Stock prices,financial markets, stock market crash, oil stocks, trading programs, profitable strategies, Technical trading, market technician, successful traders, expert systems, individual trading, institutional trading,buy and sell signals,pattern recognition, trading method, trend-following method,developing a trading strategy, head-and-shoulders formation

free download links about online stock trading, forex, futures, stock investing, market, trading systems
Reasonable Expectations Give Achievable Results

The combination of technology, optimism, and the success of some highly visible traders can easily lead to unrealistic expectations. We see the profits before the risks, then we attack and eliminate the risks with out questioning the validity of the profits. When developing a trading strategy, the failure to understand what is most likely to happen can waste time and effort, as well as cost a great deal of money.

Every profit opportunity has risk. Larger profits have larger risk. Each systematic approach to trading has its own risk and reward trade-offs. And a trade-off is always an unpleasant compromise. It is a mistake to think that you can find a trading strategy that has no losses, or an arbitrage that will absorb unlimited funds.

A pure trend-following method has smaller losses and larger profits. It is classified as a "conservation of capital" approach. To keep losses small, it is necessary to close out trades quickly. The system continual ly tries to fmd a trend but exits as soon as prices move in the wrong direction. Therefore, there are more losing trades than winning ones. If you increase your tolerance for risk, by using a slower trend or larger stop-loss, the system will have a larger percentage of winning trades but larger losses and equity swings. Make the trend very slow and the stop-loss far away and you have a passive portfolio of one open trade.

Countertrend strategies typically allow more risk to achieve frequent, small profits. When the strategy errs, there are large losses. Its trade-offs have the same, continuous relationship as a trend-following method, but in reverse. As the targeted profits get smaller, they are more fre quent. In exchange, the fewer losses get larger.

Trend and countertrend traders have similar choices: smaller losses or smaller profits more often, offset by larger profits or smaller losses less often.

Commodity arbitrageurs compete for "riskless" opportunities, such as a location arbitrage. Profits can be made when the differential between the same product selling in separate locations is greater than the cost of transportation, insurance, and other carrying charges. Competition can be so keen that the arbitrageur accepts both small profits arid small vol ume, a situation that, at some minimum, is not worth the effort.

There is no secret way to produce constant trading profits. Some methods are better than others, but none of them are immune from these trade-offs. It is important to identify and understand the alterna tives before making a final choice (see Table 3-1).

We all know that there is higher risk with higher profits and that the only system that doesn't have losses is the one that doesn't trade. When you sit in front of a powerful computer with sophisticated tools, however, you tend to forget that these limitations still apply.

Profits seems to stand out when you look at the trading results of hun dreds of strategies. The high leverage of options and futures pushes many of the test returns well above 100 percent per year. These are exceptionally high compared with most investment returns and can be intoxicating at first. But presented in a more traditional risk-adjusted form, with adequate capitalization to insure safety, results are scaled

down to a more sensible range of 5 percent to 25 percent annually. This is discussed in Chapters 4 and 11. You may discover that the risk-adjust ed returns of a very profitable trading method are no better than a con servative bond fund.

Developing a trading method is hard work. The main idea can be the result of years of watching market patterns, or the formation of precise mathematical relationships. The process of testing and verifying is also difficult. Each calculation needs to be checked and entry and exit prices must be reviewed. It can be so tedious that, at some point, you would willingly accept good test results and stop looking for errors. You begin to look at larger individual losses without also reviewing the larger profits. But errors cause both good and bad results, and a program is not valid until it is error free.

A trading strategy that returns more than 20 percent annually, with a reward/risk ratio greater than 3.0, is an enviable achievement. If you add a stop-loss to a system to reduce its risk, you find that the returns also drop; or, by selecting specific trades with better opportunities, you reduce the frequency of trading so that each trade must generate larger profits to reach the original rate of return. Each day that you are out of the market, waiting for an entry signal, reduces the rate of return.

In the end, all systems must conform to the expected trade-offs. If they don't, you have reason to be suspicious. A system with no losses, a reward /risk ratio over 5.0, or annualized returns over 50 percent for 10 years must have compensating limitations for these benefits. You can not accept their results at face value; you must find and understand their problems. There is no room for careless optimism.

Computers excel in the manipulation of data. Great advances in technology are the result of the ability to consolidate massive amounts of information and produce solutions. What better application than solv ing a price-forecasting problem?

It is a simple task to input fundamental and economic data necessary for the interpretation of stock prices. Or information about supply and demand, which are the primary factors affecting the price of food, energy, and other commodities. It is also possible to add to the database less obvious influences (e.g. financial data, such as money supply, unemployment, and per capita income) that when combined with other events, will cause prices to shift.

The computer can be given sophisticated tools for finding relationships in the data that explain price movement. The most popular has been a form of multiple regression—finding constant relationships over time between numerous factors. More recently, neural networks are replacing regression analysis as the favorite method. For example, when crude oil stocks decline more than 10 percent below the average, prices rise. And, when OPEC calls an emergency meeting, prices rise. When the two happen at the same time, the prices rise higher. Fundamentals provide a reasonable, classic approach to predicting price movement.

In the past, the speed limitations of computers curtailed the amount of data that could be matched against each other. Calculations were lengthy and the results showed a clear relationship, but large variability. Because not all the price movement could be "explained," the predictive quality also had uncertainty. Newer, faster computers can analyze much more data. They should be able to reduce the variance and improve fore casting. Entering as much data as possible and letting the computer find the relationships is a simple extension of the same problem.

But the solution usually is not improved. Many choices of statistics, in combination, can "explain" the price movement. Which combination is the right one? Or are any of them correct? With enough data, countless patterns are produced. The power of the computer can find them all, without having any way to identify the right ones.

A common use of computer analysis is pattern recognition. In an attempt to find relationships not yet clear to others, the computer can scan, for example, unemployment data and find that a drop greater than 3 per cent in November was followed by a rally of at least 4 percent in housing starts during the following May, for all of the 18 years available.

Is that a perfect solution or a coincidence? It depends on how much data the computer scanned. If you started with the premise that unem ployment affects housing, and that more people working means more home sales, then you needed to scan only those two data series to confirm the relationship that you already believed to be true.

If you began with the entire database of U.S. statistics spanning the past 50 years, looking for a perfect relationship, this would have been one of many that you found. Another one may have been the infestation of gypsy moths relating perfectly to the import of Chilean grapes. Employment and housing sound right, but how do you know it is any more significant than moths and grapes? You can only know if you applied a logical argument, determined before testing, then validated the theory using the computer. Because we never would have thought of a trading program using gypsy moths, even the most astounding result must be ignored. Box 3-1 shows how to calculate when a rela tionship is "significant."

Big companies have more choices than individuals. When they need to solve a more complicated problem, they can buy a bigger computer and get more crunching power. They also get a very impressive computer center and a higher electric bill.

Some solutions require long calculations. Finding meaningful rela tionships among thousands of data series can require hours of comput ing time even on large machines. Yet, figuring out the navigational sequences and landing instructions for putting a space team on the moon takes a small computer only a few minutes. Why is the difference so great? The computer that guides explorers to the moon has both objectives and well-defined physical relationships. The effects of plane tary motion and gravity are precise, although intricate. Specify the time and everything else can be solved.

Although forecasting a price can be stated as a clear objective, the solution is not clear. The answer may not even be hidden in all the data stored in the computer. The method used to scrutinize the information may not be the way the market works.

You do not necessarily get a better solution by using more computer power, you just get a faster one. Power can overwhelm reason. No mat ter how big or fast, the computer only solves problems the way it is told. It does not claim to solve a problem that has no answer.

Indicators have been given credit for correctly signaling major price moves in stocks and financial markets. On-balance volume, short-inter est, customer margin debt, "insider" trading, percentage of cash hold ings of mutual funds, oscillators, contrary opinion, and countless others claim accurate forecasting ability. It is true that they have produced buy and sell signals at key points. But at other times, they signal the wrong moves or no move at all.

A indicator is an optimistic, oversimplification of a solution. It is an attempt to find a single, orderly solution to a complex problem.

You would think that finding a perfect pattern over 18 years is an impressive discovery. But not if the computer finds it by scanning a large database. For example, with 5 years of data, there are 2 A 5 = 32 possible up-down patterns:

If you scan 33 series of 5-year up-down patterns, two of those series must be identical although that doesn't prove they have anything in common. There is only a 1 in 32 chance that any two series will be iden tical. Therefore, if you were to take only two series that you thought depended on one another, and they proved to be identical, you would succeed with a 3.1 percent probability. If you were able to match 6 years, there would be a 1 in 64 chance, or 1.5 percent.

Statistics say that there has to be less than a 5 percent chance of occur rence before an event is called probably significant. But thaf s of minimum

Combining multiple indicators does not seem to eliminate the bad signals, as you would want. Instead, the new composite indicator has the same erratic properties as its components.

This disappointing inconsistency is not the fault of the indicator, but of the user. Indicators are constructed to emphasize a particular market feature. On-balance volume will confirm a trend if there is one; stochas- tics and contrary opinion will show when a market is overbought or oversold, but not whether it's a good time to buy or sell.

A reliable strategy is one that uses relevant information or specific indi cators only when they are important. At other times, these indicators have no value; forcing them to say something all the time is a misdirected approach to analysis. The market does not always have something to say.

Standard econometric analysis, often based on multiple regression analysis, tries to form continuous relations between the data. That requires all data to interact in the same way throughout history, even interest. It needs to have less than a 1 percent chance (1/100) of occur ring to be called significant, and less than .1 percent (1/1000) to be high ly significant.

Finding the importance of a pattern of results requires only a few simple steps:

Step 1: How many different combinations are there in a series of 18 ups and

the series. Then there are 2 X 2 X 2 x ... x 2 (18 times), or 262,144 total combinations.

Step 2: What are the chances of having two identical series? If you entered one more than the total combinations, or 262,145, series of 18 years into the computer, one of those series must be a duplicate. When half the data (131,072 series) have been tested, there is a 50 per cent chance that one of them is a duplicate.

Step 3: When is there a "significant" 1 percent chance? When you compare only two data series of 18 items, there is only a 1 in 262,144 chance of them being the same (which is very small). There is a 1 percent chance of finding two identical patterns in 2,622 series. Selected at random, statistics states that it is not likely to be coinci dence.

when markets are being driven by different forces, or by nothing at all. That technique is similar to forcing a relationship. This problem can now be avoided by using a neural network for threshold analysis to isolate dif ferent market scenarios. A single event, or combination of events, will trigger the use of a selected set of information applying only to that situ ation. For example, an unusually large jump in inflation will turn the focus to interest rates, and begin monitoring the Producer Price Index, the Goldman Sachs Commodity Index, the yield curve, and the U.S. dollar, looking for confirmation. More about this can be found in Chapter 9.

An interesting feature common to price charts and system testing is that many of the patterns depend on the length of the chart. You cannot find a head-and-shoulders formation with only one week of daily data, or a major bull market with one year. Performance expectations are also dis torted by using too little data, or selected data. It is not necessarily the profile of the profitable trades that change as you look at more price examples, but the risk. A sustained sideways market and an exception ally volatile period will create prolonged large losses not seen in a smaller test.

The more data you use, the more patterns you see. A test of 500 days of data will have longer sustained moves than a test of 250 days; 1000 days is that much better than 500 days. More data means more of every price pattern and their combinations. There will be more new highs and new lows, whipsaw periods, and price shocks. In total, the risk gets greater as you use more data, and the profits rarely keep pace with them.

When you begin trading, you are adding the current data to your data history. After a year, you have increased a 1000-day test to 1250 days. The combined periods will contain new or more extreme patterns, just as though you had tested 1250 days, and as the data grows, so does the trading risk. As a result, when you begin trading, the risk may be larg er than is seen in testing.

Computers and systems have added valuable structure to both individual and institutional trading. Arbitrage, strips, multiple decision-making pro grams, and expert systems would not be possible without current tech nology. It has also contributed to risk control and dynamic asset alloca tion. It allows theories to be tested without real losses.

The use of computerized strategies is not the "easy" answer. It is real ly a more disciplined and limited form than traditional trading. It requires users to carefully review their procedures and decision making, and put them in an orderly form. In doing this, many traders question what they have been doing. Discretionary trading—the ability to apply instinct and select outside factors—is what makes a great trader. A com puter is not going to make an exception. For those traders who consis tently benefit from their market judgment, a computerized program often serves as a guideline. It can tell them the trend direction, show key price levels, and give them an idea of how others are positioned.

Some successful traders already use systems as a basis for trading. They are able to select which trades have more potential, enter at a bet ter price, or exit before profits disappear. You might find that they attribute much of their success to the system, while a careful, systematic test of the method will show that it does not produce net profits at all. The traders' skill is what actually makes it work. Yet these traders swear to the system's success.

Many traders are unsure whether computerized strategies improve their performance or only contribute to their peace of mind. It is said that, because life is a struggle against disorder, some traders would rather use a bad system than no system at all. Systems are good because they are definitive and can be validated by testing. Not all systems work simply because they are clear. This book is intended to help you understand systems and improve your trading performance.

When a market technician holds a position supported by fundamentals, expectations of success increase. But the rules of most systematic trad ing are not compatible with the equity fluctuations needed to maintain a trade based on government policy, supply and demand, or corporate expectations.

The reason for technical analysis is to cope with the unpredictable way that prices react to fundamentals, the difficulty in assessing objec tives, and the need to control risk. In exchange for improving control over an investment, it is necessary to sacrifice what seems to be an "obvious" bull move because prices retrace from their recent highs and exceed preset risk levels.

A technical system grinds out profits, alternating with losing trades. It may take a big loss on a short position during a bull move. Although traders can claim that they "know the system was wrong," it is the discipline of the program that ultimately succeeds.

Technical trading is not glamorous. It will never let you say that you bought at the lows and took profits at the top. But trading should be a business, and a systematic program is a plan to profit over time, rather from a single trade.

If you can keep your head, when all about you are losing theirs, then maybe you haven't, heard the news.

Conversations with other traders can lead you to believe that they knew more than you did about a surprising market move. You lost and they profited. However, if the move was a price shock, then no one could have known it was going to happen. They were probably caught on the right side of the market while you were on the wrong side.

Newspapers highlight the big winners or those newsletters that gave the right advice. Considering the number of reports published, some must have been short before the October 1987 stock market crash. Their advice could have been mediocre before that and worse, but you always find out who was right after a price shock. You can always assume that half the traders you know were right about a move, and those are the ones most likely to talk about it.

High expectations are essential to success, but unrealistic ones just waste time. Computers do not tell you how to profit in the market, they can only verify your own ideas. Using a computer to develop trading programs is a sensible, conservative approach. As with other tools, it requires skill, which comes from study and practice. As you become more proficient, you will learn more.

Because it is only a tool, results of system testing must be compared fairly with all other investments. Returns should be risk adjusted, and investments must be properly capitalized.

Be suspicious of unrealistic results, even when they are profitable. Strategies are trade-offs between many features. With more complex tools, there are many more chances for error. Check the details careful ly before accepting the results. There is no substitution for careful work.

Use new technology cautiously. Increased computer power takes the pressure off the individual's need to conceive a profitable strategy before testing. Many of the new methods discussed later can be a great asset or a crutch; their benefits are entirely in the hands of the user.

Researching and developing trading programs are unique activities among businesses. There is no assurance that some problems have a solution. Other times, successful plans and opportunities are short lived; even market patterns that have been reliable for many years can disappear. Part of developing a program is identifying when it no longer works.