Base scores combine an assessment of a stock’s fundamental attractiveness (relative to its regional/sector peers) and level of controversy (relative to the regional index). The principle is that the more attractive and controversial a stock, the more likely it is to deliver future outsized returns. Scores are not trading recommendations but a distillation of fundamental insights to help investors get rapidly up to speed on a stock’s investment thesis. It is recommended to build portfolios with base scores well above 5.0 which is typically the score of market indices. A score of 5.0 indicate an expected 1-year return in line with the index.


Each investible stock within its region is given a score from 0 to 10:

  • There are 4 defined regions: Europe, Japan, North America and Emerging Markets. While for the most part, the assigned region corresponds to that of primary listing, the main criterion is risk, so a Chinese Tech stock with a US primary listing will be assigned to Emerging Markets
  • Developed markets stocks get further assigned a sector theme: Production (commodities and industrials), Transaction (financials and real estate), Consumption (staples and discretionary), or Innovation (tech and pharma). Conversely, emerging markets stocks get further assigned to 3 sub-regions: East Asia, West Asia and Rest of Emerging Markets.
  • Scores above 8.5 regroup the top 10% of stocks most likely to be future winners over the next 12 months in a given region (Top 20% for stocks with scores above 7.5, 30% above 6.5), with around 65% chance of doing better than their regional index (2011-2017 average) – this doesn’t mean the stocks are outright buys, just that they are a good hunting ground for opportunities.
  • The odds drop toward 50% chance as scores approach 5.0 – this doesn’t mean these are bad stocks, just that the machine doesn’t find sufficiently salient information to take sides either way.
  • Stocks with scores below 1.5 (bottom 10%) carry a 65% chance of underperformance -- this doesn’t mean the stocks are outright sells or shorts, in fact a small minority will even end up outperforming most suddenly and spectacularly (e.g. following a rights issue, a take-over, a litigation settlement, a commodity price rally), it’s just that the machine sees on balance too negative an array of signals to nudge users toward them.

The model looks for an information edge based on a stock’s array of fundamental characteristics, the combination of which determines its overall fundamental attractiveness, that is, how much excess return can be expected for this stock per unit of risk. The factors are the same across regions and sector themes although their relevance vary between them – likewise the model will dynamically adjust the importance of each factor to the economic cycle, as a stock may look less fundamentally attractive in periods of economic expansion than it does in a recession.

The proprietary factors used to compute a stock’s edge, or expected risk-adjusted excess return, are all translated from the world of fundamental investing, not from data mining. They can be broadly assigned to three categories:


Both that of the:


(high score = strong business position) with factors reflecting the capacity of the firm to create (sustainable/growing) economic value, such as:

  • Long-term Economic Value Added (or EVA, a measure of returns above/below the company’s cost of capital), EVA momentum, and acceleration of EVA momentum – where we differ in our approach to EVA is in the way we adjust a company’s cost of capital and specifically its cost of equity by considering its historical returns and returns volatility rather than simply relying on the covariance of historical share price relative to a market index. As a result, a company with historically low and volatile fundamental returns will typically get assigned a higher (and more reliable) cost of capital than the very wide and unstable range of costs of capital that result from only exploiting share price patterns.
  • Long-term volatility of returns on capital and long-term returns to shareholders per unit of share price volatility
  • Sustainable growth potential without taking on more debt


(high score = safe capital structure) which gauges the firm’s capacity at honouring its financial commitments to creditors and shareholders, for instance:

  • Level of indebtment and non-debt long-term liabilities
  • Dividend sustainability (in relation to expected free cash flow generation and historical dividend growth)
  • Likelihood of financial distress (Merton’s distance to default formula)


Consisting of:

Market Valuation

Of shares, (high score = undemanding valuation) taking a multi-dimensional view, from country to sector-relative and from equity to firm-based metrics, for instance:

  • Free Cash Flow yield (as a percentage of enterprise value)
  • Country relative and sector relative price multiples
  • EV/EBIT premium (or discount), which reflects how much long-term growth (decline) or margin increase (decrease) would be required to justify the company’s current EV/EBIT multiple

Management Behaviour

(high score = high alignment with shareholders’ interests) inferred by management’s share dealings, corporate actions, and reporting quality, looking at:

  • Change in number of shares outstanding
  • Change in insiders’ shareholding
  • Change in operating accruals


Consisting of:

Fundamental momentum

(high score = significantly improving business prospects) to uncover trends in earnings, returns, and investments, for instance:

  • Consensus earnings revision (past 3 months)
  • Price response following last period’s earnings report
  • Change in return on assets in relation to change in capital expenditures as a percentage of sales (next 4 quarters vs. last year) – this metric favour either low EVA stocks that are about to pass their peak capex level and start enjoying higher returns, or high EVA stocks that keep investing for growth; conversely it penalises low EVA stocks that keep (over)-investing and high EVA stocks that have reached “peak growth”

Technical momentum

(high score = positive stock sentiment) reflects the overall stock sentiment observed over the past couple of quarters, as illustrated by:

  • Net institutional buyers
  • Relative share price strength
  • Change in price to book multiple in relation to earnings growth trajectory


The stock controversy is then factored in with the above factors.

Controversy reflects the stock’s “active risk”. It denotes how much a stock tends to “behave” differently than what its historical correlation to the market index would suggest. The more “different” (controversial) a stock, the more its potential for “ab-normal” (preferably above-normal) returns. Conversely, very low controversy stocks are more likely to deliver “benchmark-like returns” if all goes well, and they carry the added risk of suddenly becoming controversial and to underperform as a result. While having a portfolio with above average controversy score is desirable, too much controversy or holding too many extremely controversial stocks (scores above 9.5 / 10) usually leads to overly aggressive risk profiles that expose the portfolio to large (absolute and/relative) drawdowns. Note that high (low) volatility does not necessarily equal high (low) controversy: a highly volatile mining stock may be “reliably” volatile relative to the bench and carry a low controversy score, whereas a much less volatile beverage stock may end up with a high controversy score thanks to a very idiosyncratic share price behaviour.

Butterwire’s engine produces two base scores: an upcycle and a downcycle one, depending on how good or bad it estimates global economic conditions to be at the time of computation. As a result, while both scores are reported, the app singles out the one that it believes carries the most reliable information given prevailing market conditions. For instance, the scores represented below points to an already high and rising score of 9.5 in the prevailing downcycle conditions, while the picture painted would have been different in upcycle conditions with a mediocre and fast declining score standing at 6.8.

Butterwire's Upcycle and Downcycle Base Scores

For the more mathematically minded:

base scores are the cumulative normal distribution (from 0% to 100%) of (ex-ante) stock alphas, multiplied by 10 (so for instance a stock with 0% alpha will have a base score of 50% x 10 = 5). Stock alphas are estimated in a way that yields a satisfactory out of sample information coefficient (ie. correlation between ex-ante and ex-post alphas) through economic cycles. Underpinning the information coefficient are z-scores computed using the (winsorised and normalised) stock values calculated for each of the factors described above, conditional on their region of risk and sector theme (as the regional/theme groupings described above were found to represent optimal clusters in terms of covariance x size). In summary, the components of alphas are confirmed ex-post and estimated ex-ante as follows: