Algorithmic trading

From Wikipedia, the free encyclopedia

Algorithmic trading is the use of computer programs to trade financial instruments (e.g., stocks, bonds, etc.) in electronic markets, usually for and on behalf of a customer. This is mostly useful only in markets that offer some sort of electronic execution and have a public, electronic market.

Changing market structure, a focus on transaction costs by investment managers and pressure to reduce expenses within the industry as a whole accelerated the adoption of this automation technique in the world's financial markets.

Contents

[edit] History

Beginning in 2000, the US markets changed their minimum tick size from 1/16th of a dollar to 1/100th of a dollar (a single penny), in a process called decimalization. Therefore, stock exchanges and Electronic Communication Networks could hold limit orders at prices in increments of 1 penny. Prior to this the minimum tick size in most circumstances was 1/16th of a dollar and in order to improve the best bid or offer of one of these public limit orders a new order would have to have been entered 1/16th better (lower for sell orders, higher for buy orders). The effects of this on order book and market micro-structure were profound and far-reaching. This effectively reduced the profit in a stock that market makers could charge customers seeking immediate execution, as market makers in a stock typically make a profit proportional to the difference between the highest (best) bid and lowest (best) offer prices for that stock. This drove some dealers out of business, and encouraged many to incorporate a "fee-based" (or agency) business model. Another effect of decimalization was decreased apparent liquidity and more volatile prices. This is because it now cost 1 penny to step in front of another order, whereas before it cost 1/16 of a dollar (6.25 cents). The reduced liquidity and slightly more volatile pricing made trading large numbers of shares more difficult. The solution was trade in smaller increments over a longer time. This change in market microstructure was countered by innovation in automation, with the initial concepts and techniques growing out of prior experience in index arbitrage and statistical arbitrage trading. Thus, algorithmic trading went mainstream.

[edit] Trading Strategies

When using a trading algorithm, it's common to select a strategy or objective for the algorithm. The typical example problem is optimizing large institutional trades; for example, how to break up a large volume of shares to sell over time, such that the price obtained will not be greatly affected by the increased supply or demand of the order into the market place. Some of these strategies are as follows (please be aware that these descriptions are non-technical and only meant as an introduction)

[edit] Trend following

Main article: Trend following

Trend following is a strategy which tries to identify a trend in the price for a certain security, trade with that trend, and exit when the trend appears to be failing. This may be achieved by the use of a range of different tools such as trendlines, regression, and moving averages.

[edit] Contrarian (reverse)

Contrarian strategies involve entering a position against the current trend, having determined through indirect means that the trend may be going to reverse, or at least undergo a correction. The evidence for a contrarian trade may include divergence of a range of momentum indicators, which is associated with a weakening trend. Overbought/oversold indications from a range of indicators may also support a contrarian trade.

[edit] Benchmark

The most common benchmarks are VWAP and recently, decision price or arrival price (Implementation shortfall). These strategies generally break up the order into smaller pieces, ideally reducing market impact whilst achieving the desired benchmark as they trade. Which benchmark you should use largely depends on how big the order is and what your trading goals are. VWAP is still the global favourite but implementation shortfall is gaining ground rapidly, as it generally provides a better match for the decision-making process.

[edit] Order book

This is a relatively new class of trading strategies that firms developed after NASDAQ made full order book information available. As the field progressed, financial professionals have proposed increasingly complex strategies. Some of these are based on methods developed in fields such as: econometrics, financial mathematics, stochastics, artificial intelligence, and multi-agent systems.

[edit] Purpose

"Algo" trading is used to achieve two different ends. The first is to generate new orders and the other is to improve the execution efficiency of existing orders (which may have been based on unrelated strategies). For example, one can use trendlines, regression etc. to create a new order for a set of securities while using Benchmark to break down an existing order into smaller more efficient chunks so that they can be better executed.

[edit] See also

[edit] External links

[edit] Software