This article was published in August 2007 in SFO Magazine (now defunct). The original published article has been lost with the shutdown of SFO Mag but the text below is the original unedited article I submitted for publication.
Introduction
Most readers have probably seen at least one of the movies in the Arnold Schwarzenegger trilogy of Terminator movies. These movies have machines fighting machines with humans slightly in control of one side. These days, it feels as if trading the markets is just like that – machines against machines with humans barely in control. It’s this feeling and conversations with a number of experienced traders about the radical changes they’ve observed in price action over the last three years that formed the genesis of this article.
Anyone who has been trading for a while will probably, in the course of an extended price action discussion, mention how quickly and radically price moves occur now. Of course, there have always been quick and radical price movements in the financial markets. However, today, it seems that “quick and radical” is the norm rather than the exception. There are a number of reasons for this change:
- All market participants are more aware of technical levels. Everyone sees the same levels and plays for or against those levels. Everyone sees the same patterns and thus dog-pile in and/or out of the instrument as the break from the pattern unfolds.
- Information flows much more rapidly now. This means that the effects caused by new information is assimilated and acted upon by market participants in increasingly shorter timeframes.
- The has been a exponential increase in the number and power of automated trading systems (‘black boxes”)
- The associated radical increase in the number and power of trade execution algorithms (different from black-boxes).
- The increase in the number of market participants in any given market due to globalization/liberalization of markets, exchanges and money flows.
- The increase the amount of capital available thanks to the easy money policies of many central banks.
In this article I’m going to touch on the effects of all of these changes but focusing primarily on the impact of algorithms on trading styles. I will not profess expertise in this subject – I am simply writing as a market participant/observer that has felt the impact of these changes. I believe that that there is value if a trader can understand the reasons and drivers behind some of these changes – it can provide insight into adapting, identifying and taking advantage of new opportunities in the market.
Definitions
Before continuing, I must make a distinction between “black boxes” and “algorithmic” trading. When reading articles related to computerized trading I noticed that there seemed to be a distinction by some authors regarding these two terms. Frankly, I was surprised because instinctively I thought computerized trading could all be categorized as “black box” trading. But, for the purposes of this article I’m going to assume that distinction, using the terms “algorithm” and “black-box” to mean different things (albeit, they are related).
The term “algorithm” or “algorithmic trading” will be used primarily to refer to trade execution only algorithms. In algorithmic trading the job of the algorithm is to get the best price over a particular period of time. It’s not trying to determine trend nor is it a product that tries to obtain a profit based on buying and selling. In other words, its job is to buy 100,000 shares of ABC and get the best AVERAGE price possible over the course of an hour, 15 minutes or the day. The decision to buy (or sell) is someone (or something) else’s. The execution of that decision is the job of the algorithm.
Any algorithm that is not execution only and involve decision making (trend, entries, exits, profits and losses) we can categorize as Black-boxes – their job is to try to turn a profit on a strictly mechanical basis. This means determining what the trend is, when to buy or sell, what the stop level is and what the exit levels should be.
Please note that I am only making the distinction between the terms for the purposes of this article. I am not sure I agree or that everyone agrees that there should be a distinction between the two terms. But, for my limited purposes for this article, the distinction works.
The effects of algorithms
The job of the market is to bring demand and supply into balance and, in the process, determine the “market” price of a commodity or stock. One of the consequences of this function is that price moves from one level to another. Prior to around 2002 this price movement tended to be more orderly than not. A rising market would be filled with retracements, zigging and zagging towards its new price level. Traders could count on those retracements and make a good living playing them. That is no longer the case. Today, the movement between price levels (or, more accurately, value areas) tend to be more abrupt. There are fewer retracements and the time it takes to move from one level to another has become shorter and continues to compress. In other words, the market has become more “efficient”.
There are multiple of reasons, related to system and algorithmic trading for this:
- In the equities market more and more of the sell side institutions (i.e.: brokerage houses) use algorithmic trading to process client orders.
- More and more of the buy side institutions are internalizing their execution using cheap computing power and proprietary or semi-proprietary algorithms.
- In all markets, there are more “index” funds or ETFs where the job of the fund or ETF manager is to match the overall performance of an index basket.
- Computers are needed simply to find sources of liquidity, especially in equities. There has been a large increase in the number of venues where the same equity can be traded – it is almost impossible for a trader to efficiently hunt for and place orders at all those sources.
These changes have the potential to change the game for every player in the market permanently.
An Algorithm for Everyone
In the “old” days there were two distinct roles on Wall Street – those that bought (or shorted) stuff and those that brokered the transaction. Respectively these two camps are known as the “buy side” and the “sell side”. The buy side is comprised primarily of fund managers (and other brokerage customers); the sell side is comprised primarily of brokerage firms. In those “old” days, the buy side would give an order to the sell side (brokerage) who would fill that order using human ingenuity. The sell side is a competitive business however. So, as electronic trading became more prevalent, the sell side started to use simple algorithms to more efficiently fill orders and to provide an edge. This allowed them to guarantee to their clients (the buy side) a better fill price for large orders in order to get the business. Obviously, all sell side firms had to get into the algorithmic business or they would go the way of the buggy whip. Thus was laid the foundations for the first wave of “robot” battles in the industry.
The second wave was laid by the explosion of cheap computing power and the continued rise of electronic execution. This made more intelligent algorithms possible and made software that was easy for buy side firms to use possible. Thus, today, a buy side firm can buy or lease software with pre-built algorithms for order execution. The software can easily be customized to use a proprietary fill algorithm as necessary. Suddenly, instead of humans competing against each other, you have super-fast buy-side algorithms competing for fills. The sell side firms have been mostly relegated to supplying clearing, customer service and brokerage services at rock-bottom rates. Some have morphed into technology vendors of sorts – providing algorithms, connectivity and technical know-how.
Overall though, it has become a free for all. Bots battling bots to try to get the best fill.
The Battles
In order to understand what kinds of battles are being waged by these order fulfillment algorithms we have to get into the psyche of index fund and ETF managers. Many of these funds are indexed to the overall performance of the market or a segment of the market. Thus, a manager has done his/her job if they simply meet the performance of the segment that they are indexed against. So, what does this mean in practical terms? It means that if the market or their market segment is going up, they have to buy – the price at which they buy is not necessarily a consideration as long as their final price is an acceptable average over some time period. As long as this average is similar to the average that other fund managers will obtain everything is ok. Let’s understand this because it’s important: Usually, the final price is not a consideration – it is the average price obtained over a time period compared to the average trading price over that time period that counts. In fact, one of the biggest growth areas for some of the technology vendors right now relates to products that measure and monitors the effectiveness of these algorithms.
For many of these managers it is simply not acceptable to miss the trade – if their indexed market is going up, they have to be in. Thus, many execution algorithms are designed to ultimately execute their buys and sells regardless of price. Of course, these algorithms will try to get better prices but if the market starts to move then they will simply execute at the market because they have to get the orders filled. This adds fuel to the fire as each algorithm successively starts to submit what are effectively market orders. The net effect are markets that flood in one direction with no let-up. Spikes are more frequent and more pronounced with prices returning back to original pre-spike levels just as rapidly as they moved once the algorithms are done with their games.
What about hedge funds and other actively traded funds? It’s no secret that there has been an explosion of these players. Many of these players are trend followers. Thus, if it looks like a trend is about to begin you can be sure that there are going to be quite a large number of players that are going to want in on that trend. This means that more money than ever is poured into (or taken out of) a market in a much smaller period of time. It means that markets can be easily driven far above or way below their true value more often and can stay at those levels for far longer than before because there is ever more money to sustain the trend beyond “rational”.
The Effects
The effect of all this can easily be seen in today’s markets. What starts off as a small creeper mode quickly turns into a crescendo and then everything just peters our. There are no retracements – price moves up (or down) in one shot and then goes sideways for what seems like forever. There are more v-spike reversals than ever before. It’s a big, fast game of musical chairs – everyone dog piles in and then dog piles out. If you’re not quick enough then you’re left holding the bag.
Another major side effect is the lack of respect for support or resistance points. Major violations of support and resistance points are common now. I firmly believe that this is related in no small way to algorithmic execution. An algorithm that needs to buy will buy – it has no knowledge of support and resistance zone. Its primary purpose is to get the best average price – even if that average price drives the market above what looks like resistance (or below what looks like support).
Further, there are major effects on stop sizes. Because these algorithms introduce lots of volatility into markets, your stop levels have to be much further away than ever before. This means that, for the same account size, you have to play with reduced size. It used to be that 2 ATRs could be considered “outside the noise.” Now, it’s 3 ATRs and, in many cases 3.5 ATRs. Ouch if you take a 3 ATR stop on the same unit size you were trading 4 years ago!
Not all the news is bad. The volatility introduced by these algorithms (and other changes) means that a trader can day-trade more markets than ever. The really old hands will remember when corn could not be day traded, when the range was less than 5 cents. Or Sugar. Or even gold. Now, the length of line is so much larger and provides lots of opportunities just “playing the noise” if a trader so desired.
Trading Techniques
Probably the best trading technique that will work in this environment are breakouts. You’ll have more false breakouts but the risk/reward ratios will more than make up for those. And, you will be guaranteed to be in on every move. It will just be far more frustrating, especially if your normal bread and butter trade is retracements after a breakout.
In order to make up for the numerous small losses, traders may have to get used to pyramiding when the break is finally for real.
If a trader wants to play the retracement game then the trader will be forced to look for retracements on a much lower timeframe. If the breakout is from a 5 minute formation the first 5 minute retracement may not occur until the move is almost complete. Instead, the trader will have no choice but to look for retracements on charts as low as 30 seconds or 1 minute.
Tools
Since markets move so much more rapidly today it has become harder for the individual discretionary trader to watch and manage the same number of markets and positions as they did just a few years ago. New tools are needed and software vendors are rising to the challenge. Prop firms are rolling out their own algorithms to help with trade execution. Trading platforms are introducing more robust “chart” trading so that the user will have not have to turn away from the screen in order to place and manage trades. New semi-automated software products are increasingly available to make it easier to manage one side of a trade.
For example, the discretionary trader could enter a position manually and then let the software manage the exit using a trader defined algorithm such as a parabolic, fixed ATR, keltner bands or combinations. This frees the trader from monitoring the position allowing her to focus on finding more opportunities. Even black-box building tools are getting easier to use increasing the number of potential traders than can use them.
Summary
Probably the only guarantee you can ever have in the trading game is that it will change. Over the last three or four years there have been lots of changes and the effects on the markets have been pronounced. Once a trader understands the drivers behind these changes, it’s easier to react and take advantage of them. Algorithms and blackbox trading is here to stay. Their use will grow – some estimates suggest that 50% of equity trading will be done using algorithms by 2010 up from 25% today . But, as with anything else in the markets, the whole scale attempts to use software to extract maximum value from the markets will themselves create numerous pockets of inefficiencies that can be exploited. The smart, adaptable individual trader will be able to seek out and exploit these new pockets. So, I say “Long live Algorithms”!
Follow Me!