price strength relative to economically linked peers | Market |
| Quality | Sustainability of earnings, cash conversion, accruals and balance-sheet strength | Fundamental |
| Low Volatility | The part of volatility associated with future returns, adjusted for other factor exposures | Market |
| Earnings | Forecasts of earnings and revenue surprises, built from analyst estimates weighted by each analyst’s track record | Alternative |
| Transcripts | Tone, wording and patterns in management commentary on earnings calls | Alternative |
| Innovation | Patent activity and specialised hiring (H1B visa filings) at company and industry level | Alternative |
The three alternative data families use the same kind of institutional data licensed by hedge funds and asset managers. The four traditional families are refined versions of factors that appear in most academic and practitioner factor models.
Value
What it measures. Whether a company is cheap relative to what it is expected to earn. SharpeMetrix uses forward earnings expectations, taken from its own Earnings model, instead of trailing earnings alone, and combines several measures: price to earnings, price to book and cash-flow yields. Each company is compared with its sector peers, because a bank, a utility and a software firm carry very different “normal” multiples. Valuation is also adjusted for earnings quality, so a low multiple on unsustainable earnings does not score as cheap.
Why it has worked. Fama and French (1992, 1993) documented that stocks with high book-to-market ratios earned higher average returns than those with low ratios, and built that observation into the three-factor model. Lakonishok, Shleifer and Vishny (1994) argued that part of the value premium reflects investors extrapolating past growth too far. Asness, Moskowitz and Pedersen (2013) found value premiums across countries and asset classes, not only in US equities.
What to expect. Value can lag for long stretches. The late 1990s and much of the 2010s were difficult periods for simple value strategies. This is one reason value is one input of seven rather than the whole model.
Momentum
What it measures. Whether a company’s shares have been outperforming. Standard momentum compares a stock with the whole market. SharpeMetrix compares each company with economically linked firms, such as customers, suppliers and close competitors, and controls for market-wide moves. The goal is to isolate company-specific strength rather than a sector or market trend.
Why it has worked. Jegadeesh and Titman (1993) showed that US stocks with the highest returns over the prior three to twelve months tended to keep outperforming over the following months. Chan, Jegadeesh and Lakonishok (1996) linked part of this to slow reaction to earnings news. Cohen and Frazzini (2008) found that news about a firm’s major customers is reflected in the supplier’s share price with a delay, which is the basis for using economic links. The research note Momentum Among Economic Peers covers this in more detail.
What to expect. Momentum can reverse sharply after market turning points. Daniel and Moskowitz (2016) document these “momentum crashes”, which tend to follow steep market declines.
Quality
What it measures. Whether reported earnings are durable. The Quality family looks at the split between recurring and one-time items, whether earnings turn into cash, balance-sheet strength and the size and composition of accruals. Accruals are weighted by how persistent they have historically been.
Why it has worked. Sloan (1996) found that companies whose earnings were driven mostly by accruals rather than cash flow tended to see weaker future earnings and returns. Novy-Marx (2013) showed that gross profitability predicts returns about as well as book-to-market, and that the two work well together. Asness, Frazzini and Pedersen (2019) found that profitable, stable, well-managed companies earned higher risk-adjusted returns than lower-quality companies.
What to expect. Quality tends to hold up in weak markets and can lag in sharp, speculative rallies.
Low Volatility
What it measures. The part of a stock’s volatility that is associated with future returns, after controlling for the company’s other factor exposures and sector. Volatility driven by changing fundamentals is treated differently from volatility driven by trading and sentiment.
Why it has worked. Classical finance suggests riskier stocks should earn more. The evidence has often pointed the other way. Ang, Hodrick, Xing and Zhang (2006) found that stocks with high idiosyncratic volatility had low subsequent returns. Frazzini and Pedersen (2014) showed that portfolios of low-beta assets earned higher risk-adjusted returns than high-beta portfolios, and attributed the effect to investors who cannot use leverage bidding up riskier assets. Baker, Bradley and Wurgler (2011) argued that benchmark-relative mandates discourage managers from exploiting the effect.
What to expect. Low-volatility tilts can trail in strong, narrow rallies led by high-beta stocks.
Earnings
What it measures. Whether a company is likely to beat or miss consensus expectations for earnings and revenue. The model does not treat all analysts equally. It weights each analyst’s estimate by how accurate that analyst has been in the past, and adds information from estimate revisions, management guidance, peer trends, signs of earnings management and the company’s own earnings history.
Why it has worked. Bernard and Thomas (1989) documented post-earnings-announcement drift: prices continue to move in the direction of an earnings surprise for weeks afterwards. If surprises can be forecast, and markets react to them slowly, a forecast of the surprise carries information about returns. Consensus is a simple average and adjusts slowly; weighting forecasters by accuracy is a direct way to improve on it.
Transcripts
What it measures. The language management uses on earnings calls: tone, word choice, the confidence of forward-looking statements, and how all of these compare with the same company’s previous calls. SharpeMetrix uses a natural language processing engine trained on more than 200,000 calls since 1999, combining finance-specific word lists with word embeddings and a machine learning model.
Why it has worked. Loughran and McDonald (2011) showed that general-purpose sentiment dictionaries misclassify many common financial words and built a finance-specific alternative. Price, Doran, Peterson and Bliss (2012) found that the tone of earnings conference calls, especially the question-and-answer section, helps explain returns around and after the call. Mayew and Venkatachalam (2012) found that vocal cues in managers’ speech carry information about future performance. The white paper What Earnings Calls Reveal covers the method and its pitfalls.
Innovation
What it measures. How much a company invests in new products and processes, and how productive that investment is. The signal uses patent data and H1B visa filings, which track the hiring of specialised talent, at both company and industry level.
Why it has worked. Chan, Lakonishok and Sougiannis (2001) found that R&D-intensive companies with poor recent returns went on to earn higher returns, consistent with the market undervaluing R&D that is expensed rather than capitalised. Hirshleifer, Hsu and Li (2013) found that firms producing more patents and citations per dollar of R&D earned higher subsequent returns. Cohen, Diether and Malloy (2013) showed that firms with a record of turning R&D into results were undervalued. See Innovation as a Leading Indicator.
From raw data to a 0–100 score
Raw measures do not compare well across companies. A price-to-earnings ratio of 15 means one thing for a utility and another for a software firm, and patent counts mean little in sectors that do not patent. Each family is therefore converted into a 0–100 score that reflects rank, not raw level:
- Compute the raw measures for each company from the latest data available at the time of scoring.
- Adjust for sector and comparability where it matters. Value and Quality are judged within sector; Momentum is judged against economic peers.
- Rank and scale. Companies are ranked within the comparison group and the rank is scaled to 0–100. A score of 80 means the company ranks ahead of roughly 80% of its comparison group on that family.
Ranks are robust to outliers. One company with an extreme multiple does not distort everyone else’s score.
Combining the seven
The seven family scores are blended into one composite signal each month. The weighting is set by rules, not by month-to-month judgement, and the same composite drives the multi-factor models and the Power List. Single-family models, such as the momentum, value/quality and earnings-signal models, use the relevant family directly.
Why combine at all
The case for combining rests on correlation. A useful way to think about a signal is its information coefficient (IC): the correlation between the signal and subsequent returns across stocks. Grinold (1989) showed that a strategy’s expected risk-adjusted return scales with IC and with the square root of the number of independent bets.
If two standardised signals each have an IC of 0.03, an equal-weighted blend of them has an IC of:
IC(blend) = (0.03 + 0.03) / √(2 + 2ρ)
where ρ is the correlation between the two signals. The arithmetic for three cases:
| Correlation between signals (ρ) | Blend IC | Change vs one signal |
|---|---|---|
| 1.0 (same signal twice) | 0.06 / √4 = 0.030 | none |
| 0.5 | 0.06 / √3 = 0.035 | about +15% |
| 0.0 | 0.06 / √2 = 0.042 | about +41% |
These are stylised numbers to show the mechanics, not SharpeMetrix estimates. The point holds generally: a signal that measures something different adds more than a second version of something already measured. Value and momentum are a well-known example. Asness, Moskowitz and Pedersen (2013) found them negatively correlated with each other, so a combination was steadier than either alone. Earnings-call language, patent activity and analyst accuracy are drawn from data that price and accounting factors do not use, which is the reason to include them.
From signal to portfolio
The composite ranks stocks. Portfolio construction turns the ranking into holdings, with limits set in advance:
| Rule | Concentrated models (30–50 stocks) | 100-stock and optimised books |
|---|---|---|
| Maximum position | 10% | 3% |
| Sector weight vs benchmark | ±7% | ±7% |
| Rebalance | Monthly | Monthly |
| Turnover limit | Applied | 20% for optimised and index-enhanced books |
The Power List is the plain ranking with no construction limits beyond equal weighting.
Limits of the approach
Every signal family described here has had multi-year periods of weak results. McLean and Pontiff (2016) found that published anomalies weaken after publication, and Harvey, Liu and Zhu (2016) warned that many reported factors are products of data mining. A systematic process does not remove these risks. It makes them visible, applies the same rules each month, and spreads exposure across several independent sources of information.
References
- Ang, A., Hodrick, R. J., Xing, Y., & Zhang, X. (2006). The cross-section of volatility and expected returns. Journal of Finance, 61(1), 259–299.
- Asness, C. S., Frazzini, A., & Pedersen, L. H. (2019). Quality minus junk. Review of Accounting Studies, 24(1), 34–112.
- Asness, C. S., Moskowitz, T. J., & Pedersen, L. H. (2013). Value and momentum everywhere. Journal of Finance, 68(3), 929–985.
- Baker, M., Bradley, B., & Wurgler, J. (2011). Benchmarks as limits to arbitrage: Understanding the low-volatility anomaly. Financial Analysts Journal, 67(1), 40–54.
- Bernard, V. L., & Thomas, J. K. (1989). Post-earnings-announcement drift: Delayed price response or risk premium? Journal of Accounting Research, 27(Supplement), 1–36.
- Chan, L. K. C., Jegadeesh, N., & Lakonishok, J. (1996). Momentum strategies. Journal of Finance, 51(5), 1681–1713.
- Chan, L. K. C., Lakonishok, J., & Sougiannis, T. (2001). The stock market valuation of research and development expenditures. Journal of Finance, 56(6), 2431–2456.
- Cohen, L., Diether, K., & Malloy, C. (2013). Misvaluing innovation. Review of Financial Studies, 26(3), 635–666.
- Cohen, L., & Frazzini, A. (2008). Economic links and predictable returns. Journal of Finance, 63(4), 1977–2011.
- Daniel, K., & Moskowitz, T. J. (2016). Momentum crashes. Journal of Financial Economics, 122(2), 221–247.
- Fama, E. F., & French, K. R. (1992). The cross-section of expected stock returns. Journal of Finance, 47(2), 427–465.
- Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33(1), 3–56.
- Frazzini, A., & Pedersen, L. H. (2014). Betting against beta. Journal of Financial Economics, 111(1), 1–25.
- Grinold, R. C. (1989). The fundamental law of active management. Journal of Portfolio Management, 15(3), 30–37.
- Harvey, C. R., Liu, Y., & Zhu, H. (2016). …and the cross-section of expected returns. Review of Financial Studies, 29(1), 5–68.
- Hirshleifer, D., Hsu, P.-H., & Li, D. (2013). Innovative efficiency and stock returns. Journal of Financial Economics, 107(3), 632–654.
- Jegadeesh, N., & Titman, S. (1993). Returns to buying winners and selling losers: Implications for stock market efficiency. Journal of Finance, 48(1), 65–91.
- Lakonishok, J., Shleifer, A., & Vishny, R. W. (1994). Contrarian investment, extrapolation, and risk. Journal of Finance, 49(5), 1541–1578.
- Loughran, T., & McDonald, B. (2011). When is a liability not a liability? Textual analysis, dictionaries, and 10-Ks. Journal of Finance, 66(1), 35–65.
- Mayew, W. J., & Venkatachalam, M. (2012). The power of voice: Managerial affective states and future firm performance. Journal of Finance, 67(1), 1–43.
- McLean, R. D., & Pontiff, J. (2016). Does academic research destroy stock return predictability? Journal of Finance, 71(1), 5–32.
- Novy-Marx, R. (2013). The other side of value: The gross profitability premium. Journal of Financial Economics, 108(1), 1–28.
- Price, S. M., Doran, J. S., Peterson, D. R., & Bliss, B. A. (2012). Earnings conference calls and stock returns: The incremental informativeness of textual tone. Journal of Banking & Finance, 36(4), 992–1011.
- Sloan, R. G. (1996). Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review, 71(3), 289–315.