< Back to Research

Butterwire’s Three-Year Track Record: Delivering What and How It Set Out To Deliver

Raphael Fiorentino
Raphael Fiorentino
10th January 2020 - 5 min read

Revised version, 22/01/2020

Anecdotal track record: Butterwire’s inaugural global portfolio

It’s been 3 years since we launched our first website to introduce Butterwire’s approach to stock selection, portfolio construction, and global macro nowcasting, as well as to showcase what it set out to accomplish. So in Jan-17, one year before the beta version of the Butterwire app was released and 27 months before commercial launch, we used the website to post and report on a typical Butterwire portfolio: concentrated, global, low factor/high stock-specific risk exposures, low turnover. This inaugural portfolio (available as “butterwire2017” in the Portfolios section of the app) returned 61% to date with 0 turnover and has outperformed the MSCI ACWI index by 19%.

Portfolio Performance

More than the performance itself which has plateaued since the global macro context turned more adverse around mid-2018, we hoped the portfolio would illustrate how Butterwire seeks to contribute to performance, and it did deliver as:

  • A few holdings (notably Ferrari, Apple, Tatneft, Hoya, VMWare) delivered the kind of outsized returns that alone determined the overall portfolio performance, while only one holding (Pandora) ended up a major detractor
  • None of the outperformance is attributable to factor timing, with stock specifics (i.e. idiosyncratic risk) pretty much the only driver of returns (see below)
  • Despite numerous stock flags and alerts appearing along the way, the portfolio retained a satisfactory return/risk profile throughout the period despite no action taken, and maximum drawdowns remained within expected limits (-5.4% relative).

Excess Return & Risk Factors

Systematic track record: Butterwire’s Candidates Indices

The Butterwire engine tracks several baskets of stocks based on their “Interesting Research Candidate?” status. Each basket is rebalanced on the first day of each month and performance is tracked and made available in the Markets section. Notable indices include:

  • The “All Interesting Candidates” index which is composed of all the stocks within each region, regardless of size, with an assigned candidate status of “Interesting” or “Interesting but High Octane”. Typical number of stocks in these baskets are 300 for North America and EM, 200 for Europe, 100 for Japan.
  • The “Large Cap Recession-Resilient Candidates” is a subset of the above that excludes any candidate with a below regional median market cap or a recession resilience score under 7.5. Typical number of stocks in these baskets are 100 for North America and EM, 70 for Europe, and 35 for Japan.
  • The “Large Cap Shorts” index consists of stocks with an assigned candidate status of “Potential Short” and a market cap above that of the regional median. Typical number of stocks in these baskets are 60 for North America and EM, 40 for Europe, and 20 for Japan.

Below is a summary of how each of the above indices behaved over the past 3 years within each region, relative to the corresponding regional MSCI index. All the regions have been contributing positively on the long and the short side (ex-Japan). Furthermore, the level of active return is generally in line with the level of active risk and the realised maximum drawdown, thereby producing reasonably clustered levels of Information Ratios (IR) across indices and regions.

Candidate Performance

Returns Table

The same applies when switching from a regional to a sectorial view, albeit with a broader range of return/risk profiles as reflected by the below graphs where the Innovation meta-sector (Tech + Pharma) consistently delivered wide long-short spreads in contrast to the narrow spreads for the Transaction meta-sector (Financials + Real Estate). Likewise, the range of IRs delivered across sectors is wider than that observed by region.

Candidate Performance 2

Returns Table 2

Track record in perspective

Consistent ability of the engine to beat a random selection

Between Jan-17 and Dec-19, 62% of all Innovation (i.e. Tech + Pharma) stocks covered by Butterwire in North America outperformed the corresponding MSCI index, and half of those did so by over 61% (see graph below). By contrast, only 23% of the North American Transaction (i.e. Financials + Real Estate) stocks managed to outperform, and of those who did, only half did so by more than 21%. And so a random draw from the universe of Innovation stocks had an expected excess return of +24% while one from the set of Transaction stocks had an expected excess return of -22%. In other words, picking an outperformer from the Tech sector was a lot easier than in Banks...

Looking at the engine performance relative to a random stock selector is a way to adjust for these differences. As illustrated in the graphs below, the engine beats a random selector by a similar margin whether one looks at the high-performing Innovation sector vs. the low-performing Production sector. The engine therefore managed to add value in a consistent manner, extracting fundamental signals that proved relevant regardless of the relative difficulty of finding outperformers.


Persistence of fundamental signals and timeliness of alerts support low portfolio turnover

The graphs below represent the relative performance by regions and sectors of a basket composed of the 235 large cap recession resilient long candidates (interesting + interesting but high octane) generated by the engine as of Jan-17 out of a global coverage of 3,300 stocks at the time. Consistent with the notion that a fundamental investment thesis typically plays out over several quarters, the performance of this “0 turnover” basket beats that of the corresponding Candidate Index analysed in the previous section, which is rebalanced monthly and therefore generate liable to significant turnover.


Of course, it is equally unrealistic to extrapolate aggregate results from a monthly rebalanced index (too much turnover) than it is to do so from a 0% (i.e. too little) turnover basket over 3 years. This is where timely alerts (“Exit?”, “Take Profit?”, “Check Thesis!”) help trigger a review of a stock’s fundamentals to decide whether to change a portfolio position, as illustrated by looking at some of the top and bottom performers from the Jan-17 basket (see list below).

Recession Resilient Candidates

Longfor was the top performer in the basket with 253% excess return over 3 years but heeding the “Take Profit?” advice of the engine from April 2019 would have shaved off a third of the overall performance as the stock continued its upward trajectory throughout 2019 (as a reminder, a “Take Profit?” alert points to a heightened likelihood of extreme future returns, with a slight tilt toward underperformance vs. outperformance). In the case of ANTA Sports, “Take Profit?” alerts started appearing as early as August 2017 and selling the stock back then would have resulted in losing pretty much all the 183% excess return of the past 3 years. On the other hand, “Exit?” alerts were systematically issued for Pandora from Mar-17 until Aug-18 and selling out of the stock at any point during the period would have avoided or mitigated the impact of the stock’s -74% excess return (as reminder, an “Exit?” alert points to a heightened likelihood of material future underperformance). Same for Alliance Data Systems from Jul-17, etc. Portfolio turnover can therefore be better kept in check by complementing one’s own stock-specific knowledge with the timely occurrence of a flag or alert on the stock.