"The name Hedge Fund is actually a misnomer since it implies a trading strategy where speculative risks are NOT taken; one where the risks in one investment are counterbalanced by other, separate investments".
Over a period of many years I've been tutored on a broad array of subjects by an ex-colleague Frank Dunn. He and I worked together at Kleinwort Benson in the mid-nineties and have been friends ever since. His 'thing' is physics & mathematics and he describes himself thus;
"Frank Dunn is a professional commentator on Hedge Fund strategies & advisor to numerous financial professionals. Along with many noted academics, he has argued for some time that risk models need updating; to reflect the dynamic nature of sentiment risk; that is the changing behaviour of market participants. He uses a variety of highly mathematical techniques to model this; built upon the theory of Social Networks." So naturally this interests me ;)
Frank has been applying social network theory to predicting the movement of the financial markets for the past 10 years. His models detect changes in human behaviour early on and so once you know what mood investors are in, you can predict their decision making behaviour; which is, apparently the easy bit.
I appreciate this subject matter won't be for everyone, but it's one of those pearls of wisdom that has to be placed in the public record as a point of reference, for posterity and for our central bankers. A paper that describes a credible and proven means of generating wealth in a socially responsible way. Where a hedge fund actually 'does what it says on the tin'. “There is actually the opportunity for value creation in being socially responsible.” - Brian Walker
I have to stress that none of what follows constitutes legal, investment or financial advice. So with kind permission;
The route to investment safety:
A brief description of Smoother Composite Investing
The End of Capitalism?
Statistics can fool you. In fact it is fooling your government right now. It can even bankrupt the system (let's face it: use of probabilistic methods for the estimation of risks did just blow up the banking system). Nassim Nicholas Taleb, from a recent Essay entitled The Fourth Quadrant
Many academics other than Taleb have been warning of the frailty of our risk models for many years. The favourite aphorism amongst EU researchers who I met working on the Complexity in Finance initiative was: ”your risk models will fail just when you need then most”. The truth of that should now be painfully self-evident.
The reasons why they were all ignored can be summed up in one word: greed. Everyone was in too much of a hurry making money to reflect on the risks they were taking. However, it is important to understand that there were different characters in the villainy.
Most of the financial services industry is numerically dyslexic; so relies on others to ‘do all the numbers’. Those ‘others’ are the ones Taleb singles out for his most vitriolic criticism. In the same essay he says:
My outrage," he says, "is aimed at the scientist-charlatan putting society at risk using statistical methods.
However, even Taleb struggles to understand the issues clearly. He says there are ‘no known tools’ to model certain features; when in fact the tools exist. They are little-known & poorly understood as is most good wisdom; but they do exist.
The title of his book, Fooled by Randomness, is a mistake. Neither financial markets nor other features of our social and natural world are random; to assume they are random is itself misleading & highly dangerous.
Events & outcomes are uncertain, not just because our models are not good enough; but also because real-world causality is complex: effects have many causes; & causes have many effects. This network of interdependencies between events is sometimes called non-localisation.
Understanding financial markets in these terms enables newer, more robust measures of risk to be employed.
Is there any true Hedge Fund Out There?
As recent events have revealed, most Hedge Funds are run on a highly speculative basis; with little or no understanding of the finer aspects of market risk.
The name Hedge Fund is actually a misnomer since it implies a trading strategy where speculative risks are NOT taken; one where the risks in one investment are counterbalanced by other, separate investments.
The textbook example is a comparison between the transport & energy sectors: when energy costs go up, transport companies usually suffer; & vice versa. A hedge between an airline company and oil company should then smooth out the worst fluctuations in energy prices.
Unfortunately this traditional hedging approach was always a naïve and simplistic strategy. It no longer works; if it ever did. The key flawed simplifying assumption is that the interdependence between the two investments is one stable, single relationship; called the correlation. This is NOT the case.
Markets are fractal in nature: within each phase or trend, there are sub-phases & sub-trends; with sub-sub-phases and even smaller, shorter phases still. Typically, at any time there are 12 separate timescales exhibiting phase-like or trend-like behaviour.
The patterns at each timescale are subtly different. Benoit Mandelbrot coined the term Self-affinity to highlight the patterns are different; compared to simple fractals which have identical patterns and are thus known as ‘self-similar’.
There are thus 12 components to the return in an investment (the trend gradient); 12 components to any sensible risk measure; & 12 components to the co-dependency between any two connected investments. In global financial markets, all investments are now connected in some degree or other; in mathematical terms, the 12-component co-dependency vector is always non-zero.
To hedge the fluctuations in an investment, you need to devise a strategy that smoothes out the adverse price moves caused by fluctuations in timescales shorter than the investment horizon; that is how long you intend to hold the investment.
No-one has been able to do this properly; until now.
Fast-Moving markets are excessive fluctuations of the shortest timescale
The credit crunch is a new & different phase of market behaviour; owing in large part to fast-moving markets; & the excess volatility they cause. Fast-moving markets have been an ever-present feature of the credit crunch.
They have always existed to some degree in headline commodity markets, such as Oil & Gold. An important piece of news would send all traders scurrying to their brokers; creating a sharp move in a matter of minutes; sometimes even seconds. Exchanges have devised mechanisms to limit such moves; by simply preventing further trades from occurring.
However, the contagion has now spread. Stocks, currencies, indeed any investment at all can move several full percentage points on the back of a news item; frequently unconnected to the sector in question; but caused by a domino effect of co-dependencies with the relevant markets.
Fast-moving markets highlight an undeniable facet of price patterns: short-term fluctuations are event-driven and therefore inherently unpredictable. It is not possible to know all the news items of any day in advance; from storms, earthquakes & hurricanes through to what the US Treasury or Federal Reserve might say, or how a committee or government might vote.
Many traders do try to profit from such short term moves by ‘living on their nerves’ and chasing the trends within such fluctuations. Sadly, these intraday traders merely contribute to the instability; amplifying the fluctuations to larger & larger spikes, making current markets the maelstrom of nerves we all witness. One thing they are definitely not doing is trying to hedge these risks.
To provide lower-risk trading strategies, you need to measure the co-dependencies of markets under these very stressed fast-moving market conditions; then choose combinations of investment that minimise the effect of these high-frequency fluctuations.
It turns out that, if you can do this, then measuring the rest of the 12-component co-dependency vector becomes a comparatively simple exercise.
High-Frequency Investment Smoothing: the Process Itself
Our new software is designed specifically to overcome this problem; so that we can offer smoothed (i.e. properly hedged) investment combinations. We have also designed chart production software; so that these investment combinations can be viewed graphically.
In practice what this means is that, for each potential investment we might consider (e.g. stock, currency, commodity or financial future, but avoiding options at the present time), each night we search & scan all other potential investments for the ‘best combinations’. We make our ‘potential investment’ list (what we call our investment universe) as wide as possible; or rather as wide as broking costs permit us to do.
Our guiding principle of maximum diversity; trading as many different investments in as small amounts as possible; has to be offset against the expense of so many transactions. We run a specific optimisation trial to decide on the correct basic trade size for a given set of brokers’ charges.
For all items in our investment universe we collect & clean high-frequency (tick-based) market data (price, volume, etc). For futures contracts we have our own proprietary techniques for merging data series from electronic & open outcry pit-based markets; and for contracts with differing expiry dates.
‘Free Lunch’ criteria
Once we have identified smoothed pairs of the ‘best combinations’ of investments, we then repeat the process seeking to add further investments to the combination; thereby creating smoothed complex composites.
The criteria that we apply to decide what are the ‘best combinations’ do vary. The general aim is to find investments that are counter-dependent (have negative co-dependency) in shorter timescales; but have positive co-dependency in the intermediate & longer term.
Roughly speaking we apply a kind of Dedekind cut (ref. any standard maths text on set theory & the real numbers) to the pattern features we detect in the investment combination. Patterns that reflect features that are longer term than the intended investment timescale are regarded as part of the projected reward; whereas patterns that represent fluctuations of a shorter term than the intended investment timescale are regarded as part of the predicted risk.
The startling thing about this technique is it eliminates risk far more than reward: in financial markets, there is such a thing as a free lunch!
Investment Timescale or Horizon
This begs the question of what is the intended investment timescale. That in itself varies and is critically dependent upon prevailing market sentiment, globally, regionally and within the sector and its most closely-related areas.
Currently most investment combinations are stable for only between 2 to 8 weeks; whereas in calmer times, this would extend out to many months; sometimes years.
The success of the criteria we apply are judged by a technique called competing experts: those criteria with the most successful portfolio performance in recent weeks, as well as the most stable predictions moving forward are used to build complex composites.
Order Preparation, New entries
Finally, when complex composites are judged to meet all our smoothness criteria, they are examined to see whether the timing is right for market entry. This can be compared to traditional charting methods; except of course the price series is that of the investment composite.
Perhaps the truly astounding nature of our investment composites is revealed by the fact the charts they produce give such clear & unambiguous technical signals. In other words, by filtering out short-term fluctuations we have rendered the job of the Technical Analyst considerably simpler.
Our own entry signals are nevertheless not those of the typical chartist trader. Instead we exploit features of the patterns we have detected to identify the lowest risk entry points.
Order Preparation, Adjustment of existing Holdings
Then the same overall process is repeated for all existing holdings; making sure complex composites are still stable propositions moving forward; with or without rebalancing.
The output from all this process is something we call the conditional order list; which contains all the suggested adjustments (full or partial entries & exits) of the whole portfolio. This is intended to be provided on an automated basis via the intelligent interface of the broking system(s).
Handling Actual Trade Execution: Block Trades
For very profound reasons, the instabilities of the current markets have led us to require all trades (apart from a few ‘headline markets’ discussed shortly) are executed at or around the same time of day. This can be seen as a kind of block trade or even a delayed program trade.
In simple terms, this enables us to ignore many of the effects of the unbalanced portfolios that would result from only partial execution of the suggested trades. Such effects might lead to unpredictable risk scenarios and are thus to be avoided.
Even with this constraint there is a need for scenario planning for each day’s trading: dependent upon the patterns revealed by the morning’s (i.e. European) price moves; followed by the opening of the (US-based) trading pits (primarily in New York & Chicago); different subsets of the conditional order list may or may not be executed.
Once the market has evolved sufficiently for a list subset to be enabled for trading; the orders can be confirmed to the broker. Orders are typically of the ‘Entry @ Limit’ kind but they can be accompanied by triggers.
Each successful entry order must be followed by two ‘exit strings’: a list of prices where some or all of the holding must be cut; with the upside string being known as Target Exits; and the downside string usually called Stops.
The distance of prices in the target exit string from the entry price depend on the pattern of projected reward; whereas the distance of prices in the Stop string from the entry price depend on the pattern of predicted risk.
Simply put, if an investment combination stays within the bounds of what we were expecting, we stay with it; if not, we get out. We trade NOTHING that does not do what it is supposed to.
Controls & Checks
Once the full scenario is clear, several portfolio balancing checks are run.
The most important of these is whether the portfolio is still self-hedging: one aim of trading in a widely diverse set of holdings is that they are as risk-independent as possible; that is the exposure to any one particular market or risk is minimised.
However, fast-moving markets can shift this balance very quickly; so it can be necessary to trade some key ‘headline markets’ several times a day. These markets include: the 6 major currencies, the major US & EU stock indices; some commodity futures related to energy, metals & agriculturals.
The important constraint on this intraday futures trading is that it is only allowed in one direction: it must reduce exposure to that market of the portfolio as a whole.
This high-frequency trading will take some while to automate fully.
Another related check is on weak or rogue combinations. If a particular investment composite is requiring constant adjustment; or if it shows an unacceptable sequence of adverse days, even within pattern; or if it exhibits anyone of a number of undesirable instabilities; it is removed from the portfolio.
The basic outlook is there is no need to remain in any investment; that falls short of expectations; there are always more desirable opportunities arising each and every day; the scan & search software reveals where they lie.
The ultimate aim of all these controls and the criteria by which they are judged (by comparing the outcome if they had not been applied) is the reduction in adverse movements of the whole portfolio. We think it possible to find yet further improvements so that the number of ‘successful days’ consistently exceed 90%.
What kind of portfolio do we get?
The essential philosophy is thus to deliver the best possible portfolio for the given market environment. It does this by recognising that different market environments exist.
These can be categorised in terms of the 12-component vectors for Risk, Reward & Co-dependency for each major sector: an empirically measurable view of the state of human sentiment; that is far more detailed & precise than anything previously conceived.
However, it must be stressed that the investment strategy requires no fundamental view of the market. Its aim is to grow a portfolio that is immune to the short-term fluctuations created by unpredictable news events; regardless of which phase of market sentiment currently predominates.
We describe such a portfolio as being Risk-Neutral: in summary, a portfolio that attempts to eliminate all known risks, as can be measured by human interaction with them; leaving only the unquantifiable unknown left to ponder. We shall have to deal with that later.
This document should nevertheless have set out why our performance is so different to anything else; and will remain so. In absolute truth it is the only Hedge Fund.
CEO Market Maths Ltd