The Handbook of Energy Trading

By Stefano Fiorenzani, Samuele Ravelli and Enrico Edoli

To thrive in today's booming energy trading market you need cutting-edge knowledge of the latest energy trading strategies, backed up by rigorous testing and practical application

Unique in its practical approach, The Handbook of Energy Trading is your definitive guide. It provides a valuable insight into the latest strategies for trading energy—all tried and tested in maintaining a competitive advantage—illustrated with up-to-the-minute case studies from the energy sector.

The handbook takes you through the key aspects of energy trading, from operational strategies and mathematical methods to practical techniques, with advice on structuring your energy trading business to optimise success in the energy market.

• A unique integrated market approach by authors who combine academic theory with vast professional and practical experience
• Guidance on the types of energy trading strategies and instruments and how they should be used

Soaring prices and increasingly complex global markets have created an explosion in the need for robust technical knowledge in the field of energy trading, derivatives, and risk management. The Handbook of Energy Trading is essential reading for all energy trading professionals, energy traders, and risk managers, and in fact anyone who has ever asked: 'what is energy trading?'

• Language: English
• Publisher: Wiley
• Release date: Dec 8, 2011
• ISBN: 9781119954552

Book preview

To our families

List of Figures

List of Tables

Preface

Over the past ten years energy trading has proved one of the most capital attractive of businesses. Both in the US and in Europe energy trading has become a fundamental activity for utilities (vertically integrated energy companies), investment banks and hedge funds, enticing a lot of experienced staff from other (more established) trading sectors such as equity, FX or fixed income. Probably due to this fast, and perhaps unexpected, growth in importance, energy trading has developed without the support of either precise or ad hoc benchmarks in terms of organization, management and control, something its complex nature desperately requires.

Energy trading is not just of interest to energy traders, it is also relevant to a very wide range of other professionals engaged in the business, such as energy analysts, risk managers and, last but not least, energy trading managers. This book is intended to cover in a detailed way all the various aspects of energy trading from the most macro to the most micro. Dealing so thoroughly with every topic gave us the chance to analyze many aspects of the business in depth with the aim of providing support to energy trading managers, energy traders and energy trading analysts as well as trying to bridge the worlds of practitioners and academic research. Obviously, in order to be at the cutting edge of every topic discussed we could not avoid the occasional use of technical language or a pretty complex mathematical one. But, we genuinely believe that anyone choosing to read this book is ready for something more than the basics and we wanted to provide them with something more advanced.

Trading in general and energy trading in particular is not a pure science but rather a complex mix of scientific techniques and emotional behaviours. It is impossible to gain a complete understanding of the topic using just one approach. For that reason we will try to explore and critically revise all the potential trading approaches that can result in profits in the energy sector. In particular, the book condenses our combined professional experience in the field, covering quantitative analysis, trading, risk management as well as the managerial aspects of the energy trading business.

In order to keep our readers on their toes in every chapter we alternate theoretical discussion with practical applications and examples. In our view practical applications are important to clarify and exemplify theoretical discussion and sometimes also to show that the use of mathematical complexity is justified and not merely an exercise in aesthetics. We have tried hard to avoid unnecessary complexity, just as we have always tried to avoid dangerous simplifications.

The book seeks to answer questions like: why is energy trading different from any other trading business? Where are those differences? Are energy markets efficient markets or do inefficiencies allow for extra profits? How should we structure an energy trading business in order to maximize the probability of being successful? How many energy trading strategies and energy trading instruments can we use and how should we use them in order to be successful?

The book consists of six chapters, ranging from practical studies of the efficiency of energy markets to quantitative analysis of derivatives and portfolio strategies. Despite the highly theoretical nature of some of the topics analyzed, we have been careful not to lose sight of the practical aspects. The first chapter gives a detailed analysis of the efficiency of energy markets, combining traditional statistical analysis tools with our own market insight. Next follow the presentation and critical revision of the most used energy trading strategies, focusing on different approaches, different markets and trading instruments. The presentation is tailor-made for energy markets, and not merely a simple application of existing theory to energy markets. With many factual examples, and numerical schemes for practical implementation, we intend to provide helpful insights for readership in both academia and industry.

The book ends by dealing with the more macro aspects of energy trading (designated as metatrading elements) such as macroeconomic trends determination, organizational, capital allocation and risk management issues. We believe that this book is comprehensive in setting out for the reader all the relevant problems connected with an energy trading business. On some issues firm solutions to problems have been provided, on others we were simply happy to be the first to engage in discussion!

Stefano Fiorenzani

Samuele Ravelli

Enrico Edoli

The views, opinions, positions or strategies expressed by Samuele Ravelli are his alone, and do not necessarily reflect the views, opinions, positions or strategies of E.ON Energy Trading SE.

Acknowledgements

We would like to express our gratitude to the following people for their conceptual and material help.

to Carlo Occhiena for many profitable discussions on technical trading techniques, which helped us fine tune the details of directional trading strategies;

to Tiziano Varigiolu for his generous advise on both conceptual and technical matters.

Writing a book together means continuously challenging each other, reviewing every section and try to organize the various constituent parts into a whole. So we would like to thank one another for the common collaborative effort aiming at the very best result.

We are also grateful for the challenging comments we received from our anonymous referees preparatory to publishing.

The responsibility for any remaining errors is ours alone.

Stefano Fiorenzani

Samuele Ravelli

Enrico Edoli

Chapter 1

Energy Markets as Efficient Markets

1.1 The Efficient Market Hypothesis

Are energy markets efficient markets? In what sense can a commodity market be considered efficient? And what are the trading implications of market efficiency? These are the questions we would like to answer in this introductory chapter. The third question, in particular, has a strong impact on the overall meaning of the book, but obviously, before discussing the trading implications of market efficiency we need to define the concept formally and to try to understand if energy markets can be considered efficient in some sense.

The theoretical origins of the Efficient Markets Hypothesis (EMH) are connected with pioneering studies of modern financial economics. The first formal definition and in depth analysis must surely refer to the studies of Roberts (1967) and Fama (1970). Whenever we say that a market, a financial market particularly, is efficient we basically refer to efficiency in the informational sense. In fact, in a competitive market, asset prices dynamically reflect relevant information flows and consequently a certain market is more informationally efficient if it is able to reflect more relevant information in its asset prices. Information is thus relevant if it can predict future price dynamics. According to Fama's original definition, a capital market is said to be efficient with respect to a certain information set if security prices would be unaffected by revealing that information to all market participants. Hence, the description of the information set is the base for the definition of market efficiency.

The classic taxonomy of the information set exposed by Roberts (1967) characterizes three different forms of market efficiency: weak, semistrong and strong.

In weak form market efficiency the information set considered includes only the history of prices themselves. If the market we are considering is efficient in the weak form, revealing the whole history of security prices to market participants wouldn't impact their behaviour in the market. In other words, future prices cannot be predicted by analyzing prices from the past. Under this framework, price realizations are time independent of each other. Knowing last price realization or the whole history of prices wouldn't change the ability to predict future prices. This form of market efficiency has strong mathematical and trading implications and will be discussed throughout the chapter.

In its semistrong form, the information set includes all the so-called publicly available information, while in the strong form of market efficiency private information is also considered as a part of the available information set. Under these two frameworks, price adjustments to new information should be immediate and of a reasonable size. Consistent upward or downward adjustments after the initial change should be interpreted as informational inefficiencies.

Independently of the nature of the information set used for the definition of a particular market efficiency form, it is clear that the only way information should be reflected in prices is through real market transactions. Hence, even if the information flow is huge and fast in its development but the frequency of effective transactions is not proportional, the market efficiency paradigm loses its importance. In other words, a market can be efficient only if it is sufficiently liquid, otherwise the information flow cannot be reflected in price signals. Market efficiency is strictly connected to perfect competition between a wide universe of market operators (no single one of which is able to impact price dynamics). If this were not the case, information would be uploaded into prices as discrete packages resulting in sizable price jumps. Often energy or other commodity markets suffer when liquidity problems arise and this is something that has to be seriously considered when testing for market efficiency.

1.1.1 Trading Implications

The efficient market hypothesis has aroused interest in the public debate not because of its theoretical appeal, but mainly for its trading implications. In fact, if the disclosure of a certain information set to market participants does not affect their price forecast, no extra profits can be extracted on the basis of that information set. In the weak-form efficiency case the information set is basically the complete knowledge of price history. This implies that if a market is efficient in weak form, trading based on technical or statistical analysis cannot provide expected excess returns, while fundamental trading can still be profitable. In a semi-strong efficient market not even fundamental trading can be expected to overshoot the market, while the strong efficiency form is that theoretical situation in which not even with private information can we expect consistently to realize extra returns. Obviously, the last situation is a purely theoretical one since if a strongly efficient market were to exist, trading would be meaningless in it (as also a book on trading!).

Looking at how many people work in the trading industry around the world we might well be inclined to say that, in general, financial markets are far from being efficient. But, what is a market in practice? Is it possible that a market can be considered efficient for a certain asset class and not for another? It can happen that information fully reflected into a certain security's price (and hence irrelevant for its better prediction) can be useful to better predict the future price of an other security. It can be that (especially in those markets characterized by a mix of financial and industrial investors) market segmentation allows for an inefficient mechanism of transmission of information and consequently potential excess profits.

A full understanding of the degree of efficiency of the market we are operating in, conditional on the information set we are endowed with, allows us to understand which trading activity is worth setting up and what is the best way of allocating our risk capital. It is interesting to analyze in depth and specify the concept of market efficiency within different market sectors in order to verify the concrete possibility of making money out of our energy trading activity.

1.1.2 Informational and Mathematical Implications

So far we have defined and analyzed the EMH as an information-based concept. In practice, a market is considered efficient if it is able to reflect the information flow immediately into asset price dynamics. However, we have not really formally defined the concept of information. We said that a market framework is efficient at time t with respect to a certain information set φ(t) if revealing that information to market players will not change their future price forecast. This implies, obviously, that market players are already endowed with an information set Φ(t) already containing φ(t) [formally, images/c01_I0001.gif ]. If we imagine that agents formulate their future price forecast by means of the conditional expectation operator, we may write the following expression using the Law of Iterated Expectations:

images/c01_I0002.gif

which formally states that redundant information is worthless in improving (or at least changing) forecasting errors. Discussing the classic taxonomy of information sets related to market efficiency forms we saw that our information set can be the whole history of asset prices or, equally, the whole set of publicly available information. We need to give a formal definition to all those sets before analyzing the mathematical and statistical implications of EMH.

1.2 Mathematical Framework

In this section we introduce the mathematical and statistical background useful to describe and analyze the EMH and briefly present some mathematical theory.

1.2.1 Information Set: σ-algebras

From a mathematical point of view the concept of information set (available at time t) is given by the definitions of σ-algebra and filtration: let Ω be an abstract space, that is with no special structure, and let 2Ω denote all subsets of Ω, including the empty set ∅. Let images/c01_I0003.gif be a subset of Ω. images/c01_I0004.gif is a σ-algebra if it satisfies the following properties:

i. images/c01_I0005.gif and images/c01_I0006.gif

ii. If images/c01_I0007.gif then images/c01_I0008.gif , where Ac denotes the complement of A

iii. images/c01_I0009.gif is closed under countable unions and intersection: that is if A1, A2, … is a countable sequence of events in images/c01_I0010.gif then images/c01_I0011.gif and images/c01_I0012.gif are both also in images/c01_I0013.gif .

The pair images/c01_I0014.gif is called a measurable space.

Even if the general definition does not require a special structure for Ω, we will always consider Ω as the space of all possible events for a given experiment. In probability theory, an event is a set of outcomes to which a probability is assigned. For example, if the experiment is the roll of a dice, an event is a five appears, and it is intuitive that the probability assigned to this event is images/c01_I0015.gif .

Now the problem is how to describe spaces of events that have a complex structure, with an infinite and uncountable number of events, collected in the σ-algebra images/c01_I0016.gif . First of all we have to define a probability measure.

Given a measurable space images/c01_I0017.gif we say that the real-valued function images/c01_I0018.gif defined on images/c01_I0019.gif is a probability measure if it satisfies:

i. images/c01_I0020.gif

ii. For every countable sequence images/c01_I0021.gif of pairwise disjoint elements (i.e. images/c01_I0022.gif

whenever i ≠ j) one has:

images/c01_I0023.gif

The triplet images/c01_I0024.gif is called a probability space.

The last mathematical object needed to describe the price evolution of an asset in a mathematical framework is the random variable (and the stochastic process). A random variable X is an unknown quantity that varies with the outcome of a random event. Before the random event, we can know which values X might possibly assume, but we do not know exactly which one it will take: let images/c01_I0025.gif be a probability space and let images/c01_I0026.gif be a measurable space. A random variable is a images/c01_I0027.gif -measurable function images/c01_I0028.gif , i.e. such that images/c01_I0029.gif one has:

images/c01_I0030.gif

Since images/c01_I0031.gif we can consider the following probability on E, named the law of X:

images/c01_I0032.gif

An interpretation of this definition is that the preimages of the well-behaved subset of E (the elements of images/c01_I0033.gif ) are events, i.e. elements of images/c01_I0034.gif , and hence are assigned a probability by images/c01_I0035.gif . One can prove that μx is a probability measure on images/c01_I0036.gif in the sense of the definition given above.

At this point it seems clear that a price in the market can be viewed as a realization of some random variable: at time t, given a suitable set of information images/c01_I0037.gif , the price realization will be described by images/c01_I0038.gif . But how to describe the sequence of prices and the flows of information? A possible way might be to model a sequence of information sets and a sequence of random variables.

Let us start with the sequence of information: given a measurable space images/c01_I0039.gif , a filtration is an increasing sequence of σ-algebras in images/c01_I0040.gif , i.e. images/c01_I0041.gif is a filtration if

i. images/c01_I0042.gif for each t

ii. if t1 ≤ t2 then images/c01_I0043.gif

Let us assume that the σ-algebra images/c01_I0044.gif is the information set available at time t, which is increasing. If we want to have a mathematical description of the information set given by the history of prices we have to introduce the concept of σ-algebra generated by a set and by a random variable: if C ∈ 2Ω is a subset of Ω then the σ-algebra σ(C) generated by C is the smallest σ-algebra containing C; if images/c01_I0045.gif is a random variable then the σ-algebra generated by X, denoted by σ(X), is the σ-algebra images/c01_I0046.gif , generated by the subset images/c01_I0047.gif .

A stochastic process is a sequence of random variables indexed by a set images/c01_I0048.gif : images/c01_I0049.gif . In particular, for each fixed t, the mapping w → Xt(ω) is a random variable; for each ω ∈ Ω fixed the mapping t → Xt(ω) is a deterministic function of t, called the trajectory of X for the outcome ω.

We are now (finally) ready to describe the evolution of a price in the market: at every time t a certain information set images/c01_I0050.gif is given and a certain event images/c01_I0051.gif happens; then the realization of the stochastic price x = Xt(H) follows and the market operator can see it.

If we are in a weak efficient market then the only information available at time t is the history of the prices in the time interval [0, t), so the information set at time t is given by the σ-algebra images/c01_I0052.gif that is the smallest σ-algebra generated by the history of the process.

In semi-strong form efficiency the information set is clearly bigger than images/c01_I0053.gif : we also have to consider the public information images/c01_I0054.gif ; in this case the σ-algebra is given by images/c01_I0055.gif ; in strong efficiency the information set images/c01_I0056.gif is given by the history of prices images/c01_I0057.gif , the public information images/c01_I0058.gif and also the private images/c01_I0059.gif information, so images/c01_I0060.gif . It follows that

1.1 1.1

1.2.2 Conditional Expectation

Another important concept in probability theory, useful here to describe the efficiency of the market, is that of conditional expectation. Let images/c01_I0062.gif be a probability space, images/c01_I0063.gif a random variable and consider a σ-algebra images/c01_I0064.gif included in images/c01_I0065.gif . We would characterize images/c01_I0066.gif , i.e. the expectation of X given that we have access to the information in images/c01_I0067.gif . It is not trivial to formalize this notation, so we start with a simple intuition. Recall that the unconditional expected value is given by

images/c01_I0068.gif

which is a weighted average of the values of X with respect to the weights given by images/c01_I0069.gif . Suppose now that we have access to this extra information: we know that all the possible events lie in a subset of Ω (in particular, a subset of images/c01_I0070.gif ), i.e. the only possible events one can observe form a subset images/c01_I0071.gif such that images/c01_I0072.gif . It seems reasonable to normalize our probability measure images/c01_I0073.gif so that we have the