Alpha Capture Systems: Past, Present, and Future Directions

François-Serge LHABITANT*
Professor of Finance, EDHEC Business School

Denis MIRLESSE*
Director, Steadfast Advisory Services Ltd.

Abstract: Alpha Capture Systems (ACS) describe a set of commercial and private systems
that track investment recommendations and their subsequent performance to create an alpha
generating portfolio. This paper presents an overview of the ACS evolution from the first
sell-side to the more recent buy-side systems, from capturing brokers ideas to mutual funds
and even hedge funds positions. We discuss the key features of the most recent ACS and flag
their weaknesses. We also suggest various improvements that we expect to see in future ACS
and which, based on our experience, should lead to enhancing their performance significantly.

1 Introduction
Following a pioneering study by Alfred Cowles (1933), the performance of analysts’ stock
recommendations has been discussed for nearly nine decades with mixed conclusions – see
Hobbs and Sign (2015), Crawford et al. (2017), or Su et al. (2019) for recent reviews. Due to
serious limitations in terms of data availability for buy-side analysts, most studies have focused
on publicly-distributed sell-side analysts’ recommendations.

Sell-side analysts typically work at brokerage firms or banks, and they are the researchers who
have a public face. They follow a list of companies that generally operate in the same industry
or sector, gather information, build detailed financial models, and provide regular research
reports and notes to their firm’s clients, hoping to attract execution flows and increase trading
commissions. Their reports typically include a set of financial estimates, a price target, and a
recommendation usually formulated as “buy”, “sell”, or “hold”, depending on their
expectations of a stock’s future performance. Their investment horizons range from several
weeks for trade ideas related to catalyst events such as news or earnings announcements, to
several quarters or even years for the more fundamental, longer-term investment ideas.

By contrast, buy-side analysts operate outside of the public eye. They typically work for
managers of mutual funds, hedge funds, pension funds or large family offices. They also track
down financial news and company information, they build financial models and ultimately
issue stock recommendations, but exclusively for the benefit of their in-house portfolio

*
The views, thoughts, and opinions expressed in this paper are those of the authors and do not necessarily reflect
the views or policies of the organizations they are, have been, or will be affiliated with. Certain information
herein is based on data obtained from third-party sources believed to be reliable. However, we have not verified
this information, and we make no representations whatsoever as to its accuracy or completeness. The document
is a general communication that has been prepared solely for informational and educational purposes and does
not constitute an offer or a recommendation to buy or sell any particular security or to adopt any specific
investment strategy. Contact: [email protected]

Electronic copy available at: https://ssrn.com/abstract=3873884

managers. Most of the time, their research process is expected to be more reliable / less
conflicted than that of sell-side analysts, as they have no commercial purpose. However, one
should remember that although buy-side analysts conduct proprietary research, buy-side
analysts often have an interdependent relationship with their sell-side colleagues, whose
research they can read. Based on the 2015 Bloomberg Survey “Broker vote: what do you value
most in research”, the buy side views sell-side trade ideas as the most valuable sell-side
research output, more so than ideas from traditional fundamental research. More and more
investors use trade ideas as an integral part of their investment process, and the number of
analysts and brokerage houses issuing them is rising – see Birru et al. (2019). The difficulty,
of course, is to separate the wheat from the chaff, ideally in an automated way.

A series of commercial and private systems have been developed to collect and track the
performance of investment ideas. These are generically called Alpha Capture Systems (ACS).
Some are quite simple and limited to collecting lists of stocks and calculating their
performance. Others require more sophisticated inputs, such as portfolios of recommended
stocks including some weightings, and/or offer advanced tools to rank investment idea
providers and ultimately create an optimized alpha-generating portfolio from their stock
recommendations. Some are exclusively based on sell-side recommendations, while others
explore investment ideas from the buy side, with a universe including not only mutual funds,
but also secretive hedge funds and their regulatory filings.

The purpose of this paper is to explicitly investigate the differences between the major ACS
approaches available in the market, the nature of their inputs, and the limitations that they face
as a result of their choices. To the best of our knowledge, this is the first study that provides
such a comparison. In addition, we discuss various improvements that we expect to see
implemented in the next generation of ACS. Based on our experience and on empirical tests,
these improvements should lead to higher alpha generation.

The structure of the paper is as follows. Section 2 reviews the early days of ACS, from the first
approaches geared at capturing trade ideas voluntarily contributed by the sell side, to the more
recent systems that attempt to extract position-level information from the buy side, mostly
long-only mutual funds, and from insiders. Section 3 describes the most popular current
systems, which have been extended to include hedge fund positions in their approach. Most of
these systems suffer from fundamental weaknesses. Section 4 discusses possible improvements
to ACS, and their key benefits. Section 5 concludes.

2 The early days of Alpha Capture Systems (ACS)

2.1 Extracting alpha from sell-side ideas

ACS came to prominence in 2001 when the hedge fund company Marshall Wace started using
a system to systematically track the performance of trade ideas received from various sell-side
institutions. Its initial goal was to reward these institutions based on the profitability of their
ideas, rather than using other less measurable factors. Marshall Wace quickly realized that these
trade ideas could be optimized to try to outperform the market. It therefore went on to establish
a series of MW TOPS funds1, which relied exclusively on its proprietary ACS. Their success,
both in terms of performance and asset gathering, encouraged other money managers to

1
The acronym TOPS stands for “Trade Optimized Portfolio System”.

Electronic copy available at: https://ssrn.com/abstract=3873884

develop similar systems and funds. Third-party commercial ACS quickly followed – let us
mention YouDevise’s Trade Information Monitor, Thomson Reuters’ StarMine, First
Coverage’s Community, FactSet’s AlphaMetrics, as well as Trade Ideas Limited, a common
platform used by Citigroup, Credit Suisse, Dresdner Kleinwort and Merrill Lynch.

The philosophy of these systems differed in various ways. TOPS was clearly based on polling
as many contributors as possible through a web-based system and selecting a large amount of
their ideas. As an illustration, in 2006, the system had thousands of individuals contributing
tens of thousands of ideas in real time. Each contributor had to manage a virtual portfolio and
allocate some virtual capital to their most compelling ideas – long or short – with a 200-word
rationale for each. Some small sector-focused virtual portfolios had as few as 6 positions, while
larger ones had as many as 25. TOPS sorted and prioritized these ideas by geography, industry
sector, market capitalization and contributor accuracy, amongst other metrics, and used a
constrained multifactor optimization model to filter, evaluate and ultimately weight them in the
various portfolios (each TOPS fund typically held close to one thousand live positions).

By contrast, some of TOPS’ largest and most public rivals used a more selective and/or
concentrated approach. For instance, the GLG Partners Esprit Fund gathered investment ideas
from a select group of 51 discretionarily-selected brokers, who only delivered their best long-
term fundamental ideas to a web-based system. Similarly, the Gartmore Alphagen Acamar
Fund invited about a dozen of their favorite, most-trusted brokers to submit their highest-
conviction ideas, with a cap of four to six recommendations per quarter. The result was a
concentrated portfolio of 60 to 70 equally-weighted positions, with some constraints in terms
of risk and diversification (country and sector limits of 25 percent, rolling stop loss of 10
percent on any single position).

Another important element in these systems was the incentives given to contributors. For
example, Marshall Wace rewarded contributors on a time-lagged basis by allocating
commissions to the most “successful” idea providers (as determined by Marshall Wace),
regardless of whether TOPS funds had selected their ideas. Following selection, actual trade
orders for those ideas were directed to the best execution and service firms, no matter where
the idea came from. Many other systems not only included various inducements to participate,
but even evolved the practice into a competition, where the winners were announced publicly
every quarter. Name and fame…

Over the years, ACS gradually became an accepted and essential practice. More and more
investors started using ACS to analyze sell-side ideas. To prove their value, more and more
sell-side sales professionals started communicating their trade ideas through ACS rather than
usual channels. Not surprisingly, this led to several potential regulatory issues in different
jurisdictions. Let us mention a few:
• How to ensure that brokers would not give preferential treatment (i.e. communicate their
best ideas faster) to more profitable clients such as hedge funds, rather than to traditional
asset managers?
• How to prevent that inputting ideas in an ACS might result in leaking inside information,
or relevant information not generally available to clients?
• Should trade ideas sent to an ACS be regarded as a desk communication or as fundamental
research? If the latter, then idea contributors should be subject to the same disclosure
obligations as research analysts. If the former, then the communication should not include
a long-term target price or an investment rating.

Electronic copy available at: https://ssrn.com/abstract=3873884

• If an idea contributor is remunerated in cash, as opposed to receiving execution flows,
should the contributor be considered as an investment advisor? If so, then the contributor
should also be forced to register as an investment advisor and be subject to the Investment
Advisors Act (1940).

Surprisingly, the blessing for ACS came from regulators. In its Market Watch Newsletter, the
Financial Services Authority (2006) reviewed alpha capture systems (ACS). It gave ACS a
clean bill of health, warned that they could be abused, but also praised the use of audit trails
created by trade ideas in ACS for compliance purposes, and listed a series of good practices.
A few years later, the European Market Abuse Regulations required salespeople and sales
traders to disclose their 12-month stock recommendation histories, similarly to research
analysts, which forced brokers to become more systematic in tracking sales activity –
something very easy to do when using an ACS. Finally, the most recent boost was given with
the European regulators’ rewrite of their financial rulebook, known as MiFID II, which imposes
a clear separation between research payments and trading commissions.

In 2011, the TABB Research Group estimated that there were around a dozen of fund
management groups running similar strategies to TOPS, plus many third-party providers
offering access to commercial alpha capture platforms, including FactSet and TIM group, the
market leaders. By then, some banks had also developed their own proprietary systems – for
example, Deutsche Bank’s Raptor and Morgan Stanley’s TIDE. Several additional providers
have joined the market since. Even academic literature has started to explore the topic, with
the first conclusions being that trade ideas possess predictive value and stock-picking capability
– see for instance Thomas (2011).

2.2 An alternative source of investment ideas: buy-side holdings and insider trades

While they are potentially a great tool to collect investment ideas, ACS rely on the voluntary
contributions of research analysts, brokers, and other market participants. In addition, they are
relatively expensive and complex to set up. An alternative free source of investment ideas can
be found in portfolio holding disclosure (PHD). PHD is where the underlying securities of an
investor, together with their value and portfolio weight, are mandatorily made available to the
public, usually with some lag time.

PHD requirements were initially designed to provide greater transparency in the mutual fund
industry and to allow investors to understand where their money was invested. Prior to May
2004, US-registered investment companies such as mutual funds had to disclose all of their
portfolio holdings to the SEC by filing an N-30D Form semi-annually, within a maximum of
60 days after the end of each quarter, and make that information available to shareholders upon
request. More frequent disclosure using Form N-30B2 was possible on a voluntary basis, but
was not required. From May 2004, the SEC introduced new Forms N-CSR and N-Q to report
all portfolio holdings and effectively increased the disclosure frequency from semi-annually to
quarterly2. As a result, it is possible for an investor to access and observe the content of any
mutual fund portfolio on a quarterly basis, with a maximum 60-day lag.

2
Complete portfolio schedules for the second and fourth fiscal quarters are filed on Form N-CSR (in conjunction
with the filing of annual and semi-annual reports to shareholders), while complete portfolio schedules for the first
and third fiscal quarters are filed on Form N-Q, within 60 days of the end of the quarter.

Electronic copy available at: https://ssrn.com/abstract=3873884

Another important source of information about portfolio holdings derives from the reporting
obligations imposed in Sections 13 and 16 of the US Securities Exchange Act of 1934. These
obligations target a broader universe of investors, namely all holders of public company
securities such as mutual funds, as well as hedge funds, asset managers, and even large family
offices, whether they are registered with the SEC or not, and regardless of where they are
located (US or non-US). A major difference with the reports discussed previously is that such
reports disclose holdings at the management company aggregated portfolio level, rather than
for each individual fund or portfolio.

Forms 13F: In 1975, the US Congress enacted Section 13(f) of the Securities Exchange Act
to increase the public availability of information regarding the securities holdings of large US
equity investors. It requires investors having investment discretion on portfolios of over $100
million of Section 13(f) securities3 to disclose their long holdings within 45 days of the end of
each calendar quarter. The disclosure must be made electronically on Form 13F through the
SEC’s Electronic Database Gathering and Retrieval (EDGAR) system. Forms 13F include the
CUSIP number, number of shares, and the total market value (based on closing prices) of the
long holdings as of the last business day of the quarter.

Once a Form 13F has been filed, the SEC is required to make it publicly available. However,
investors may request confidential treatment for some of their holdings for public interest
reasons or for their own interest protection. In such a case, they file their Form 13F as usual,
but exclude their confidential holdings and indicate (where appropriate) that some positions
have been omitted. Separately, they file a paper request for confidential treatment with the
SEC, which includes the list of confidential positions, the reasons for requesting their
nondisclosure and the length of time for which confidential treatment of these positions is
requested. If accepted by the SEC, the confidential holdings will eventually be disclosed to the
public through a Form 13F add-new-holdings amendment reported on EDGAR, which must be
filed within six days of the end of the confidential treatment period. If rejected by the SEC, a
new Form 13F amendment that reveals the confidential holdings must be filed within six days
of the rejection. In practice, the SEC rarely grants confidential treatment, apart for large merger
arbitrage positions, or for ongoing programs of acquisition or disposition – see Pekarek (2007),
Robertson (2008) and Martin (2009) for a discussion. As such, confidential holdings represent
only a small portion of most hedge funds’ portfolios.

Schedule 13D: Under Sections 13(d) of the Securities Exchange Act, any investor or group of
persons acting together that acquires beneficial ownership of more than 5% of a class of a
company’s equity securities must file a Schedule 13D electronically via the SEC’s EDGAR
system within 10 days after the acquisition.

Schedule 13D includes, amongst other things, the identity of the acquiror (or if applicable, each
member of the group), the source and amount of funds used to acquire the securities, the total
number of securities held, the purpose of the acquisition, including any plans or proposals for
future purchases or sales of target stock or for any changes in the target management or board
of directors or any major corporate transaction affecting control of the target, such as a tender
offer or business combination. Moreover, any material changes to the information in the initial
filing (including any acquisition or disposition of beneficial ownership of securities equal to
1% or more of the class) must be promptly reported in an amendment.

3
The SEC publishes a list of Section 13(f) required securities on a quarterly basis, called the “SEC Official List”.
It includes mostly equity securities traded on a US securities exchange (e.g., NYSE, AMEX, NASDAQ), shares
of close-ended investment companies, shares of exchange-traded funds, certain other securities such as ADRs,
along with specific equity put and call options, warrants, and convertible debt securities.

Electronic copy available at: https://ssrn.com/abstract=3873884

Schedule 13G: Under some conditions, an investor may ask for an exemption and file a short‐
form statement on Schedule 13G in lieu of a Schedule 13D. This applies, in particular, if the
investment is passive, that is, the investor has no intention to exert control or influence the
management of the company, and the ownership is below 20%.

Note that under Schedule 13G requirements, investors must also include beneficial ownership
of options and rights in their share count when the right to acquire is within 60 days of the end
of the quarter.

Forms 13H: Large traders must report their transactions to the SEC. Essentially, a large trader
is defined as any person who directly or indirectly exercises investment discretion over
transactions in listed US equity securities and listed options in an aggregate amount that equals
or exceeds 2 million shares or $20 million in value in a calendar day, or 20 million shares or
$200 million in value in a calendar month. Unlike the reports discussed above, Forms 13H are
confidential and remain exempt from disclosure under the Freedom of Information Act.

Section 16 Forms: Section 16 of the Securities Exchange Act sets the reporting obligations
applicable to persons considered to be “insiders” of public companies. Insiders include officers,
directors, and beneficial owners of more than 10% of any class of equity security that is
registered pursuant to Section 12 of the Exchange Act.

Insiders must file three types of reports with the SEC, all of which are publicly available:
• When becoming an insider, a Form 3 must be filed within 10 calendar days. Form 3 lists
all the company’s equity securities (including derivatives) that were owned by the insider
immediately prior to becoming an insider.
• In most cases, when an insider executes a transaction in the company’s equity securities
(including derivatives), a Form 4 must be filed by the close of the second business day
following the transaction. Form 4 discloses the insider’s various transactions in company
securities, including the amount purchased or sold and the price per share.
• Transactions made by insiders that were not reported when they occurred, for instance
because of an exemption or a failure to report, must be reported on a Form 5 within 45
days after the end of the company’s fiscal year.

Relatively few people are subject to section 16, but these are generally prosperous, influential,
and informed people. Whether they actually possess material, nonpublic information when they
trade is completely irrelevant. They must file what is required, or face SEC fines and
disgorgement of profits.

2.3 Extracting alpha from mutual funds

PHD is useful from a client information perspective and can potentially be used to free ride on
someone else’s research by running copycat portfolios. In its simplest version, an investor
could easily collect the list of holdings of successful fund managers from regulatory filings
(say 13F Forms, for instance, published with up to 45 days delay after each quarter end) and
replicate them in his portfolio, without paying any fee to the managers.

Would such a simple approach work? A large body of academic literature answers positively.
As an illustration, Martin and Puthenpurackal (2008) show that a hypothetical portfolio

Electronic copy available at: https://ssrn.com/abstract=3873884

invested in Berkshire Hathaway’s holdings a month after they were publicly disclosed would
have outperformed the S&P 500 by an annual average of 10.75% from 1976 to 2006. Similarly,
Hugues et al. (2010) evidence that one could achieve similar investment results to Warren
Buffett by simply following his trades as implied by Berkshire Hathaway’s quarterly 13F
filings. Frank et al. (2004) and Verbeek and Wang (2013) show that copycat funds have earned
returns that are statistically indistinguishable, and possibly higher after fees, compared to
original funds.

Copycatting mutual funds is relatively easy, particularly since their regulatory filings are made
electronically. However, becoming a smart imitator comes with two main issues. The first issue
is technical; how does one go about identifying the managers to replicate? Most people will
simply try to replicate managers that have achieved great performance in the past, but there is
no guarantee that these will remain winners. The second issue is more philosophical;
replicating an active manager only makes sense if that manager adds alpha over their
benchmark. Since most mutual fund portfolios are highly constrained in terms of deviating
from their benchmarks, their manager’s ability to take large active bets is limited. Most of the
time, stock weightings are primarily driven by the portfolio’s benchmarks rather than by the
manager’s conviction. One should therefore wonder whether mutual funds PHD are the best
place to extract great investment ideas.

Hedge fund managers may be a better pond to look for investment ideas, as their portfolios
should represent their views rather than restrictive investment guidelines4. Since they are
subject to similar PHD as mutual funds (with the exception of Forms N-CSR and N-Q), it is
tempting to seek to capture some of their alpha by buying some of the same stocks they own.
Recent academic research seems to support this idea. For instance, Bae et al (2011) analyze
hedge fund holdings and find that hedge funds do have superior forecasting ability (almost four
times as great) when compared to other institutional investors. They also uncover a positive
link between the level of hedge fund ownership in a stock and the stock’s subsequent
performance. Angelini et al. (2019) and Anton et al. (2021) show that the best ideas of hedge
funds deliver economically meaningful and statistically significant returns that outperform the
S&P 500. Extracting alpha from hedge fund holdings could therefore be an interesting path to
explore.

It should be noted that an alternative approach to extracting alpha from hedge funds would consist
in replicating their “betas”. Several ETFs use quantitative approaches such as rolling linear
regressions or Kalman filters to estimate the risk exposures of hedge funds and replicate them
dynamically and mechanically. Let us mention the IQ Hedge Multi-Strategy Tracker ETF (QAI),
the ProShares Hedge Replication ETF (HDG) and the iM DBi Hedge Strategy ETF (DBEH).
These beta-driven clones tend to be more liquid and cheaper than hedge funds, but their results
have generally disappointed from a performance perspective – see Fischer et al (2016). We will
therefore not discuss them any further and rather focus on position-level replication.

4
A lot of academic and practitioners research has evidenced the lack of alpha in hedge funds, particularly in recent
years. In our opinion, this is not unexpected as most researchers have focused on the average hedge fund – for
instance looking at the performance of a hedge fund index. Who would be dumb enough to allocate money to a
basket of a few thousand of hedge funds, each charging 2 and 20 on their individual portfolio, and with no netting
of fees? On average, hedge fund managers are expected to be average, minus their fees. In fact, there may be
more unskilled managers in hedge funds than in mutual funds due to the lower regulatory requirements and
scrutiny. So, looking for hedge fund alpha in the average hedge fund is a waste of time. The real question is
whether a small minority of hedge funds generate alpha, and whether these funds can be identified.

Electronic copy available at: https://ssrn.com/abstract=3873884

3 The present: ACS and hedge fund replicators
The first two hedge fund alpha replicators are the Global X Guru ETF (GURU) and the
AlphaClone Alternative Alpha ETF (ALFA). Both were launched in mid-2012 and rely on
similar proprietary rule-based methodologies using exclusively 13F Filings from hedge funds.

3.1 GURU

GURU tracks the Solactive Guru Index, which is calculated and published by Solactive AG
and is built as follows:
Idea contributors universe: The universe of potential idea contributors is a list of several
thousands of hedge fund names compiled from a number of sources including Morningstar,
Bloomberg, Bloomberg magazine and Barron’s Top 100 Hedge Funds as published on a
regular basis.
Idea contributors selection: The pool of idea contributors is selected once a year, annually
at the end of January. It includes all the funds that fulfill the following conditions according
to their most recent 13F filing: (1) at least $500 million of assets under management; (2)
their largest holding must represent at least 4.8% of their portfolio; (3) their year-on-year
portfolio turnover must be below 50%.
Holdings universe: The holdings universe is determined on a quarterly basis, 7 business
days after the official 13F filing date. It includes the holdings of each selected hedge fund
that fulfill the following conditions: (1) listed on a US regulated stock exchange; (2) have a
market capitalization of at least $100 million; (2) have an average daily trading volume in
the last three months of at least $10 million; (4) have an average monthly trading volume of
at least 75,000 shares in each of the last six months; (5) have an allocation of at least 4.8%
of the total hedge fund portfolio market value.
Holdings selection: Each hedge fund’s largest holding is selected for the index. In addition,
each hedge fund’s second- and third-largest holdings that are in the current GURU portfolio
(which means they were selected in the previous quarter) remain selected.
Portfolio construction: On a quarterly basis, 7 business days after the official 13F filing
date, the index is reallocated equally to the holdings selected in the previous step. Specific
rules deal with extraordinary events and corporate actions between index rebalancing dates
– see Solactive (2016).

3.2 ALFA

ALFA tracks the AlphaClone Hedge Fund Masters Index, which is calculated and published
by AlphaClone Inc. It is built as follows:
Idea contributors universe: The universe of potential idea contributors consists of 500
funds selected from the Pertrac database, which tracks over 6,000 hedge fund managers.
Selection criteria include fund size, length of 13F filing history, and their investment
approach.
Idea contributors selection: The pool of idea contributors is made of 10 hedge fund
managers selected semi-annually (February and August) by a scoring process that aims at
finding managers that outperformed and did it often. The scoring process runs multiple

Electronic copy available at: https://ssrn.com/abstract=3873884

 simulations (clones) that simulate investing in a manager’s disclosed holdings (e.g. their
largest 1, 3, 5, 10, 20 holdings, etc.). Each manager is then assigned a profile that is the
average of their clone monthly performances. All profiles are analyzed to determine by how
much the manager outperformed the total market (the average batter) and how often they
outperformed. Each fund receives a score based on the results of that analysis, and the 10
funds with the highest score are retained as idea contributors.
Holdings selection: The 5 largest holdings from each selected manager’s based on their
13F-HR filing.
Portfolio construction: On a quarterly basis, index constituents are selected and equally
weighted.

It should be noted that ALFA has changed its investment objective since its inception. Prior to
December 27, 2017, it was tracking the performance of the AlphaClone Hedge Fund Downside
Hedged Index, which selected and weighted its constituents differently, and more importantly,
allowed using short S&P 500 positions in an amount equal to the value of the fund’s longs
positions for risk management purposes. This occurred in October and November 2015 and
again in January, February and March 2016.

3.3 Other copycats

Following the initial marketing success of the idea of imitating the smartest investors, a large
amount of ETFs using variations of their methodology were launched, sometimes with a
particular focus on specific hedge fund strategies or segments of the equity market. Let us
mention, in particular, the Global X Guru Activist Index ETF (ACTX), the Global X Guru
Small Cap Index ETF (GURX), the Global X Guru International Index ETF (GURI), and the
AlphaClone International ETF (ALFI). Most of these products failed to attract assets and have
been shut down.

Another line of rule-based copycats is based on filings from insiders such as corporate officers,
directors, and large stockholders. For example, the Invesco Insider Sentiment ETF (NFO),
which tracks an equal-weighted index of the 100 US stocks selected by publicly available
buying trends of corporate insiders, price momentum, and trailing 12-month volatility.
The Direxion All Cap Insider Sentiment Shares (KNOW) tracks an equal-weighted index of
the 100 US stocks selected on the following basis: (1) an upward analyst rating of the earnings
per share (“EPS”) estimate for the current fiscal year; (2) the percentage change in the EPS
estimate for the current fiscal year as of the recalculation date and prior month end; (3) the
percentage of insider holders having increased their holdings in the past month as of the
recalculation date; and (4) the absolute change in the insider holdings measured as a percentage
as of the recalculation date and prior month end.

A slightly different approach used to be offered by the Direxion iBillionaire Index ETF (IBLN).
IBLN tracked the iBillionaire Index, which is calculated and published by iBillionaire Inc.
It was built as follows:
Idea contributors universe: The universe of potential idea contributors consists of
investors with a personal net worth of at least $1 billion that is calculated and verified by
industry publications, and whose primary source of wealth is financial markets and
investments. These investors must file a Form 13F with the SEC.

Electronic copy available at: https://ssrn.com/abstract=3873884

 Idea contributors selection: Based on 13F Filings, IBLN selects 10 billionaires based on
several criteria and requirements, e.g. (1) their 13F implied portfolio must be valued at
$1 billion or higher, (2) it must have at least 10 securities; (3) it must have an annual turnover
below 50%; and (4) the selected billionaires’ 13F implied portfolio must have an 8-year
return that places them amongst the top 15 billionaires.
Holdings selection: the top 30 large-cap US equities where the 10 billionaires have the
highest combined allocation.
Portfolio construction: On a quarterly basis, index constituents are selected and equally
weighted.

The resulting portfolio had a large-cap bias by design and was more concentrated than GURU
or ALFA. Unfortunately, despite its over-the-top name, IBLN failed to attract assets and closed
on 6 April 2018.

3.4 The Goldman Sachs Hedge Fund VIP Index

The real game changer in hedge fund ACS was the creation of the Goldman Sachs Hedge
Industry VIP ETF (GVIP) on 1 November 2016. GVIP tracks the Goldman Sachs Hedge Fund
VIP Index, which is calculated and published by Solactive AG. It is built as follows:
Idea contributors universe: The universe of potential idea contributors consists of all the
managers that file a Form 13F and that are classified as “hedge funds” in their Investment
Adviser Registration Depository (IARD) filings with the SEC.
Idea contributors selection: Based on their most recent Form 13F, all the managers that
have at least $10 million of disclosed assets, and hold no fewer than 10 and no more than
200 distinct US equity positions, are selected as idea contributors.
Holdings selection: From each idea contributor’s Form 13F, the largest 10 holdings by
dollar value is extracted. The 50 stocks that appear most frequently in the resulting screened
list of the top 10 holdings of the various idea contributors are selected as index constituents.
In the case of a tie in ranking, stocks will be included based on an alphabetical sort by ticker.
Note that stocks that Goldman Sachs Asset Management is restricted from holding for legal
or regulatory reasons cannot be included in the index.

With an annualized return of 15.1% since 2001 (pro forma until the end of October 2016), the
GS VIP index would have beaten the S&P 500 in 62% of the quarters and outperformed it by
approximately 80%. This seems to corroborate the idea that the GS VIP index represents smart
money, although some of its outperformance may be linked to a mega-cap bias (it has a median
market capitalization of $48 billion versus $13 billion for the S&P500).

10

Electronic copy available at: https://ssrn.com/abstract=3873884

4 The future: some thoughts on improving ACS
In this section, we suggest and discuss several improvements that we expect to see implemented
in future ACS.

Broaden the universe of idea contributors beyond hedge fund managers

Many ACS limit their universe of idea contributors to one specific pool, for example hedge
fund managers listed in a given commercial database, and impose some key requirements such
as minimum assets under management, minimum past performance, a given pre-specified
strategy, etc. While convenient, these choices are highly restrictive and have important
unintended consequences.
First, as illustrated by Lhabitant (2007), there is very little overlap between hedge fund
databases. Consequently, ACS based on different databases will start with very different
universes of idea contributors.
Second, following the Dodd-Frank Act’s tougher reporting requirements, in particular Form
PF5, several prominent hedge fund managers have shed outside investors, shut down their
funds and effectively restructured their businesses as a family office6. Famous examples
include George Soros, Leon Cooperman, David Tepper, Bob Karr, Louis Bacon, Carl Icahn,
Stanley Druckenmiller, just to name a few. These talented stock pickers no longer run funds,
but are now picking stocks for their own benefit. Due to their size, their portfolios continue
to be subject to regulatory filings, including 13F, 13D and 13G. Restricting idea contributors
to just hedge funds excludes these talented stock pickers from the universe, which is not a
great idea.
Third, there are still talented managers that run long-only portfolios. Warren Buffet is likely
the best-known example, even though his value bias has been painful in some years.

We recommend that ACS extend their idea contributor universe to include anyone that
discloses potential great investment ideas on a regular basis. The technical nature of the
investor, and even whether it is a buy-side or a sell-side investor, should be irrelevant.

Use as many data sources for position data and update them as frequently as possible

As previously discussed, many ACS exclusively rely on Forms 13F collected quarterly from
the SEC website. As a result, they only update their knowledge about investment ideas once a
quarter, usually with a 45-day lag. While convenient, this choice is highly myopic.

First, there are multiple alternative data sources, including some from vendors that provide
holdings information at different dates than the SEC. For example, Schwarz and Potter
(2016) compare SEC filings with Thomson Financial Mutual Fund Holdings data7. Out of
77,555 unique portfolios across both sources, 32% are in Thomson but not in SEC filings,

5
Form PF requires fund advisers to disclose their strategies, products, performance, changes in performance,
financing information, risks metrics, counterparties and credit exposure, percentage of assets traded using
algorithms, and the percentage of equity and debt, amongst others.
6
Under Dodd-Frank, family offices benefit from a special exemption from regulation under the Advisers Act,
meaning that the requirements for disclosure and investor protection imposed on family offices are far less
onerous than for hedge funds.
7
The Thomson Financial Mutual Fund Holdings database was previously known as the CDA Spectrum
database, or the CDA Investment Technologies data.

11

Electronic copy available at: https://ssrn.com/abstract=3873884

 and 20% are in SEC filings but not in Thomson. When also considering the CRSP Mutual
Fund Database, only 39% of portfolios overlap in all three data sources. While it is
surprising to see funds voluntarily disclosing additional non-SEC portfolios to data vendors,
this information may be extremely useful for ACS as it allows them to update their inputs
intra quarter.
Second, Forms 13F may be amended after the initial filing, either to correct errors or
omissions, or to include certain reportable securities that were not listed in the original
public filing for confidentiality reasons8. Managers making amendments must re-submit
their entire (corrected) filing within six business days of either the SEC notification of denial
of confidential treatment, or the expiration of the confidential treatment period. They must
also include a cover page indicating that the new filing is an amendment and may add brief
explanatory remarks. In some instances, multiple amendments are required. For instance, in
the case of an error correction (e.g., number of shares or fair market value misstated) plus
the inclusion of new holdings, the manager must first file a Form 13F stating the old holdings
and ticking the box designating the amendment as a restatement, and then a second Form
13F showing the new holdings and ticking the box designating the amendment as adding
entries for new holdings. If ACS ignore these intra-quarter amendments, they may miss
important information, or even worse, use inaccurate information about the real content of
the underlying managers’ portfolios. Both Aragon, Hertzel, and Shi (2013) and Agarwal,
Jiang, Tang, and Yang (2013) evidence that hidden positions tend to deliver abnormally
positive returns.
Third, going beyond 13F and using other regulatory filings with the SEC also allows for
more frequent updates. Very few ACS seem to use Forms 13D and 13G, which are typically
filed intra quarter, even though their content may be highly significant in terms of underlying
portfolios. In particular, if a Form 13D or a Form 13G is filed after quarter end but before
Form 13F is available, an ACS using Form 13F would look at an outdated portfolio. Ignoring
the information available in such forms is likely to lead to an unnecessary lag of several
months in terms of updating portfolio holdings.

We recommend that ACS use as many data sources as possible and update the information they
use as inputs as often as possible. In particular, for ACS scraping the SEC website for Forms
13F, they should do it on a continuous basis to capture any late or amended filings.

Always validate and correct the data

A major issue with regulatory filings is the poor quality of the underlying data. As evidenced
by Anderson and Brockman (2018), errors in 13F files are not a rarity, they are the norm.
Examples include number of shares inconsistent with other public filings, inaccurate closing
prices for stocks, discrepancies between dollar values and percentages, CUSIP that do not
match the text stated security, etc. The problem has been explicitly acknowledged by the SEC
Office of Inspector General (2010), who admitted that “no SEC division or office monitors the
Form 13F filings for accuracy and completeness”. The main issue is likely that “no SEC
division or office has been delegated authority to review and analyze the 13F reports, and no
division or office considers this task as falling under its official responsibility.” No action
seems to have been taken, apart from the SEC adding a caveat to all 13F documents stating that
“the reader should not assume that information is accurate and complete”. Unfortunately, errors
in portfolio holdings are also common in databases commercialized by reputable third-party

8
Such holdings must be disclosed through a revised Form 13F within six business days of either the SEC
notification of denial of the confidential treatment request, or the expiration of confidential treatment.

12

Electronic copy available at: https://ssrn.com/abstract=3873884

providers. As an example, Schwarz and Potter (2016) document that prior to the fourth quarter
of 2007, one in five CRSP Mutual Fund Database portfolios had 25% or more of their positions
reported inaccurately.

We recommend that ACS challenge the accuracy of any information they use, even if it is
provided by the SEC or collected, organized, and displayed by reputable third-party data
providers. In our experience, a considerable amount of work is needed before such data can be
validated, corrected, and ultimately trusted.

Identify talented idea contributors, not talented fund managers

Having a large universe of idea contributors is conceptually great, but it often creates more
noise than information. On average, managers are average, and more than 80% of them have
failed to beat the market over the past 10 years. Therefore, what an ACS needs is not a large
amount of contributors but rather a small pool of talented ones.

One may think that, in essence, selecting skilled idea contributors is not very different from
selecting skilled fund managers. While it is true that both activities share the same goal – trying
to avoid false positives, selecting skilled idea contributors presents several difficulties. To
illustrate the problem, let us assume that 30% of active managers have skills and let us say we
can differentiate between skillful and unskillful ones with a 65% success rate. As such, if we
select one manager from the universe, there are four possible outcomes:
• the manager has skill, which we correctly identify (19.5% chance); this is the Graal,
and what ACS are trying to do.
• the manager has skill, which we mistake for no skill (10.5% chance); this is a missed
opportunity, which may be corrected in the future.
• the manager has no skill which we correctly identify (45.5% chance); this is fine, but it
does not create value for any long-only ACS.
• the manager has no skill, which we mistake for skill (24.5% chance); this is the false
positive situation, which may lead to a disaster.

These percentages implicitly assume that skilled managers always have great investment ideas,
while unskilled ones always pick dogs. Reducing the percentage of false positives can be
achieved by raising our success rate and/or building a universe made of more skilled managers.
There is no other recipe for success here, and it should be obvious that selecting managers
based solely on the size of their assets is unlikely to be the right approach. In addition, the
shorter the time horizon, the more randomness will be the dominant influence in outcomes. In
summary, making the right pick is not easy and straightforward.

Further, an important fundamental difference between selecting skilled idea contributors and
selecting skilled fund managers is the nature of what should be analyzed. For an ACS, what
matters is not the performance of a flagship vehicle listed in a database, but the future
performance that can be extracted from investment ideas as and when disclosed by a manager.
As an illustration, a great hedge fund manager with fantastic shorting skills may end up being
a poor idea contributor, because most of his alpha would come from his short book, on which
there is no transparency, rather than his long book. Similarly, a great trader is likely to be a
poor idea contributor because of the high turnover of his investment ideas combined with a lag
imposed by disclosure timing. ACS should therefore focus on great investors rather than great
traders. Clearly, being skilled does not always guarantee the delivery of great investment ideas.

13

Electronic copy available at: https://ssrn.com/abstract=3873884

Focus on highest conviction ideas

In theory, portfolio managers should leverage their stock-picking skills by concentrating capital
in their highest conviction ideas, as these are expected to have a higher return. Unfortunately,
in practice, managers often end up diluting their portfolio by adding lower conviction positions.
In many instances, this is encouraged by institutional investors who want to see some volatility
management relative to a benchmark, or who perceive portfolios with more holdings as better
diversified, less volatile and providing better downside protection. It is true that, on average,
additional portfolio diversification reduces volatility and the risk of very poor returns.
However, what is rarely mentioned is that, when applied by talented active managers,
diversification also reduces the risk of very high returns. Simply stated, concentrated portfolios
are exposed to significant stock-specific (“idiosyncratic”) risk, layered on top of the equity
market risk (“systematic”) that all stocks contain. On average and across the market,
idiosyncratic risk is not rewarded, but when a given manager is talented, the idiosyncratic risk
of his portfolio is extremely valuable because it has a high positive expected return (“alpha”).
Diluting it by adding lower conviction positions is therefore likely to result in performance
deterioration.

Academic research has studied the performance of high conviction positions – see for instance
Yeung et al. (2012), Hight (2019), Panchekha (2019), Lazard (2020) or Anton et al. (2021). All
these authors use position size as a proxy for conviction. Their conclusion is that high
conviction positions tend to sustainably outperform the diversified portfolios from which they
were derived, and that this outperformance exists across benchmarks, risk models, and high
conviction definitions. To be successful, ACS should therefore focus not only on the best
managers, but also on managers’ highest conviction positions rather than all their positions.

In our opinion, ACS should in fact go one step further to identify for each manager the optimal
number of high conviction positions. When using only public information, the simplest
approach consists in approximating conviction with active position sizing in the portfolio9.
By creating clones of the manager’s top N active positions, with N varying say from 1 to 10,
and comparing their historical performance, one can get useful insights on the impact of the
high conviction positions, and figure out whether there has been an optimal N, at least
historically. It also allows to spot managers who are good at picking stocks but not at weighting
them in their portfolio. Position level information for such managers is useful, but weighting
should obviously be ignored. Of course, all this information may be changing over time and
needs to be dynamically re-assessed and monitored. In fact, this optimal number should not be
the defining characteristic of the investment strategy, but should instead be the outcome of
what opportunities a manager sees in the market.

Identify idea contributor skills and weight their ideas accordingly

A clear issue with many ACS is that once an idea contributor has been selected, all their
investment ideas are treated equally, with no consideration for their skillset. In our opinion,
investment ideas should be ranked based on their expected quality. In practice, the latter can
only be determined by analyzing dynamically the investment idea of each contributor to
understand where their skills and weaknesses are. As an illustration, an investment idea in a
sector where a manager has proven stock-picking skills should be considered of a higher quality
than if it comes from a sector where the manager has historically failed to deliver alpha or has

9
For absolute return portfolios, the active position size equals the weight. For a benchmarked portfolio, it is the
difference between the weight of a position in the portfolio minus its weight in the benchmark.

14

Electronic copy available at: https://ssrn.com/abstract=3873884

never allocated previously. As a result, ACS need to move away from producing a binary
outcome for idea selection (included/not included), and rather produce a continuous measure
that reflects the estimated quality of each idea.

Similarly, some managers are very talented at portfolio construction and make a large
allocation to their best performers; others are just average, or even bad, and should abstain from
being involved in any position sizing activity. This can easily be analyzed by tracking the
cumulative performance of each of the ranked positions for a given portfolio over time. In an
ideal portfolio, we would like to see the performance of the largest position being always higher
than that of the second largest position, the performance of the second largest position being
always higher than that of the third largest position, etc. In reality, things are not as ideal, but
one can rapidly identify managers who are talented at sizing positions and managers who are
not. In the case of managers talented at position-sizing, the weighting of stocks in the portfolio
should be considered as an additional relevant information to assess the quality of the
investment idea. In the opposite case, all the ideas should be considered as being of equal
quality, because the portfolio weightings are just noise.

Select standalone investment ideas

Another important element for an ACS is whether an investment idea is standalone or only
makes sense as part of a more complex set of positions. Consider for example the case of a
stock that is up for a takeover offer. An arbitrageur who believes in the success of the upcoming
takeover will go long the target and short the acquirer and may size both positions large in his
portfolio due to their expected convergence. If the ACS only collects 13F filings, the long
position will likely be wrongly identified as a high conviction position, while the manager may
have no positive view on that stock. Other examples are pair trades, which include companies
listed on multiple exchanges, preferred versus common stocks, or liquidation situations (e.g.
Altaba versus Alibaba). Such pairs are often attributed a large allocation due to offsetting long
and short positions. More importantly, they do not reflect a vote of confidence in the long
position. We therefore recommend that ACS screen high conviction positions versus a list of
corporate actions and exclude the ones that are involved in such events.

Identify leaders, not followers, and measure their ideas time decay

When analyzing a manager’s positions, an ACS should ascertain whether new high conviction
positions are revealing new information, or if they are simply following a consensus trade. The
former is a great source of input for an ACS, while the latter just implies increasing exposure
to crowded trades. Crowded trades are not necessarily bad, but they often carry more potential
downside and tail risk – see Brown et al. (2019). Before entering in a crowded trade, ACS
should try to understand the rationale behind the crowdedness and more importantly, avoid
entering them too late.

There are situations where several managers end up with the same investment idea at the same
time due to the same exogenous signal, such as index inclusions or exclusions, earnings
announcements, analyst recommendations, takeover news, or post-bankruptcy restructuring,
just to name a few. This is fine. There is also a number of situations where managers are just
… copycatting. For instance, a fund manager might prefer to rely on his own analysis to select
stocks where he has an informational advantage, while imitating others’ trades in stocks where
he is at a disadvantage. Not surprisingly, the imitated managers are often leaders in their field,
so that their high conviction ideas become crowded once they are revealed. In such cases, the
ACS should try to follow the leader rather than the copycats. In addition, all copycats are not

15

Electronic copy available at: https://ssrn.com/abstract=3873884

all created equal. For instance, Lesmond and Stein (2020) provide ample evidence that
imitation is pervasive in the mutual fund industry, and that only skilled managers who
outperform with their own stock picks are the ones who can select the best funds and stock
trades to emulate.

An important element for assessing the quality of a manager ideas is their alpha time decay.
The intuition behind alpha decay is best illustrated in the context of a manager having superior
information about a mispriced stock. The manager decides to buy this stock and makes it a
large position in his portfolio because he expects it to generate a large and significant alpha.
What really matters for an ACS is the timeframe over which this alpha is expected. If the
manager information is based on next quarter earnings, then due to the reporting lag, the
superior information may be public when the manager position is disclosed, and most of the
alpha may already have been captured. In some instances, the manager may even have exited
from this (just) reported position. If the manager information is based on a more fundamental
and longer-term view, then one should expect a slower market price response to new
information, so that there is still alpha to be extracted once the idea is disclosed. Talented
leaders with slow alpha decay are obviously the best idea generators for ACS.

Portfolio construction

Once an ACS has selected a short list of investment ideas, the last remaining question is how
to build a portfolio from these ideas. The simplest approach in the absence of any benchmark
is the equally weighted portfolio. It is not exposed to estimation errors, it maximizes the
Herfindahl index and often dominates many others in terms of risk-adjusted return and turnover
– see Lhabitant (2017). If investment ideas come from a wide variety of contributors, then the
resulting portfolio should be relatively diversified. Alternative portfolio construction
approaches may be used, but they require forecasting a large amount of data, and are subject
to estimation risk.

A commonly heard argument is that positions with deeper conviction, as gauged by the size of
the position relative to the manager’s overall 13F portfolio size, should be given higher
percentage allocation and vice-versa. We disagree with this view; for us, a deeper conviction
should increase the likelihood of the stock being selected, but not necessarily influence its
weighting in the final portfolio.

5 Conclusions
On average, stocks are average, and only a few of them are really valuable/can generate a
performance superior to that of the index. As an illustration, Bessembinder (2018) analyzed
the performance of all US stocks in the Center for Research in Security Prices (CRSP) database
from 1926 to 2016 and found that “most common stocks do not outperform Treasury bills over
their lives" and “slightly more than 4% of the firms contained in the CRSP database collectively
account for all of the net wealth creation in the US stock market since 1926”. On average, fund
managers are also average, minus their fees. In its 2019 SPIVA Scorecard, S&P report that
89% of large-cap mutual funds underperformed the S&P 500 over the past decade. The
situation is not better with mid cap (84%) or small cap mutual funds (89%) versus their
respective benchmarks. Finding that there is no alpha on average should not be a big surprise
since alpha generation is a zero-sum game. Fortunately, the absence of alpha on average does
not mean there is no alpha anywhere. Some stocks are very valuable, and some investors are

16

Electronic copy available at: https://ssrn.com/abstract=3873884

truly talented stock pickers. This is where ACS step in, with the key idea that a collective brain
of successful stock pickers should be able to deliver a significant and steady alpha.

In this paper, we have reviewed and discussed several ACS and their approaches to build such
a collective brain. Some of them are simple, not to say simplistic. Mechanically riding the
coattails of past winning investors and cloning their top positions with a lag does not guarantee
superior returns. One needs to think cautiously about each stage of the process and use all the
available information to complement the partial insights about idea contributors and their top
positions. The recommendations that we have made go in this direction. Their implementation
allows the creation of long-only alpha-generating strategies that not only beat the market in the
long run, but also beat their peers – see the Appendix for a discussion.

From there on, there is plenty of room for further research. For instance, one should look at the
alpha generated and understand what drove it over time, e.g. change in factor exposures, sector
or characteristics timing, etc. Another interesting direction would be the creation of long/short
or even equity market neutral alpha-generating strategies from position data. This could be
done using short/shorting? ETFs or baskets as hedges, but also by using data on large short
positions where publicly disclosed (for instance in Europe), or even by shorting some of the
positions that appear the least in the portfolios of the selected and supposedly talented stock
pickers. At the extreme, one could also consider shorting the high conviction long positions of
a pool of consistently non-talented investors. After all, shorting negative alpha may be a
sustainable path to generating positive performance.

6 References
• Agarwal V., Jiang W., Tang Y., and Yang B. (2013), “Uncovering hedge fund skill from the
portfolio holdings they hide”, Journal of Finance, vol. 68, pp. 739-783.
• Anderson A.M. and Brockman P. (2018), “An examination of 13F filings”, Journal of
Financial Research, vol. 41 (3), pp. 295-324.
• Anton M., Cohen R.B. and Polk Ch. (2021), “Best ideas”, working paper, Harvard Business
School, available at SSRN: https://ssrn.com/abstract=1364827
• Aragon G., Hertzel M., and Shi Z. (2013), “Why do hedge funds avoid disclosure? Evidence
from confidential 13F filings”, Journal of Financial and Quantitative Analysis, vol. 48, pp.
1499-1518.
• Bessembinder H. (2018), “Do stocks outperform treasury bills?”, Journal of Financial
Economics, vol. 129 (3), pp. 440-457.
• Birru J., Sinan G., Liu X. and Stulz R.M. (2019), “Are analyst trade ideas valuable?”, Fisher
College of Business working paper No. 2019-03-015.
• Brown G., Howard Ph. And Lundblad Ch. (2019), “Crowded trades and tail risk”, working
paper, University of North Carolina (UNC) at Chapel Hill, available at
https://ssrn.com/abstract=3326802
• Brown S. and Schwarz C. (2011), “The impact of mandatory hedge fund portfolio disclosure”,
working paper, New York University.
• Crawford S., Gray S., Johnson B.R. and Price R.A. (2018), “What motivates buy-side analysts
to share recommendations online?”, Management Science, vol. 64 (6), pp. 2473-2972.
• Cowles A. (1933), “Can Stock Market Forecasters Forecast?”, Econometrica, vol. 1 (3), pp.
309-324.

17

Electronic copy available at: https://ssrn.com/abstract=3873884

• Frank M.M., Poterba J.M., Shackelford D.A. and Shoven J.B. (2004), “Copycat funds:
information disclosure regulation and the returns to active management in the mutual fund
industry”, The Journal of Law and Economics, vol. 47 (2), pp. 515-541.
• Financial Services Authority (2006), “Markets division: newsletter on market conduct and
transaction reporting issues”, Market Watch, issue no 17, 25 September, available on
https://webarchive.nationalarchives.gov.uk/20081112215652/http://www.fsa.gov.uk/pubs/ne
wsletters/mw_newsletter17.pdf
• Fischer M., Hanauer M.X. and Heigermoser R. (2016), “Synthetic hedge funds”, Review of
Financial Economics, vol. 29 (1), pp. 12-22.
• Hight C. (2019), “The concentration manifesto”, working paper, Alpha Theory.
• Hobbs J. and Sigh V. (2015), “A comparison of buy-side and sell-side analysts”, Review of
Financial Economics, vol. 24, pp. 42-51.
• Hughes J.S., Liu J. and Zhang M. (2010), “Overconfidence, under-reaction, and Warren
Buffett’s investments”, available at https://ssrn.com/abstract=1635061.
• Jensen M.C. (1968), “The performance of mutual funds in the period 1945-1964”, Journal of
Finance, vol. 23, pp. 389-416.
• Lazard (2020), “Less is more: a case for concentrated portfolios”, working paper, Lazard Asset
Management.
• Lesmond D.A. and Stein R. (2020), “The (mutual fund) imitation game: on the performance of
smart copycat funds”, available at https://ssrn.com/abstract=3061898
• Lhabitant F.S. (2007), “Hedge fund indices for retail investors: UCITS eligible or not
eligible?”, Journal of Derivatives and Hedge Funds, vol. 12, pp. 275–289.
• Lhabitant F.S. (2017), “Portfolio diversification”, ISTE Press – Elsevier.
• Martin E.E. (2009), “The intersection between finance and intellectual property: trade secrets,
hedge funds, and section 13(f) of the Exchange Act”, New York Law School Law Review, vol.
53 (3), pp. 575-97.
• Martin G.S. and Puthenpurackal J. (2008), “Imitation is the sincerest form of flattery: Warren
Buffett and Berkshire Hathaway”, working paper, University of Nevada, available at
https://ssrn.com/abstract=806246
• Panchekha A. (2019), “The Singular Impact of High Conviction Overweight Positions for
Active Managers”, working paper, Turing Technology Associates Inc., available at
https://ssrn.com/abstract=3472625
• Pekarek E. (2007), “Hogging the hedge? Bulldog's 13F theory may not be so lucky”, Fordham
Journal of Corporate and Financial Law, vol. 12 (6), article 3.
• Robertson J. (2008), “Hedge funds and public disclosure requirements, is the SEC telling
secrets?”, Nevada Law Journal, vol. 8 (2), article 11.
• Schwarz C.G. and Potter M.E. (2016), “Revisiting mutual fund portfolio disclosure”, The
Review of Financial Studies, vol. 29 (12), pp. 3519-3544.
• SEC Office of Inspector General (2010), “Review of the SEC’s Section 13(f) reporting
requirements”, report no 480, available at https://www.sec.gov/files/480.pdf
• Solactive (2016), “Guideline relating the Solactive Guru Index”, version 1.6.
• Su C., Zhang H., Bangassa K. and Joseph N.L. (2019), “On the investment value of sell-side
analyst recommendation revisions in the UK”, Review of Quantitative Finance and
Accounting, vol. 53, pp. 257-293.

18

Electronic copy available at: https://ssrn.com/abstract=3873884

• Thomas J.W. (2011), “Capturing alpha in the alpha capture system: do trade ideas generate
alpha?”, The Journal of Investing, vol. 20 (1), pp. 11-18.
• Verbeek M. and Wang Y. (2013), “Better than the original? The relative success of copycat
funds”, Journal of Banking & Finance, vol. 37 (9), pp. 3454-3471.
• Yeung D., Pellizzari P., Bird R. and Abidin S. (2012), “Diversification versus concentration ...
and the winner is?”, Working Paper Series 18, The Paul Woolley Centre for Capital Market
Dysfunctionality, University of Technology, Sydney.

19

Electronic copy available at: https://ssrn.com/abstract=3873884

Appendix
This appendix provides a quick comparison between the pro-forma performances of the GS
VIP and SACS portfolios, from the 1st of April 2008 until the 31st of December 2020. GS VIP
tracks the GVIP portfolio discussed in Section 3.410, while SACS tracks the portfolio
recommended by the Steadfast Alpha Capture System, which was implemented by one of the
co-authors of this paper and incorporates the improvements discussed in Section 411.

SACS uses a small group of 30 long only investors and hedge fund managers as idea
contributors. These managers have been selected on a discretionary basis and are considered
as being talented in terms of US stock picking. They have a long-term investment focus rather
than a trading one and therefore a low portfolio turnover. They must report their holdings on a
regular basis either in regulatory reports, commercial databases, on their website or in the press.
SACS collects historical data about idea contributors, their underlying high conviction ideas,
and their associated corporate actions. It uses both Bayesian statistics and machine learning
techniques to identify the key behavioral characteristics, investment strengths, weaknesses, and
biases of all idea contributors. This information is then used to analyze all the high conviction
ideas of all contributors. It should be noted that SACS may decide to underweight or even
completely ignore a contributor or reduce the signal strength of some of its ideas if they display
characteristics where that contributor has no clear evidence of historical alpha generation skills,
or if their alpha decay is deemed to be too high relative to their reporting lag. As a result, being
selected as a contributor does not guarantee that your ideas will be retained; having a high
conviction in a particular stock does not either guarantee that this stock will be selected.
Ultimately, SACS produces a list of 50 stocks ranked by the system’s confidence in its
predictions (the recommendation strength, which can be seen as the equivalent of the
significance of an estimation in statistics). The final portfolio allocates equally to the top 30 to
35 stocks of that list.

As evidenced by Figure 1 and Table 1, on the long run, SACS has outperformed GVIP. This
should not come as a surprise, given its more advanced analytics. SACS has also significantly
and quite consistently outperformed the S&P 500. SACS volatility is slightly higher than GVIP
and the S&P 500, but this likely comes from the more concentrated nature of its portfolio (30
to 35 stocks rather than 50 or 500). SACS drawdown results are also impressive. In absolute
terms, SACS maximum drawdown is slightly lower but remains large and occurs at the same
date – one need to remember that the portfolio remains long only and concentrated. However,
the time to recover is significantly shorter, with only 219 days instead of more than 1’000.

From a style perspective, both SACS and GVIP exhibit a growth and a large cap bias. However,
their two portfolios do not overlap much as illustrated by Table 2. GVIP tends to own crowded
stocks and momentum names, which have typically been a magnet for large hedge funds, while
SACS often picks stocks before they become crowded. Overall, all these results clearly show
that the recommendations of Section 4 are additive to performance and deliver much better
results than just selection managers and positions based on size.

10
GS VIP tracks the Goldman Sachs Hedge Industry VIP ETF (GVIP) since its creation on the January 11,
2016. Prior to this, it tracks the performance of the components of the Goldman Sachs VIP List, which is
the model behind GVIP.
11
SACS tracks Notz and Stucki’s AMC Steadfast Equity Portfolio since its creation on November 20, 2020.
Prior to this, it tracks the performance of the components of the portfolio recommended by the Steadfast
Alpha Capture System, which is the model behind the AMC Steadfast Equity Portfolio. Performances are
adjusted for U.S. withholding taxes on dividends, trading costs, and a hypothetical management fee of 1%
p.a. and operational costs of 0.2% p.a.

20

Electronic copy available at: https://ssrn.com/abstract=3873884

 Source: Steadfast. Indices source: Bloomberg. All with dividends re-invested.
Past and simulated performance is no reliable indication of future performance. While best efforts were deployed in
producing such illustrative returns, no assurance or guarantee can be given that the actual trading results achieved in
future will be as successful or similar to the results presented herein. Investment in any investment product is speculative
and involves significant risk, including the risk of loss of capital.

Figure 1: Comparison between the pro-forma performances of the GS VIP and SACS
portfolios, from the 1st of April 2008 until the 31st of December 2020

Source: Steadfast. Indices source: Bloomberg. All with dividends re-invested.

Past and simulated performance is no reliable indication of future performance. While best efforts were deployed in
producing such illustrative returns, no assurance or guarantee can be given that the actual trading results achieved in
future will be as successful or similar to the results presented herein. Investment in any investment product is speculative
and involves significant risk, including the risk of loss of capital.

Figure 1: Comparison between the pro-forma performances of the GS VIP and SACS
portfolios, from the 1st of April 2008 until the 31st of December 2020

21

Electronic copy available at: https://ssrn.com/abstract=3873884

 GS VIP Weight SACS Weight
(%) (%)
Sea Ltd 2.4 DoorDash 3.5
Carvana Co 2.3 Avis Budget Group 3.3
Twitter Inc 2.2 Sea Ltd 3.3
Adobe Inc 2.2 BridgeBio Pharma Inc 3.3
TransDigm Group Inc 2.2 eBay Inc 3.2
Maxim Integrated Products Inc 2.2 T-Mobile US Inc 3.1
Salesforce.com Inc 2.2 Avantor Inc 3.1
Square Inc 2.2 Alphabet Inc 3.1
Applied Materials Inc 2.2 International Flavors & Fragrance Inc 3.1
PayPal Holdings Inc 2.1 Amazon.com Inc 3.1
Source: Steadfast, Goldman Sachs.

Table 2: Top 10 positions held by the GS VIP and the SACS portfolios, as of June 14,
2021

22

Electronic copy available at: https://ssrn.com/abstract=3873884