Honing An Algorithmic Trading Strategy

part 2:

Honing an algorithmic trading strategy

41

Chapter 4 Choosing the right algorithm for your trading strategy

51

Chapter 5 Anonymity and stealth

59

Chapter 6 Customising the broker’s algorithms

■ Chapter 4 Honing an algorithmic trading strategy

Choosing the right algorithm for your trading strategy What are the options buy-side traders need to consider in selecting an algorithm best suited to a particular investment style? Tracy Black* and Owain Self**

trading styles and Ia nvestors benchmarks can be different for number of reasons. Index track-

algorithms without going into the complex mathematics behind their construction.

ers and fund managers may have specific price goals such as the close price, whilst others may be more constrained to the price at which the investment decision was made. Some investors may be investing into 3-4 days volume of a stock and others in smaller more frequent trades. Reactions to changes in prices and volumes over the course of the implementation of their trade and, generally, the degree of impact they are willing to have in order to complete, will differ from investment case to investment case. Whether or not an investor’s trading style involves large slow money orders or smaller high frequency trades, there is a place for algorithms if used appropriately. In this chapter we will explore the attributes of the more commonly available

Algorithmic choice When deciding if you can utilise an algorithm to execute a particular order a number of questions need to be answered. Besides the obvious question of ‘what is my benchmark?’ other factors will ultimately dictate whether algorithmic trading is an option and, if so, what type of algorithm and which parameters to apply. First, you need to assess whether the stock is suitable. Blue chip liquid names that trade a large percentage of their volume on the order book will be good candidates. Small/mid cap stocks that trade a very small percentage on the order book are only suitable with correct parameterisation. The reasons for this are obvious, a computer can only react to

■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

41

**Owain Self , executive director – Equities, UBS Investment Bank *Tracy Black, executive director – European Sales Trading, UBS Investment Bank

■ Chapter 4 Honing an algorithmic trading strategy

■

“VWAP algorithms do not generally take into account absolute volume levels and will still try to complete the order even if this would cause additional market impact.”

42

information that is electronically fed to it, it cannot participate in off market prints and it cannot make phone calls to negotiate block trades. Secondly, you have to decide what proportion of the order you want to execute via an algorithm. You might want to put your entire order into an algorithm if it suits your benchmark or you may want to combine algorithms with more of a traditional trading service, such as Block Trading or DMA. Making the process even more complex for the client is the fact that not all algorithms are equal. Brokers have different names for algorithms that have similar trading styles and sometimes two firms will offer an algorithm under the same name, which will execute very differently. This creates a minefield for the client. Only through education from brokers on what to expect of their algorithms and through actually ■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

using them will a client be able to know which algorithms at which firms best suit their trading style. Undetermined benchmarks Algorithms can generally be split into two types of benchmark, predetermined and undetermined. First generation algorithms try to obtain a yet undetermined benchmark, such as VWAP, where the benchmark will be determined over the life of the order. The more recently developed Implementation Shortfall algorithms will be measured against a benchmark predetermined at order creation. VWAP (Volume Weighted Average Price) has been the most commonly used algorithm historically. As things have evolved, VWAP has gained its critics but it still has its uses. Ultimately used with the aim to minimise market impact, VWAP is useful for executing trades where you don’t necessarily have a view on a stock and want to obtain a fair price by sampling market levels over a specified period. Due to its sensitivity to changes in volume distribution, VWAP will participate relative to liquidity. However, VWAP algorithms do not generally take into account absolute volume levels and will still try to complete the order even if this would cause additional market ■ THE TRADE 2005

■ Chapter 4 Honing an algorithmic trading strategy

■ ALGORITHMIC TRADING

Figure 1: GlaxoSmithKline PLC 16.35

8.00

2004/07/12 2004/10/1

te

2004/07/12 2004/10/1

2005/07/29

Da

2005/07/29 2,800,000

Volume

2,800,000

2,400,000 2,000,000

2,400,000 2,000,000

1,600,000

1,600,000

1,200,000

1,200,000

800,000

800,000

400,000

400,000

0

8.00

Time

16.35

0

43

Figure 2: LogicaCMG PLC 16.35

8.00

2004/07/12 20,050,111 2005/01/11

te

2004/07/12 2005/01/11

2005/07/25

Da

2005/07/25 800,000

800,000

Volume

impact. It is therefore important that traders apply price and volume caps on larger trades to minimise such an impact, although this potentially means the order will not complete. The expectation of VWAP is not just about mean performance; we know that if you executed an order each and everyday of the year in the same stock, the mean performance would be acceptable, however this is an unlikely scenario. Therefore, the risk-adjusted performance and the standard deviation of the returns become important. Due to VWAP’s sensitivity to volume distributions, the performance risk is affected by its ability to predict changes in these distributions. The majority of trading engines are based on historical data. This can often mean that you have to choose your stocks carefully. Some will have a relatively stable historical trading pattern – e.g. GlaxoSmithKline (Fig. 1) – and therefore a more predictable outcome. Others, however, can have a more volatile trading pattern historically – e.g. LogicaCMG (Fig. 2). In these cases, using an average historical curve will not deliver acceptable performance, as your standard deviation would be too large. It is important that the algorithm you are using can predict trading

400,000

400,000

0

8.00

Time

16.35

patterns based upon both historical and real-time data analysis. This results in an improved standard deviation of returns even when trading less suitable stocks. ■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

0

■ Chapter 4 Honing an algorithmic trading strategy

■

“Neither VWAP nor TWAP have any macro level price sensitivity – the overall profile of execution is not affected by movements in the price of the stock. Their aim is to get the order executed by the end time, irrespective of price.”

44

VWAP’s sensitivity to the volume distribution over the life of the order may not always suit your trading style. For example, if you need to execute an order over the remaining hour of the day. If you use a VWAP algorithm to trade, then based on the volume distribution you could execute 25%-50% of your order in the closing auction. This ties your execution price to the closing price and samples fewer market prices during continuous trading. If your aim is to minimise market impact by sampling over a period of time but you have a lower sensitivity to changes in volume profiles, then TWAP (Time Weighted Average Price) can be a useful algorithm. Early versions of a TWAP algorithm simply split the order into portions of equal quantity and time. At the end of each portion, if ■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

stock had not been bought bid side the algorithm would pay the offer. Besides, the obvious footprint left behind by doing the same trade in the same size repeatedly, paying the offer only because time dictates, will lead to poor execution quality. Advanced TWAP algorithms trade in a more sophisticated manner, deciding the prices they pay in the market on the basis of how much they are ahead/behind and whether it is the right price, whilst still trying to achieve an even average. Important to note here is that neither VWAP nor TWAP have any macro level price sensitivity – the overall profile of execution is not affected by movements in the price of the stock. Their aim is to get the order executed by the end time, irrespective of price. This macro price sensitivity needs to be added by way of price limits. The inbuilt price sensitivity of these algorithms will be at a micro level – i.e. the part of the algorithm that makes the decisions on the individual executions of the order. The price sensitivity here results in the algorithm deciding how much risk it can take in order to take advantage of favourable prices, the degree of sensitivity can often be set by a risk aversion or aggression level. This will dictate to the algorithm how far it can fall behind before needing to ■ THE TRADE 2005

■ Chapter 4 Honing an algorithmic trading strategy

pay the offer and how far ahead it can get when buying bid side. Higher risk aversion generally results in a tighter standard deviation of returns at the expense of lower mean performance. The opposite is true with a lower risk aversion level. Inline/Percentage of Volume algorithms also target an undetermined benchmark. The aim is purely to participate with market volume at a rate specified by the user. The Inline algorithm is sensitive to absolute changes in volume levels. This results in it actively participating when volume trades and scaling back if volume does not permit. The level of participation sets the aggression level of this strategy – i.e. how much impact you are willing to have in order to get the trade completed more quickly. An investor who wants to trade if liquidity permits but does not want to have significant impact will choose a low level, e.g. 5%. Someone who wants to execute more quickly at the cost of added impact will choose an aggressive level, e.g. 33%. One major concern for this style of execution is that 33% has become the market default, and the market impact of such a high percentage is often underestimated. For example, when buying at a rate of 33% and 10,000 shares trade away from ■ ALGORITHMIC TRADING

■

“Generally, Inline algorithms do not have any macro price sensitivity besides price limits. The micro sensitivity is the same as in VWAP/TWAP, deciding what price to pay based on your risk (amount ahead/behind).”

you, you wouldn’t need to buy 3,300 to catch up, you would need to buy 5,000. This is because you need to be 33% of the total volume of 15,000 once you have traded. Participating at a rate of 33% means you have to trade 50% of volume that trades away and this is amplified as the target percentages get higher. For example, take two buyers at a rate of 33%, combined they would need to be 200% of any volume missed. Due to this compounding nature of Inline algorithms, stocks are often seen spiralling out of control, with algorithms participating at unfavourable levels. Generally, Inline algorithms do not have any macro price sensitivity besides price limits. The micro sensitivity is the same as in VWAP/TWAP, deciding what price to pay based on your risk (amount ahead/behind). Efficient Inline algorithms will manage risk in terms of liquidity and not simply ■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

45

■ Chapter 4 Honing an algorithmic trading strategy

■

“If you are executing a high frequency of smaller orders, the generation of which are often triggered by the price of the stock, Implementation Shortfall algorithms are the most suitable.”

46

in terms of the actual percentage it has traded. For example, it is better to target 30% +/- a normal trading size in the stock versus being 30% +/-5%. The latter will result in inconsistent risk throughout the life of the order. There will be little risk at the beginning and therefore no favourable prices, but high risk towards the end of the order. Predetermined benchmarks Predetermined benchmarks have been a more recent trend, using, for example, the mid price at initiation or the previous nights close. The algorithms to use in these situations are Implementation Shortfall and Arrival Price style algorithms (unfortunately, these names are sometimes used to describe the same or different types of algorithm). What we can do is split these into two distinct trading styles, ones with low macro price sensitivity, which for the purpose of this chapter we will call ■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

Implementation Shortfall, and ones with high macro price sensitivity, which we will call Arrival Price. Implementation Shortfall style algorithms are designed to minimise the average shortfall over a number of trades. This shortfall is measured as the difference between the execution price and the price at initiation. To minimise this we need to find the optimal level between how much we move the price (market impact) and how long we work the order (risk). We know that if we didn’t execute in the market we wouldn’t have market impact but we are exposed to movements in the stock. Conversely, if we bought the entire amount immediately, we would no longer be exposed to future movements but could have extremely large market impact. In order to optimise the execution, the algorithm will determine when and how much to trade by taking into account a number of factors, primarily the size of the order, the stock’s liquidity, volatility and the time remaining. Generally this will involve being more active in the market initially, as the stock price will be at your benchmark and becoming less active as a higher proportion of your order is completed. This is not a new concept; clients have been using this trading style for many years. ■ THE TRADE 2005

■ Chapter 4 Honing an algorithmic trading strategy

Typically a percentage of the order was traded on risk to start, then worked relatively aggressively in the market. Once the majority had been executed more passive execution would follow. This ultimately is what the Implementation Shortfall algorithms do, they decide how aggressively to trade based on how much risk they are offsetting. If the stock had low volatility then this reduces how far it is expected to move so you can afford to be less aggressive and have less impact. However, with a volatile stock you can afford more impact to reduce the larger risk. For the algorithm to optimise this trade off it needs to know the full extent of the order, otherwise it miscalculates how much risk it is actually offsetting. One thing to note about this type of algorithm is that it will not necessarily be more aggressive below the initiation level. Supporting a stock at a certain level in the market does not minimise market impact. Market impact isn’t just judged as the amount you move a stock against you, it is also a measure of how much you restrict the stock from moving for you. When using an Implementation Shortfall algorithm, you need the ability to set macro sensitivities, such as volume and price caps, but you also need to understand that ■ ALGORITHMIC TRADING

■

“If a stock is trading on the favourable side of your benchmark they [Arrival Price algorithms] become very aggressive until the order is complete. If the stock is moving away from the benchmark they become much more passive.”

these can sub-optimise the strategy. A volume cap might mean you are unable to be as aggressive at the beginning of the order when the stock is at the initiation level. A crucial parameter is some kind of risk aversion (aggression) setting. Algorithms have been built with a risk level in mind. However, your appetite might be very different. You may believe you have more alpha and therefore are willing to take more impact in order to get the trade executed quickly and reduce risk; in this case you should choose a higher risk aversion level. Arrival Price algorithms are the other style of algorithm in this space. These work similarly to that of Implementation Shortfall, but have inbuilt macro price sensitivity to your benchmark (usually mid at initiation or a set level such as previous close). If a stock is trading on the favourable side of your ■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

47

■ Chapter 4 Honing an algorithmic trading strategy

according to the price of the stock will deliver similar results.

Probability density

Figure 3

Arrival Price

Implementation Shortfall

Price achieved

48

benchmark they become very aggressive until the order is complete. If the stock is moving away from the benchmark they become much more passive. This results in a different distribution of returns to that given by the Implementation Shortfall algorithm (Fig. 3). In Implementation Shortfall we get a relatively symmetrical distribution of returns. However, with Arrival Price we get a skewed distribution. This is because we often complete our order before we get the chance to participate at more favourable prices and as we slow down when the stock moves away we potentially participate at very unfavourable levels. This skewed distribution of returns can be seen in any algorithm with macro price sensitivity. Inline algorithms that vary their participation rate ■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

Order execution The proportion of your order executed via an algorithm is also important. If you are executing a high frequency of smaller orders, the generation of which are often triggered by the price of the stock, Implementation Shortfall algorithms are the most suitable. Given the entirety of the order the algorithm can work out how best to optimise execution and minimise the shortfall on average. However, if you had a large order that was also measured relative to the price at which the investment decision was made, a standard Implementation Shortfall algorithm may not be suitable; one reason being optimal execution may take several days. The algorithm will need to know completion is not required by the end of day one and each day following, it would need to know all the details of the previous algorithms. In this scenario we don’t necessarily have to rule out algorithmic execution, it just means more control will need to be taken. You can still utilise a combination of algorithms to achieve the desired results. Many people will start trading aggressively with a small part of their order using an Arrival Price or Inline algorithm. As and ■ THE TRADE 2005

■ Chapter 4 Honing an algorithmic trading strategy

when they have completed the majority of the order or the stock has moved significantly, they will start to use the more passive algorithms such as VWAP/TWAP for remaining portions of the order; ultimately replicating the trading pattern of the Implementation Shortfall strategy. The point to note here is that an algorithm is not the be all and end all for a particular order. For example, if you had an order that was benchmarked against VWAP and you felt you could add value to the execution, you may put half of the order into an algorithm and choose levels to execute the remaining via other algorithms, a block desk or DMA. In summary, when assessing which algorithm to use you have to take into account which type best suits your benchmark. Then you need to decide on your sensitivity to changes in price and volume levels. Algorithms with low sensitivity to price movements will give you a more symmetrical distribution in returns for both rising and falling markets, highly sensitive ones will give you a skewed distribution. Sensitivity to volume changes will ultimately result in you trading with the crowd and less independently. Additionally, you need to decide on your appetite for risk. Algorithms with lower risk aversion will give you a ■ ALGORITHMIC TRADING

■

“It is becoming more common to see algorithms that are sensitive to additional factors such as momentum indicators, mean reversion and relative performance.”

better mean but at the expense of a higher standard deviation, and a higher risk aversion will tighten deviation but at the expense of the mean. The depths to which sensitivities to external factors can be introduced are endless. It is becoming more common to see algorithms that are sensitive to additional factors such as momentum indicators, mean reversion and relative performance. In order to gain access to the best possible algorithms to suit their trading styles, investors will need to retain close relationships with sell-side brokers who can deliver customisable algorithmic solutions as trading styles evolve. ■ © UBS 2005. All rights reserved.

■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

49

■ Chapter 5 Honing an algorithmic trading strategy

Anonymity and stealth What assurances can the sell-side offer to safeguard the client’s alpha capture and minimise information leakage? Richard Balarkas*

let’s be clear what we mean Fareirst by anonymity and stealth as they two quite different things: Anonymity – Refers to the expectation that information relating to the identity of the client, information which the client must of necessity give to the broker, is not divulged in the trading process or at any subsequent point postexecution. Stealth – Refers to the act of moving, proceeding, or acting in a covert way. In conflict and game play this denotes achieving ones objective without being detected, uncovering what others are attempting to conceal or obscure, and otherwise avoiding conflict. Putting it another way, anonymity is important in instances where not the order, but the identity of ■ ALGORITHMIC TRADING

who is behind the order, is itself capable of moving the price. Stealth is the act of completing any order in a manner which reveals as little as possible to the wider market in the hope of minimising impact. Are anonymity and stealth important? In 1997 the silver market was in the doldrums. From July ’97 until early ’98 its price rose 25% (at one point it was up 50%). In February ’98, Berkshire Hathaway, the investment vehicle of Warren Buffet, famous for building large if not controlling stakes in corporate stocks with long term value and a strong brand image, announced that it had been buying silver. (Buffet already had a 32-year investment history during which time Berkshire Hathaway had risen by ■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

51

*Richard Balarkas, global head of AES™ Sales, CSFB

■ Chapter 5 Honing an algorithmic trading strategy

■

“Anonymity isn’t restricted to active managers. Passive managers also need to take care that repeated habitual portfolio slices are not sending signals that others can learn to anticipate.”

52

an average 33% each year.) With regard to silver he had the same information as everyone else, essentially that demand was running ahead of supply and appeared likely to continue for the foreseeable future. Over a seven-month period he bought silver through a single broker without taking any position in futures or options. He amassed what amounted to more than 25% of the world’s supply. Years later commentators were still debating whether he had actually taken delivery of the silver, or still owned it, or leased it out… Buffet understands the value of anonymity and stealth. He needs to. Regarded by many as the oracle of the investment world, Buffet’s every move is watched via scores of internet sites selling Buffet-related books and software, hosting chat room threads, Buffet-dedicated sites, fan clubs etc. There is a whole industry out there trying to guess his ■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

next move in order to beat the market to the punch. Any public statement on his next hunch would be a self-fulfilling prophecy as the investment herd try to anticipate his move. It is not surprising that the last place to look for his ideas is the Berkshire Hathaway home page. Buffet appreciates the value that can be lost through information leakage. So did we at CSFB when we constructed our Advanced Execution Services (AES™) algorithmic trading capability. Like many of the features that are at the heart of our algorithmic trading service, the principles of protecting client anonymity and stealth trading were already embodied on our trading floor and in our trading practices, algorithms simply gave us a new medium in which to take anonymity and stealth to the next level. Valuing anonymity From a user perspective, the process of selecting whose algorithms to use should be based primarily on performance. Piles of colourful marketing literature may give some vague insight into how different broker services are constructed and delivered, and the similarities that are present in high-level marketing descriptions of tactic objectives may create the ■ THE TRADE 2005

■ Chapter 5 Honing an algorithmic trading strategy

impression that all broker algorithms are much the same thing and achieve very similar results. This is far from true. The construction of trading algorithms and their further refinement through practical use is a highly quantitative process. Clients clearly believe anonymity and stealth are extremely important, and in seeking to continuously improve the performance of CSFB’s algorithms it would be ideal to disaggregate the performance in order to focus on those components where the potential value-add is the highest. However, the retention of alpha gained through anonymity and stealth are hard components for a broker to measure. The benefits of anonymity will be readily understood by buy-side traders, many of whom are regularly handling orders that are on average multiples of ADV where revealing the size alone would be sufficient to move the price. However, knowing who is behind a trade has additional informational value. The better a money manager is perceived to be at stock selection and timing, the greater the informational component of the trade, and the greater the likelihood that if this information leaks out the market will ■ ALGORITHMIC TRADING

Anonymity

Early attempts at anonymous trading were not entirely successful

move in anticipation. And the benefit of trading with full anonymity isn’t restricted to active managers. Passive managers also need to take care that repeated habitual portfolio slices are not sending signals that others can learn to anticipate. Science of stealth Whereas anonymity has been enshrined in CSFB’s AES™ service from the start, stealth tactics can always be improved and is the area our AES™ developers find the most exciting. Many beginners think that playing poker online will prove to be completely different than playing offline and they are sometimes

■

“The retention of alpha gained through anonymity and stealth are hard components for a broker to measure.” ■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

53

■ Chapter 5 Honing an algorithmic trading strategy

■

“Even if anonymity is assured, an algorithm will not perform well unless it uses stealth in order to take advantage of other traders and other less sophisticated algorithms.”

54

correct, though usually for the wrong reasons. There is a common but misguided notion that it is much harder to ‘read’ your opponents when you do not see them sitting at the table. What most fail to recognise is that the majority of available information when making a decision comes from a variety of factors other than ‘reading’ your opponents faces. Most of the required information comes from patterns, position at the table and the hands your opponents play. It’s the same when you are trading on a public limit order book – you cannot see your opponents but if you can read their signals they may, often unwittingly, reveal their intentions. At the same time you must be careful that none of your actions are giving your game away. The winner is the one who can coax traders on the other side of the touch to cross the spread and pay the premium. The winner is the one who ■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

can, when necessary, pay inside the spread or even cross the spread without being so aggressive as to send out signals, keeping trades ‘information-less’. The winner is the one who can spot reversion, whose participation is overweight on the dips when buying and on the highs when selling. So even if anonymity is assured, an algorithm will not perform well unless it uses stealth in order to take advantage of other traders and other less sophisticated algorithms. The use of stealth is also defensive, as there are plenty of trading models out there designed to make money from reading signals generated by less sophisticated traders and black boxes. At the market micro-structure level stealth is important, and CSFB’s AES™ incorporates advanced probability and game theory tactics in order to outwit the opposition. ■

■ THE TRADE 2005

■ Chapter 5 Honing an algorithmic trading strategy

Game theory Although game theory has been studied since the 1940s, it has only recently been applied to the world of finance. Game theory champions garnered the 1994 Nobel Prize in Economics, and, today, this theory is used to analyse everything from the baseball strike to auctions. Increasingly, game theory is making its mark as a potent tool for traders. In simple terms, game theory is the study of conflict based on a formal approach to decision-making that views decisions as choices made in a game. Whether playing individually or in a group, each player in a conflict has more than one course of action available to him, and the outcome of the ‘game’ depends on the interaction of the strategies pursued by each party. Algorithms can take advantage of the fact that game theory and probability often have the edge over human intuition. To illustrate this, here are some problems where the answer does not appear to be intuitive, and in one case is actually counter-intuitive (for answers and explanations, see pages 56 and 57): Problems

55

Example 1. If you throw a die until the running total exceeds 12, what is the most likely final total? Example 2. This is a demonstration of the power of faith in random decision-making over simple logic and probability. It was inspired by the format of an old USA TV gameshow ‘Let’s Make A Deal’, hosted by Monty Hall. The conundrum is that you are on a game show and given the choice of three doors: Behind one door is £1million, behind the others nothing. You are invited to pick a door. The host, who knows what’s behind the doors, opens one of the two remaining doors to reveal there is nothing behind it. He then invites you to pick again between the two remaining doors. Is it to your advantage to switch your choice? Example 3. You are in a game of Russian roulette, but this time the gun (a six-shooter revolver) has three bullets in sequence in three of the chambers. The barrel is spun only once. The two players then take it in turn to pull the trigger. If they live, the gun is passed to the other player who then pulls the trigger, etc. Would you rather be first or second to shoot? Continues overleaf ➧ ■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

■ Chapter 5 Honing an algorithmic trading strategy

Game theory (continued) Answers & explanations Example 1. Answer: 13 The way to get a final total of 13 is to build up some total between 7 and 12 inclusive, then make a single throw of the appropriate value. The way to get a final total of 14 is to build up some total between 8 and 12 inclusive, then make a single throw of the appropriate value. Thus if we take the list of sequences producing 14, then subtract 1 from the final throw of each sequence, we will have part but not all of the list of sequences producing 13. Moreover, corresponding sequences are equally likely to occur, because they contain the same number of throws. Thus 13 is strictly more likely than 14. A similar argument shows that 14 is strictly more likely than 15, and so on. Hence 13 is the most likely total

56

Example 2. Answer: You should change your choice. The problem is called ‘counter-intuitive’, because the answer seems for many to defy instinct and logic, even after it’s been explained several times. Most contestants on Monty Hall’s show were apparently reluctant to change their original choice for fear that it was right, or because intuitively they felt that probability could not be altered by revealing one of the ‘losing’ doors. The door you originally chose was a 1-in-3 chance – i.e., the likelihood of your guessing the winning door was 1-in-3. The ‘other’ door is now a 1-in-2 chance, and the likelihood of your guessing the ‘other’ door to be the winning door is 1-in-2. You are 50% more likely to correctly guess a 1-in-2 chance than a 1-in-3 chance, so pick the other door in preference to your original choice of door. If you’re still in doubt, imagine there are 20 doors – one has the money, the others nothing. You pick a door. Then 18 doors are opened revealing nothing, leaving your choice and the one other door. Would you change your choice now? By switching doors you’d improve your chances from 1-in-20, to 50:50 evens, or (depending on how you look at it) arguably 19-in-20. Still sceptical? How about 100 doors? Pick a door. Open 98 revealing nothing, leaving two doors, one a winner and the other a loser. Would you still prefer your original 99-to-1 shot compared to the alternative that is at worst 50:50, and arguably a massive 99% chance?

■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

■ Chapter 5 Honing an algorithmic trading strategy

Game theory Answers & explanations Example 3. Answer: Player 2 is preferable. All you need to consider are the six possible bullet configurations: BBBEEE EBBBEE EEBBBE EEEBBB BEEEBB BBEEEB

➠ ➠ ➠ ➠ ➠ ➠

player 1 dies player 2 dies player 1 dies player 2 dies player 1 dies player 1 dies

57

■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

■ Chapter 6 Honing an algorithmic trading strategy

Customising the broker’s algorithms How much flexibility does the buy-side trader require to adjust and fine-tune the broker’s algorithmic models? Richard Balarkas*

s conference organisers try to A squeeze more and more ‘algo’ events into an increasingly crowded space, the topics to be discussed by their expert panels seem increasingly innovative. So whilst the majority of money managers have yet to enjoy their first algorithm experience, many conference organisers are already filling their bills with debates headlined – ‘Algorithms: does one size fit all?’ – ‘Algorithms: is customisation the future?’ – and, ‘Algorithms: canned or customised?’ What I want to do in this short chapter is explain how customisation is not a future trend, but a feature that has been around since day one. At the same time I want to show how it is incorrect to categorise all broker-provided algorithms as ‘canned’, as if they were the trading equivalent of a fast ■ ALGORITHMIC TRADING

food hamburger outlet. Not so – if you want your algorithm on organic bread with the gherkin removed from the pickle, your initials spelled out in caviar on top with strips of spring onion laying strictly north to south (Tuesday and Fridays only) – it would be our pleasure. It is important to recognise that the term ‘algorithm’ has, unfortunately, been stretched to include not only the most complex mathematical trading models but also very mechanical and simplistic trading techniques such as ‘iceberging’ (the simple drip feeding of an order into the market in predefined clip sizes). In some cases even ‘stop loss’ orders have been defined as algorithmic tactics. There is nothing disreputable about this – good results can be achieved using these tactics if you ■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

59

*Richard Balarkas, global head of AES™ Sales, CSFB

■ Chapter 6 Honing an algorithmic trading strategy

■

“The ability to tweak the parameters means there is significant scope for customising each algorithm on an order by order basis.”

60

pick the right spots, but it is important to recognise that some algorithms are more… well, ‘algorithmic’ than others. For example, those designed to anticipate volume curves, react dynamically to complex signals, and trade with stealth to minimise impact are far more advanced than their more mechanical stable mates and, as a result, are offered by fewer brokers. It is worthwhile making this distinction between the more complex algorithms and simplistic mechanical options. ‘Customising’ the latter is not particularly challenging, and it is understandable that users might perceive such tactics as all being the same regardless of the provider. Leading on from the above, if the serious algorithms are such complex beasts in which highly qualified knowledge engineers at the major brokerages have embodied the firms trading skills, the first issue worth exploring is why a client might wish to customise an algorithm at all. Algorithms are, ■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

after all, efficiency tools. Over and above deciding which order is suitable for trading through which strategy and at what point in time to execute, many traders do not necessarily want to have to consider too many other factors – it may be counter productive. As a result, we aim to ensure that our algorithms are optimised to deliver the best performance without any additional input from the end user. For many traders this approach works perfectly well. Customising to order Perhaps the first step towards customisation happens when a trader decides to adjust one of the parameters available with each tactic – typically, start and end times, price limit, aggression level, min and max percentage volumes. The ability to tweak the parameters means there is significant scope for customising each algorithm on an order by order basis, even to the extent that different tactics can be forced to perform like others, or combinations of others. For example: ■ A trader who wanted to trade ‘volume in line 20%’ but didn’t want the tactic to rigidly stick to the 20% target irrespective of price opportunities, might instead use ‘price in line with a 15% min and 25% max’ – which ■ THE TRADE 2005

■ Chapter 6 Honing an algorithmic trading strategy

would then aim to be on average 20% participation, but could speed up or slow down within the 15-25% range to respond to pricing fluctuations. The further step towards customisation is when a trader finds that his personal preference is leading him to consistently use the same tactic with the same parameter adjustments for certain sectors or markets. In this instance a request can be made to adjust the default settings for that client so that the revised parameters are always used. The revised tactic can be re-named if the client also wishes to continue using the default version. Examples of this are:

start of the trading period and less aggressive towards the end. Beyond such requests, clients also approach CSFB with ideas for custom strategies, usually variations on the menu tactics we provide, examples of which it would be inappropriate for us to reveal as they offer real competitive advantage to the client. These ideas represent the desire of traders to further automate their own trading style; in effect when they come to us with these requests they have developed their own strategy and are simply asking CSFB to put them into practice. For example: ■ When the stock gaps I like to…

■ A trader who always wants to finish by 2pm GMT ahead of US opening. All selected tactics are defaulted to finish at that time.

■ Cross asset correlations – when trading mining stocks I want participation curves that respond dynamically to commodity prices…

■ A trader in French mid-caps who is less interested in potential impact and more interested in grabbing available liquidity might ask for the TEX strategy to default to very aggressive mode for these stocks. ■ A trader who trades VWAP in the morning period but wants the curve skewed to be more aggressive/overweight at the

■ I have buy and sell baskets, I want to maintain any natural hedges as it progresses and keep both sides dollar neutral…

■ ALGORITHMIC TRADING

■

“Customisation has been around from the start. Indeed, it is hard to see how the product could have worked had it not.” ■ A BUY-SIDE HANDBOOK

■ THE TRADE 2005

61

■ Chapter 6 Honing an algorithmic trading strategy

One other area of customised automation offered on AES™ is our ‘Storyboard’ product. Storyboard automatically sends clients messages triggered by price movements, news, volume spikes etc. in the stocks they are trading in AES. Here again, all the trigger limits are configurable by the client.

62

FAQ Customisation has been around from the start. Indeed, it is hard to see how the product could have worked had it not. There is a view, inaccurate in our opinion, that broker algorithms are canned and therefore inflexible. Hopefully, the examples that have been outlined prove otherwise. There are also views expressed that all broker algorithms deliver the same performance, that they are ‘commoditised’. We have not been presented with evidence that shows this to be the case, but it is understandable how this viewpoint might add weight to the argument that the only valuable algorithm is a customised algorithm. In our experience, even using the ‘plain vanilla’ versions of our algorithms, different clients achieve different results – from good to excellent! As with all aspects of the buyside/sell-side relationship, be it research, trading or the development and use of trading ■ ALGORITHMIC TRADING

■ A BUY-SIDE HANDBOOK

algorithms, the buy-side needs to assess the merits or otherwise of insourcing versus outsourcing. Hopefully, this chapter gives those who have yet to adopt algorithms a better understanding of the current scope of customisation and flexibility that is already available. ■

■

“There is a view, inaccurate in our opinion, that broker algorithms are canned and therefore inflexible.”

■ THE TRADE 2005