Measuring market quality

Transcription

1 A Cinnober white paper Measuring market quality Lars-Ivar Sellberg, Cinnober Financial Technology AB Fredrik Henrikson, Scila AB 11 October 2011

2 Copyright 2011 Cinnober Financial Technology AB. All rights reserved. Cinnober Financial Technology AB reserves the right to make changes to the information contained herein without prior notice. No part of this document may be reproduced, copied, published, transmitted, or sold in any form or by any means without the expressed written permission of Cinnober Financial Technology AB. Cinnober and TRADExpress are trademarks or registered trademarks of Cinnober Financial Technology AB in Sweden and other countries. Other product or company names mentioned herein may be the trademarks of their respective owners. Cinnober Financial Technology AB 2

3 New trading patterns demand more from market quality analysis Today s equities markets are increasingly fragmented, and the competition for liquidity among trading venues is fierce. At the same time, algorithmic trading contributes to a growing share of trading volumes and marketplaces earnings, both in equities and other instruments. The potential negative effects of certain types of algorithmic trading are widely discussed in media, among market participants, regulators and other stakeholders. Does it increase market volatility, amplify short-term market reactions and in the long run cause traditional investors to pull out from these markets? This is still a rather new phenomenon and there are still a lot of unanswered questions. This document does not go into the details of the various types of algorithms that are deployed by traders. However, for any market stakeholder, including the trading venue itself, it is vital to understand the quality of the market and how it is impacted by algorithmic trading. It is also clear that this demands more sophisticated analysis and tools today than most have been used to. Market liquidity is often measured using the following three criteria (Bervas, 2006): The tightness of the bid-ask spread. The bid-ask spread gives an indication of the cost of immediate reversal of a position of standard amount. Market resilience. Market resilience measures the speed prices revert to equilibrium price after a large order has been executed temporarily moving the spread. When measuring market quality it is important to look at all three aspects above. For example, a simple spread measure consisting of best bid/ask price does not say much about the liquidity available at a given trading venue. Traditional order rate measurements alone don t reveal the extent to which the orders contribute to liquidity in reality. One of the key functional areas of Scila Surveillance a market surveillance system is its detailed analysis of market quality. This is not market surveillance in its traditional sense, but still of outmost importance to improve understanding of the market. The two main areas covered by this paper are: Measuring liquidity Understanding where algorithmic trading occurs and what impact is has Market depth. Market depth is the amount of volume one can buy or sell without moving the best price (i.e. without slippage). Cinnober Financial Technology AB 3

4 Measuring liquidity As competition among trading venues has increased, so has the number of methods devised to attract liquidity. Examples include: To fine-tune the parameters that define the trading venue s market model, such as the level of market transparency, counterparty information, tick-sizes, trading method etc. Trading tariffs schemas have grown increasingly complex in order to attract liquidity. Besides the actual price list, tariff schemas include more subtle measures such as giving certain order types low latency links or other advantages to selected liquidity providers. Marketing activities Equity participation schemes While there are many parameters to adjust, the actual impact from these adjustments is hard to predict since it varies among different markets and situations. If the impact from different activities isn t measured properly, market operators will end up changing parameters without understanding their true effects. In order to get an accurate picture of the market quality of a marketplace, it is necessary to examine a number of different aspects in more detail: Liquidity measures where the spread is adjusted for a fictively traded volume, e.g. showing the cost of trading 100, 1,000, 10,000 units, etc. This is especially important in fragmented markets where spreads have tightened while volumes on the best price levels have become thinner. This renders standard top-of-the-book spread measurements more or less useless. Liquidity measures as described above, also adjusted for time. Replenishment times after liquidity events not to be confused with events based on informed decisions. That is, if a large order that temporarily moves the spread is executed, how long does it take for market participants to insert new volumes that re-establish the original spread? This gives a measurement of the amount of latent liquidity in the market. Understanding the liquidity replenishment time might also constitute an important factor when determining optimal lengths of circuit breakerinitiated auctions. Average order lifetime. The contribution to liquidity by passive orders with a lifespan measured in milliseconds can be questioned, or at least needs to be evaluated in a different way than orders with longer life spans. Cinnober Financial Technology AB 4

5 Understanding the impact of algorithmic trading The previous section described ways to measure liquidity in a market. This section introduces means to understand how algorithmic trading contributes to liquidity. Scila Surveillance utilizes several methods to detect the occurrence of algorithmic trading in order to understand its impact on the market. Algorithmic trading systems utilize highly sophisticated computer programs to analyze market data based on advanced mathematic models to generate trading signals. The system can be designed to make own trading decisions or to just optimize execution of already made decisions, or both. The term algorithmic trading covers a wide range of trading strategies, some well known while others for natural reasons are non-public and carefully guarded secrets. Many types of algorithmic trading are beneficial and add to liquidity, but there are also harmful variants that even reduce liquidity. An example might be the presence of snipers in a price driven market that in the long run might drive off market makers by systematically taking advantage of weaknesses in their applications. Identifying which order books are likely to attract algorithmic trading First, the trading venue needs to understand which order books are potential candidates for hosting algorithmic trading. Two methods for this are: Measuring predictability. Not all order books are suitable for algorithmic trading. Alpha-creating algorithms 1 aim at consistently generating income over a large number of trades. To achieve this, the system tries to predict future prices, for example with the help of technical analyse s. An order book with a completely random price process is therefore of little interest. To determine if the prices for a given security appear to be random, or seem to have a level of predictability, Scila Surveillance uses a method developed by Lo and MacKinlay (1988). While a price process may appear random when seen over a longer time span, this might not be the case over shorter time spans. A strategy that uses a short position-holding time period may thus find predictability in a security, while a strategy with a longer positionholding time period finds it to be random. 1 Algorithmic trading can be divided into two main categories: Alpha-preserving and Alpha-creating. Alpha-preserving algorithms are used when an investor has through fundamental analysis or in some other manner reached a trading decision and is about to execute it. Typically alpha-preserving algorithms deal with reducing market impact, thus minimizing slippage when executing the chosen trading strategy. A trivial example is breaking up a large order into smaller chunks executed over time. Alpha-creating algorithms, on the other hand, include a diverse set of algorithms that by themselves try to create alpha. Cinnober Financial Technology AB 5

6 In Scila Surveillance, randomness can therefore be sought out at several time intervals to determine which securities a certain type of algorithmic trader is most likely to trade in. Measuring information asymmetry. Asymmetric information in a market leads to adverse selection and the possibility for more informed market participants to make money at the expense of less informed participants. Informed traders possess private information that isn t available to the general public and enables predictions of future price movements and spotting of market inefficiencies. This information could, for example, come from superior forecasts by advanced algorithms or superior news sources and can have a significant impact on the market. There are several ways to detect information asymmetry in markets. Scila Surveillance uses a method based on the measurement of the Probability of Informed Trading that was developed by Easley et al. (1996). Identifying algorithmic trading Once it has been established that algorithmic trading is likely to occur in an order book, the next step is to identify the parts of the order flow that come from algorithmic trading systems. Some methods used in Scila Surveillance for doing this are: Measuring Order Aggressiveness. It can be assumed that alpha-creating algorithmic traders possess superior information. It is also the case that in the presence of more than one informed trader, the informed traders are expected to trade more aggressively, according to Back, Cao and Willard (2000). Thus algorithmic trading can be detected and categorized by measuring how often a trader is on the aggressive side of trades. Measuring position-keeping. Analyzing size, sign (long/short) and the duration of positions can be used to both detect and categorize certain types of algorithmic trading. Detecting order flow anomalies. A way to detect informed order flow is to analyze it from a technical perspective, not directly related to the algorithms used by the trader. One example is the habit of some algorithmic trading systems to send in multiple cancels for the same order, in order to reduce impact from the inherent jitter (variation in latency) of the marketplace system. A similar example is when the algorithmic trading system submits multiple orders all marked as possible duplicate. These types of behavior are less likely to be implemented in standard systems used for the routing of retail, uninformed order flows and therefore work as identifiers of informed flows. Cinnober Financial Technology AB 6

7 Summary There is no one-size-fits-all solution for how to measure market quality. As trading patterns change and trading techniques become increasingly sophisticated, the tools used to analyze market quality must also be upgraded. Only then can the trading venue and other stakeholders better understand their market. Various types of algorithmic traders are today a growing and important client group for trading venues. At the same time, they are questioned by many stakeholders in the industry due to the potential negative effects they have on the market. It is therefore critical for any trading venue in today s highly competitive markets to learn more about the order flows from these traders, and their impact on the market. For a more thorough discussion around some of the algorithms used in Scila Surveillance, see Fredrik Henrikson s Characteristics of high-frequency trading at: Cinnober Financial Technology AB 7

8 References Aldridge, I. (2009). High Frequency Trading: A Practical Guide to Algorithmic Strategies and Trading Systems. Wiley Trading, John Wiley & Sons. Back, K., Cao, C. H., & Willard, G. A. (2000). Imperfect Competition among Informed Traders. The Journal of Finance, 5, Bervas, A. (2006). Market Liquidity and Its Incorporation into Risk Management. Financial Stability Review 8, Easley, D. K. (1996). Liquidity, Information, and Infrequently Traded Stocks. Journal of Finance, Lo, A. W., & MacKinlay, A. C. (1988). Stock Market Prices Do Not Follow Random Walks: Evidence From a Simple Specification Test. Review of Financial Studies, 1, Cinnober Financial Technology AB 8

9 Cinnober Financial Technology AB 9

10 Passion for change Cinnober provides mission-critical solutions to the world s most demanding financial marketplaces. We are passionate about one thing: applying advanced financial technology to help trading and clearing venues seize new opportunities in times of change. We build partnerships with our customers based on trust and transparency. We serve investment banks, exchanges, clearinghouses and other actors that have extreme demands on business functionality, high throughput and low latency. We currently have product-based offerings for a number of areas, such as marketplaces, post-trade management and binary markets. All solutions use our TRADExpress technology, designed for scalability and flexibility. Since our start over ten years ago, we have established ourselves as a leading provider of innovative solutions to premier financial institutions around the world. These include mission-critical systems for leading exchanges such as the Chicago Board Options Exchange, Deutsche Börse, London Metal Exchange and NYSE Liffe. We also power new initiatives and alternative trading systems such as Alpha Trading Systems, Burgundy and Markit BOAT. We are an independent provider of marketplace solutions, and do not operate a market of our own, avoiding any conflicts of interest. Our track record says it all. We help our customers turn change into a competitive advantage. Cinnober Financial Technology AB Kungsgatan 36 SE Stockholm Sweden cinnober.com Cinnober Financial Technology AB 10