books about online stock trading, forex, futures, stock investing, market, trading systems
Understanding Price Shocks

A price shock is the ultimate risk. It is an entirely unpredictable, large jump in price that is too fast to trade. Price shocks cause the ruin of more traders than any other problem. A price shock can be seen as a large opening gap, or a very volatile trading range, often three or four times the average size, sometimes bigger by a factor of 10. Because they occur infrequently but are so dramatic, price shocks are treated very differ­ ently by many analysts when they develop a trading system. Some ana­ lysts will make up special rules to be applied for specific past events; others will include shocks as a part of the normal price phenomena, to be resolved by strategy testing.

Trading Risk Is Higher Than Expectations

This chapter will show that price shocks, which are frequently small and only occasionally extreme, are the reason the risk of trading is always greater than expected. When you look back at historic price moves, especially with a computer charting or a test program, such as System Writer, TeleTrac, or MetaStock, it is easy to identify a price shock. The clear ones are seen as highly volatile days or large gaps.

How do you handle them? When you are testing or developing a strategy, you look for a trend or pattern that would have had the right position to take advantage of any major move. But is that really possi­ ble? Could you have known which direction the price would have moved? And what about all the small shocks? Many small price shocks are not obvious. Although they do not attract attention if their size does not cause any serious problems, they are just as unpredictable as large shocks.

Figure 7-1. Chart analysis with typical price shock*. The Deutsche mark chart shows eight clear price shocks. Those marked 1-5 are expect­ ed to have generated profits for a trend system; those marked A-C are most likely to have produced losses. The Gorbachev abduction first appears to be the obvious price shock, but closer study shows that there are surpris­ingly many gaps and volatile days. Unfortunately, some analysts choose their strategy by its ability to profit from these past moves.

Figure 7-1 is a chart of the September 1991 Deutsche mark, traded on the IMM. It shows the Gorbachev abduction on August 19, which pro­ duced a large profit for many traders. Those profits disappeared two days later when the market abruptly reversed. A closer look at the chart shows that other shocks were nearly as large. The point marked A indi­ cates an unexpected change of direction ending two days later and 300 points lower. The Moscow coup spanned only 250 points.

If you assume that you could have profited from a large price shock, you have mistakenly reduced your assessment of market risk. You can eliminate a price shock from a chart analysis or computer test, but you cannot remove it from real trading. A price shock is not predictable. That means you cannot assume that you would have profited from the price move, nor do you have to say that all price shocks would have caused losses. You can assume that half the shocks will be in your favor, and the other half will be against you.

Types of Price Shocks

Price shocks have no rules or patterns that can be applied in advance. Because they are always unexpected, they can occur any time and dur­ing any market environment. However, there is a distinction between a price shock that is the result of a structural change and one that is tem­ porary or ambiguous in its effects (see Figure 7-2).

(a) (b) (c)

Figure 7-2. Three patterns associated with price shocks, (a) A sudden drop to A has a sound basis but was exaggerated. Prices partially correct to B. (b) A structural shock will continue in the same direction (B) after the initial move (A), (c) A "false" shock (A), without basis, reversed as news corrected the situation. After 3 days, the effect had disappeared.

In general, wars, rumors, assassinations, and political coups have temporary effects on price (see Box 7-1 ). Weather could cause a struc­ tural change in supply but rarely does; it always results in an immedi­ ate overreaction. Lack of rain, too much rain at the wrong time, freezes, and monsoons all cause a nervous reaction. By harvest, it is clear that corn and soybeans are exceptionally healthy crops, and that Brazil is more than happy to supply the U.S. consumers with any orange juice shortfall at an agreeable price.

Price shocks based on surprise economic news are often structural; prices try to jump to a level that is a fair assessment of the news. The market may push prices a little too far, but overall a trader cannot expect to profit from a price reaction. When the Central Bank announces a rate cut of 0.5 percent, prices must move to the level dictated by that change. If a 0.25 percent cut was anticipated, then prices move down; if a 1 percent reduction was expected, then prices move up. With period­ ic economic and statistical releases, the difference between the news and the anticipation determines price reaction.

Many long, fast price moves are not price shocks. Weather-related news is often anticipated by the market. A freeze or hurricane does not occur without warning. As cold weather moves south to Florida , traders and growers become concerned about the increasing likelihood of a freeze. They start hedging by buying futures, or covering their shorts. The result is a market that starts drifting higher in advance of a freeze. A speculator using a simple moving average system may get a buy sig­ nal ahead of severe weather, the result of informed reaction to antici­ pated weather.

Similarly, a change of regulation that affects industry often has warning. A vote before Congress to change pollution control or standards has clear, sometimes measurable effects on a group of companies. A bill likely to pass finds its results already discounted in the stock price.

Impact of a Price Shock on an Investment

A price shock can cause a severe equity fluctuation in a fully funded account, but many traders use leverage whenever possible. Shocks can vary from 3 percent to 30 percent of the value of the asset. If you hold a conservative portfolio with 50 percent cash reserves and the rest allo­ cated to unleveraged stocks, or 50 percent cash and the rest leveraged at 5 percent margin in futures, then the impact of a price shock will be:

Portfolio Portion Size of Corresponding Corresponding Drop

Allocated to Price Shock Drop in Stock in Futures Portfolio

This Position (Not Leveraged) Portfolio (5% Margin)

50% 3% 1.5% 30%

(Maximum exposure) 30% 15.0% 300%

10% 3% .3% 6%

30% 3.0% 60%

The stock portfolio has no problem absorbing a 3 percent equity drop in a worst-case scenario when only 10 percent, or V5 the available trading capital, is exposed to one correlated group. Even the 15 percent loss is unpleasant but not fatal. Futures is another story. Mostly traded with high leverage, a portfolio is rarely prepared for a large adverse price move. The most conservative futures portfolio, with 5 assets allocated 10 percent each, and 50 percent in reserves, still lost 6 percent on a 3 per­ cent price drop.

Eliminating Price Shocks from System Performance

During the testing of a new strategy, most traders and analysts elimi­nate the losses due to price shocks, or gain from their moves, without being aware of it. They can

•  Select the most profitable system from an optimization test.

•  Luckily miss being in the market during a shock.

•  Test data that did not have significant shocks.

In the enthusiastic search for a great trading system, traders would select the system that performed best over historic tests. They are not critical of a system when there are profits. If a 25-day moving average had resulted in a long S&P position on the Friday before Gorbachev was abducted, it had large back-to-back losses over 3 days. A 5-day trend might have just entered a new short or closed out a long that day, and would have benefited from the shock.

Selection of the wrong system parameter can happen when you choose only the best results (i.e., as is automatically entered using a TeleTrac optimization). If the slower system netted a 5 percent return

The Kuwait invasion in August 1990 found most traders long (Figure 7-3(a)). The possibility of sustained oil shortages moved prices steadily higher. The U.S. retaliation in January was still an unknown, and the sharp reversal proved that the market's reac­ tion was a surprise to traders (see Figure 7-3(b)).

f

Figure 7-3. Kuwait price shock, (a) Iraq Invades Kuwait, (b) The U.S. retali­ates with Desert Storm.

BPH92.DLY-Daily

Figure 7-4. Conservatives win the British election.

with 20 percent risk, while the faster program returned 4 percent with a 6 percent risk (due to large losses during the Moscow coup), the num­ bers make clear that the faster program is more desirable. No one would immediately pick the system with lower profits and higher risk. What really has happened? By picking the system that profited from the coup, you have unconsciously assumed that you could predict a price shock. But that's impossible. Therefore, your conclusion is not valid. That is not to say that the slower system was better. At this point, we really do

3. The political coup in Russia on Sunday, August 16, caused a uni­ form pattern in Forex, oil, and equity markets. The S&P (Figure 7- 5(a)) moved opposite to a trend position, crude oil would have caused new longs to be set (Figure 7-5(b)), and the Deutsche mark (Figure 7-1) would have profited. However, they all would have posted large losses when markets reversed two days later, after Gorbachev's release.

CLV91.DLY-Daily

Gorbachev Abduction

Short Tests May Not See Shocks

Some systems are evaluated over recent data because older prices do not seem representative of current market conditions. The European Monetary System (EMS) has changed the spread relationship between the exchange rates of member currencies. By creating limits, prices are supported and patterns are different from pre-EMS data. Short amounts of data have the disadvantage of not having enough price patterns to develop a robust trading model. They may show only a bull market, or a few small price jumps. It would be unusual to see a large price shock in a small data sample.

Frequency of Price Shocks

You might think of every price change based on news as a price shock. Markets are filled with little jumps because of unexpected events such as periodic reports on unemployment claims, corporate earnings, an unexpected charge, Federal Reserve or Central Bank shifts in monetary policy (never announced in advance), trade balance, announcements of new government policy, crop estimates, daily marketing of livestock, amount of rain in the Midwest, or cold in the northern hemisphere. The difference between the market's anticipated assessment of a piece of information and the reality of the event causes a price shock.

Most price shocks are small. Often, the relative accuracy of market anticipation to the released information obviates a change in price. Sometimes, the difference between actual and expected is unimportant in light of other effects attracting the public's eye. After three years of prolonged recession (beginning in 1991) and steadily lower interest rates, a worse unemployment number is not as important to the market as a Consumer Price Index that signals possible inflation.

Only the bigger price shocks attract our attention, even though small­ er jumps occur frequently. Being unpredictable and frequent, shocks occur in a pattern (or lack of pattern) very similar to a random distribu­ tion. There are many small shocks and a rapidly decreasing number of large shocks.

Gaps and Ranges. Figure 7-6 shows the frequency of opening gaps and daily trading ranges and compares the S&P with the Deutsche mark for the 10 years ending with 1993. The inset chart begins with gaps and ranges of 0.5 percent; however, the frequency drops off quickly and the smaller number of large percentage moves cannot be seen. The larger chart shows only those 1-day gaps and ranges above 3 percent of the cur­ rent price. These values can be seen exactly in Table 7-1.

Table 7-1. Frequency of Price Shocks (January 1983-June 1993)

The implications of this are important. There are 51 of 3375 days (on average, 4 days each year), in which a price shock in the S&P will occur. From experience, we know that these are likely to be clustered together; therefore, we can assume that once each year there will be a volatile period of 4 days. One year, that move might be profitable, and the next year it might generate a loss. In either case, a 4 percent range is a 40 per­ cent swing in equity for futures traders, based on margin.

The chart shows that large gaps and ranges will occur periodically and will be large enough to represent a problem. Individually, the risks might be absorbed within a well-diversified portfolio, but in reality, price shocks often affect a broad range of markets and assets at the same time.

Why All the Fuss?

During testing, many trading program do not distinguish a price shock from other moves. By applying special combinations of rules, and selecting the best trend speed, the trading program can successfully be on the right side of the market whenever a major price jump occurs. It would be easy for a computer to scan stock market historic prices and identify a pattern of extreme drops. A system that sold stocks on every Friday, during the last half of October (from 1929 through 1993), cover­ ing the position on Monday afternoon, would have made a fortune. Although most trades produced a small profit or loss, a few major plunges overwhelmed the result.

What is wrong with this approach? You are collecting, classifying, and creating rules to take advantage of price shocks that are unpre­ dictable. You are attributing special traits to events that, by definition, have no traits. In general, if you have developed a system that did not show a large loss, you have done something wrong.

Using Stops for Risk Protection

In addition to a reduction in expected profits, the inability to identify a historic price shock affects risk control. You cannot assume that a price shock would produce a profit, and you cannot assume that a stop-loss would have saved even part of a loss. Resting stop-loss orders tend to be filled at the worst place, and visual stops are too slow to be effective.

•  A stop cannot get you out of a short position that is limit bid, or a
stock that moves quickly after a news release.

•  A stop will get you out at the worst price when the market begins to
trade.

•  In historic testing, a stop cannot tell that an intraday shock caused
prices to jump through the risk level and that the order was at the
high of that interval. (It is always safe to assume that you were filled
on a buy at the high of a 15-minute range, or sold at the low of a 15-
minute range.) A computer system that assumes a fill during an
intraday shock presents unachievably good results.

•  A "fast market" exists during a price shock and price quotes run late
on the screen. A visual stop could never cut losses because the event
could be over before it appears on the screen.

It would be comforting to place the stop in advance and expect that exe­ cution price, but an occasional shock is nasty, and a stop-loss rarely improves risk control.

Key Price Shock Concepts

Risk is always higher than expectations because historic testing (either computerized or manual) does not distinguish between data that can be used to forecast profits and price shocks that cannot be predicted:

•  You cannot know which shocks would have been in your favor, or
which would have resulted in losses.

•  You cannot know which days contained intraday price moves that
would not have allowed a stop-loss or a new trade entry to be exe­
cuted at a reasonable price level.

•  Many small price shocks that result in bad executions are much more
difficult to recognize when only historic price data is available.

Handling Price Shocks

So far, price shocks paint a dismal picture. They cannot be predicted, many of them cannot be seen afterward, and they can generate devastating losses. The comforting thought is that, when you accept the uncertainty of price shocks and do not assume profit opportunities, you know the worst case of risk. This is not necessarily good, but it is a safe way to evaluate trading and investment returns. Believing that a system has less risk will lead to more serious problems.

The following guidelines will help you avoid mistakes and assess risk more accurately. These points will not identify every price shock, nor is it likely that the final risk level will be as great as the real risk of trad­ ing, but it should be very close:

•  More test data gives more realistic results. Larger periods of test data con­
tain a greater variety of price patterns and more price shocks.
Parameter selection based on longer tests tends toward the longer-term
forecasts and slower trends. These can better absorb the effects of price
shocks. Faster trading strategies must show profits from price shocks,
in order to prevent losses from appearing disproportionately large. The
expected profits of a longer-term system may be lower, and the risk
higher, than a faster trading method, but the real trading results are
more likely to be similar to the slower system, and may vary far from
expectations of a fast system.

•  Use less data for parameter selection and more data for risk evaluation. If
older data is hot representative of current market conditions, it may
be more reasonable to select parameters based on a short test period.
Once those parameters are fixed, test a longer set of old data to get a
better evaluation of risk. It is not possible to find all the risk from a
small test sample. Use the old data to show more patterns of volatil­
ity and risk, and recent data for trend timing and profit patterns.

•  Find a worst case scenario in past prices. It is not difficult to look at past
charts to see obvious price shocks. Look for the largest price moves,
then consider a worst-case scenario to evaluate the risk. It is safe to
assume that what has happened before will happen again.

Creating an Artificial Data Series

A valuable transformation of data can give realistic system test results. It will be necessary to use a computer to do the following:

1. Scan the historic daily data and remove the data for the day a price shock occurred, plus the next two days of data. The day of the shock can be identified by a large opening gap or an unusually large trad­ ing range. You can select different size shocks by requiring the open­ ing gap or trading range to be 3, 5, or even 10 times larger than the average gap or range.

•  Create an index of prices without,the 3 days of data associated with
the price shock. This will close any gap created by the elimination of
data and change all the prices to percentage changes.

•  Test the trading strategy using the gap-adjusted series. This will result
in parameter selection that does not try to profit from price shocks, or
assume that it could be stopped out at unrealistic levels. It will pro­
duce a system that works in a "pseudo-normal" market (although
"normal" must really contain price shocks).

•  Run the trading strategy with the selected parameters through the
original data series, including all price shocks. The results should be
similar profits, but much larger risk. Half of the shocks should have
generated profits, and the other half large losses. If you find that
there were no shocks that caused losses, then use the profitable
shocks to indicate the magnitude of the potential losses. It should be
considered simply good fortune that a few shocks occurred in the
direction of the current position; in real trading it could be reversed.
You should manually evaluate the size of the past price shocks and
assume that it represents future risk.

You now have a realistic set of return and risk values to decide the merits of the system and the investment necessary to trade it success­ fully. The parameters selected for "normal" markets should return more consistent profits, and the final risk figures will give a realistic idea of the effects and frequency of price shocks. By removing the price shock data, the optimization process will never be able to fit parameters so that they profit from an unpredictable price shock.

The clear identification of price shocks which was used to create a gap-adjusted series also allows you to automatically recognize the same shock as the system during actual trading. When the shock occurs, you can change rules and treat the situation as a special case. Box 7-2 gives the FORTRAN code for creating this series. More on adjusting data can be found in Chapter 10.

Managing a Price Shock

You are going to take a big profit or a big loss from a price shock. Because you cannot predict when it will happen, you must assume that you will be holding a position, either long or short. We have discussed the use of a stop-loss to reduce risk and believe that a resting Stop order is more likely to capture the worst possible price. Then what are the choices?

You could hold a trade or exit it after the shock with a large profit or loss, whichever occurs. Because you can automatically identify a price shock on a computer, you can also test special strategies. For example, if the price shock is up, set a long or short position on the close, depend­ ing on the type of shock, then exit one or two days later (see Table 7-2). You can determine which shocks tend to continue and which reverse. If there are enough cases you can develop a clear price shock strategy.

Qualifying the Shock

There is a logical, accepted strategy to managing a price shock even without computerized testing. First, you must qualify the situation. If the price shock was caused by a fundamental, structural change, then only a small reversal should be expected. An announcement to raise interest rates % percent by the Central Bank means that bond prices will fall to the new level. If a \ percent increase was expected, prices will rise. It is not an issue of anticipation, but of fact. Interest rates are more definitive than most other news. Putting a price on the Gorbachev coup, plant­ ing intentions, retail drug prices, or a new national health program is not as simple. While most price shocks move further than necessary, a struc­ tural change means a permanent price shift. Some correction is normal, but a continuation of the new direction is also possible. Opportunities for recovering losses from a structural change are small.

Political news, natural disasters, and rumors dominate most other prices shocks. Assassinations are tragic but do not necessarily affect the safety or economy of a nation. Hurricanes, droughts, floods, and freezes devastate small countries and regions but rarely cause a substantial change in total supply. In the past, a freeze in Florida caused orange juice prices to soar; now, any shortfall in supply is happily filled by Brazil . Price shocks that cannot be confirmed or cannot be translated clearly into a price change are likely to move too far, too fast. These moves allow traders to recover a substantial part of their losses.

The Shock Is in Your Favor

When a price shock gives you a windfall profit, the position should be closed out immediately. Even though a structural change is likely to show additional profits, the increase in risk is greater than the potential for further profits. If the shtock causes a loss, the position can be man­ aged to recover part of the loss.

Figure 7-7(a) shows a price shock with some fundamental basis. Prices move sharply higher, then start a volatile, erratic decline. A short- term and long-term trend are shown as (1) and (2). Because prices move fast, system (1) cannot exit. The short trend would have been stopped out at the high, reversed to long and been stopped out again in The following code to create a gap-adjusted and shock-adjusted series cannot be done using TeleTrac, Easy Language, or spreadsheets because the new data series is shorter than the old one. The following code in FORTRAN reads the original data series OLD and creates an adjusted series NEW.

SUBROUTINE GAPADJ(PERIOD,GFACT,TRFACT,RDAYS)

C — "GAPADJ" subprogram for removing price shocks

C — PERIOD the number of days to determine normal price movement

C — GFACT the relative size of the overnight shock versus normal

C — TRFACT the relative size of the intraday shock versus normal

C — RDAYS the number of days to remove including the day of shock

PARAMETER (max$ = 500)

INTEGER DATE(max$),RDAYS

REAL OPEN(max$),HIGH(max$) > LOW(max$),CLOSE(max$),

+ TRANGE(max$),GAP(max$),INDEX

IF(RDAYS.LT.1)RDAYS = 1 C — Open input and output files OPEN(10,FILE = 'IN') OPEN(11,FILE = 'OUT) C — Initialize output count

N = 1 C — Read original input data

10 READ(10,1000,END=50)DATE(N),OPEN(N),HIGH(N),LOW(N),CLOSE(N) 1000 FORMAT(I6,4F8.2) C — Start output file on day of full period IF(N.EQ.1)THEN NX = 1

INDEX(NX) = 1000. WRITE(11,1100)DATE(N),INDEX(NX) 1100 FORMAT(I6,F8.2)

ENDIF

IF(N.GT.1)THEN NX = NX + 1

INDEX(NX) = INDEX(NX) + ABS(CLOSE(N)-CLOSE(N-1)/CLOSE(N-1) C — True range

TOP = HIGH(N) BOT = LOW(N) IF(CLOSE(N-1).GT.TOP)TOP = CLOSE(N-I)

IF(CLOSE(N-1).LT.BOT)BOT = CLOSE(N-I) TRANGE(N) = TOP - BOT C — Gaps

GAP(N) = ABS(INDEX(N) - INDEX(N-I)) ENDIF C — Test for a price shock

IF(N.GT.PERIOD + LAND.

+(GAP(N).GT.AVGGAP*GFACT.OR.TRANGE(N).GT.AVGTR*TRFACT))THEN C — Skip RDAYS + 1 DO 301 = 1,RDAYS PRIOR = CLOSE(N)

READ(10,1000,END = 50)DATE(N),OPEN(N),HIGH(N),LOW(N),CLOSE(N) 30 CONTINUE

INDEX(NX) = INDEX(NX) + (CLOSE(N) - PRIOR)/PRIOR WRITE(11,1100)DATE(N),INDEX(NX) C — If enough data, calculate average range and gap IF(N.GT.PERIOD)THEN SUMTR = 0 SUMGAP = 0 DO 20 I = N,N-PERIOD+1,-1

SUMTR = SUMTR + TRANGE(I) 20 SUMGAP = SUMGAP + GAP(I) AVGTR = SUMTR/PERIOD AVGGAP = SUMGAP/PERIOD IF(N.LT.max$)THEN N = N + 1 GOTO 10 ENDIF

STOP 'Data too big for array. Increase max$ and rerun.' 50CLOSE(10) CLOSE(11) RETURN END

Identification of a price shock once the factor has been determined:

IF (open > @AVERAGE((@ABS(open-close[1])/close[1]),period)*GapFactor OR © TrueRange > @Average(@TrueRange,period)*RangeFactor) THEN

Table 7-2. Price Shock Characteristics

days. The slower trend gets a windfall profit but gives back one-third before getting a trend reversal signal.

After the price shock, both fast and slow trends are catching up to the price jump while in Figure 7-7(a) prices have actually reversed direction. It is difficult to say that we are "following the trend" when the trend ended with the shock. This was exactly the situation following the 1987 stock market drop. Sensible management requires that trend positions be closed out after a windfall profit. If the change is structural, the position may be reset, but in most cases it is best to wait for a new trend signal to reenter.

Figure 7-7(b) depicts a structural change. Although prices continue higher, the faster trend system is stopped in and out because of higher volatility. The slower trend would have increased profits before encoun­ tering the same sideways period. Both strategies, however, would have been improved by taking profits immediately after the price shock.

Risk Reduction

Price shocks are accompanied by high volatility. By taking profits as soon as possible, you would not be holding a position during the period of increased risk following the first price peak. Even though a structural change produced more profits when the trade was held, the risk (mea­ sured by the volatility) was far greater than the marginal profit gained. Once this equity fluctuation is part of performance, it cannot be erased. Focusing on low risk translates into higher leverage and greater profits.

The Shock Produces a Large Loss

Once the price shock hits, and you are on the wrong side, risk is no longer an issue. The most important concern is to find the best chance for

recovering part of the loss. If the change is structural, you can expect only a small recovery, and the risk of further loss may be just as high. Timing is important. A professional trader can monitor the market, wait­ ing for a sign that trading is quieting and the surge of orders has been filled. Whatever correction is likely will come at this time. Afterward, prices may again move in the direction of the shock. A longer-term trad­ er, who may have seen the shock only after the close, would do best to get out as soon as possible, as long as the market is actively trading.

A temporary event that is likely to reverse can be managed in two ways. A conservative trader may hold the original position and wait for a rever­ sal to exit, expecting a 10 percent to 30 percent recovery. It may be closed out after one day if proved wrong and prices reach new ground, or if the reason for the shock appears to become structural. A more aggressive trader may double the original position, looking to recover 20 percent to 60 percent of the losses. In neither case should you expect to turn a loss into a profit. This is entirely a defensive management strategy.

Management Obligations

Traders often feel that they have an obligation to follow their system no matter what the circumstances. It is true that, if the rules are strictly fol­ lowed, it is easy to explain why things went wrong. Deviating from the plan, and subsequently losing more, will be embarrassing to explain.

Offsetting the rigid systematic approach is the concern that it is not reasonable to follow a "trend" when prices are moving in the opposite direction after a price shock. To resolve this dilemma, record the fol­ lowing list of clear rules to use when a "price shock day" is identified:

•  Identify the shock. A price shock occurs when a gap or trading range
is greater than a threshold value.

•  If profitable, then take profits and wait for a new trend signal.

•  // losing, then hold losses for one day after market trades. Exit if a new
extreme price, a 50 percent reversal, or a contrary trend signal occurs.

Using two clear sets of rules, a manager can justify the proper response to a price shock.

Long-Term Systems Are More Predictable Than Short-Term Systems

When choosing a trading strategy, remember that the longer-term, slower systems are more likely have returns similar to their historic tesing and expectation. Because the tests cover a long time period, they include more price shocks, more patterns, and more risk. They generate larger profits by holding the trade for longer periods; therefore, a price shock does not seem as important or disruptive to performance. The rel­ ative size gives it a chance to absorb the shocks. It is difficult to add a short-term rule to a long-term system for the purpose of eliminating a price shock. Short-term rules tend to change a slow trading system to a fast one, including large periods of being out of the market.

Summary

We prefer to trade a system which we perceive as having low risk, but we often create larger risks by unrealistically assuming that we can profit from price shocks. This leads to undercapitalized accounts and unpleasant results. A price shock affects all trading the same way, whether you are a short-term or long-term trader, a trend-follower or a countertrend trader. If a government report causes the yen to jump 300 points, all trades will be affected. It only matters whether you are long or short when the price shock hits. However, in the total performance profile or in a fully diversified portfolio, the effects of a shock on a longer-term view will be less dramatic.

By accepting the uncertainty of price shocks, you can implement alternate trading rules to limit further risk and possibly recover some loss. Once the price shock hits, it triggers a new plan. Trend-following systems do not apply to a market that has just experienced a structural change, no matter which direction prices move.

Understanding the real risk of trading is the most important part of system testing and performance evaluation. The business of trading is expected to return a profit for an acceptance of risk. Without clearly understanding the risks of each strategy, you cannot intelligently choose the best system and decide how much capital will be required to trade.

 Back to contents page