|
PureBytes Links
Trading Reference Links
|
do you know the TASC month and year the article came out...I have the hardcopy
magazines..thanks....
At 11:45 AM 1/10/2002 -0500, you wrote:
>
> Not sure if this is what you are looking for but ...
>
> Peter Gialames
>
> Here is the text from S&C V. 10:5 (220-224): Forecasting Tomorrow's Trading
> Day by Tushar S. Chande, Ph.D.
>
>
> Using linear regression as a crystal ball for forecasting the market? After
> all, if you were to be able to<?xml:namespace prefix = o ns =
> "urn:schemas-microsoft-com:office:office" />
>
> determine tomorrow's high, low and close for trend changes and placement of
> stop points, it would
>
> simplify your life immeasurably. Can it work? Tushar Chande explains how it
> can be done.
>
> Wouldn't you trade better It you could "see" the future? A simple linear
> regression can provide an
>
> objective forecast for the next day's high, low and close. These ingredients
> are essential for a trading
>
> game plan, which can help you trade more mechanically and less emotionally.
> Best of all, a regression
>
> forecast oscillator, %F, gives early warning of impending trend changes. The
> linear regression method is
>
> well known for finding a "best-fit" straight line for a given set of data.
> The output of the regression are
>
> the slope (m) and constant (c) of the equation
>
> (1)Y = mX + c
>
> Here, m and c are derived from a known set of values of the independent
> variable X and dependent
>
> variable Y. The relative strength of the linear relationship between X and Y
> is measured by the
>
> coefficient of determination r 2 , which is the ratio of the variation
> explained by the regression line to the
>
> total variation in Y. Here is a table to help interpret the values of r 2 ,
> which range from 0 to 1:
>
> The coining of the term "regression" can be attributed to Sir Francis
Galton,
> who observed in the late
>
> 1800s that tall fathers appeared to have as a rule short sons, while short
> fathers appeared to have as a rule
>
> tall sons. Galton suggested that the heights of the sons "regressed" or
> reverted to the average. Technician
>
> Arthur Merrill also had a good explanation in a recent issue of STOCKS &
> COMMODITIES, and Patrick
>
> Lafferty recently wrote on an application of multiple regression to gold
> trading. Virtually all introductory
>
> books on statistics have a detailed discussion of the linear regression
> method.
>
> Successful professional traders emphasize the importance of having a trading
> plan. A trading game plan,
>
> much like that of a football team, clearly defines specific actions under
> different conditions. The linear
>
> regression method is very useful in developing a forecast for the next
> trading day's high, low and close
>
> based on the last five trading sessions. The method is general and
> broad-based enough so that it can be
>
> used with stocks, indices or commodities. The forecast is the basis of my
> trading plan: I can define what I
>
> should do if the market rises above the forecast high, falls below the
> forecast low or stays within the
>
> forecast range. This way, I can avoid being emotional and trade as
> mechanically as possible by having a
>
> plan to rely on.
>
> FORECASTING WITH LINEAR REGRESSION
>
> I like to use at least 10 days of data and develop a forecast for the high,
> low and close. The five-day
>
> regression is a good choice for short-term trading. You can use any
length of
> regression you like. Here
>
> are the calculations with the daily close in a spreadsheet format:
>
> 1 Perform a linear regression with the first five days of data to obtain the
> slope m and constant c such
>
> that
>
>
>
> X Value Daily Close
>
> 1 Day 1
>
> 2 Day 2
>
> ....
>
> 5 Day 5
>
>
>
> 2 Forecast the next day's close with the slope m and constant c from step 1:
>
> (2) Forecast close (Day 6) = 6m + c
>
> 3 Record m, c and r 2 on the same line as Day 5. Record the forecast from
step
> 2 one day ahead, with
>
> Day 6. Note when we are using five days' data, the first forecast is for Day
> 6.
>
> 4 Step the calculation ahead one day such that
>
> 5 Record m, c and r 2 as in step 3.
>
> 6 Calculate the regression forecast oscillator, %F, as
>
> (3)
>
>
>
> %F = ((Y-Yforecast)/Y)*100
>
>
>
> where Y is the close for Day 6 and Y(Forecast) is the forecast for Day 6
from
> step 2 (from Day 5).
>
> 7 Record the oscillator on the same line as Day 6.
>
> 8 Step the calculations ahead one day at a time until the most recent day.
>
> Technically, we can use the linear regression to develop a point forecast
> (single value) for the next day
>
> (as in step 2) or a range (interval) of values with a certain confidence
> level. The interval widens, greater
>
> the variation in the data and greater the desired confidence level.
>
> I use the forecast oscillator, %F, to determine if my forecast is above or
> below the actual market data.
>
> Since
>
>
>
> %F = ((Y-Yforecast)/Y)*100
>
>
>
> where Y can be any market variable for stocks, indices or commodities, %F
> measures the percent
>
> deviation of the actual value from its forecast. In a trading market, %F
> changes its sign before a
>
> significant trend change. In trending markets, %F tends to change sign early
> in the trend. I interpret %F in
>
> the context of the r 2 Of the regression. A low value of r 2 plus a
change in
> sign of %F is a good signal of a
>
> change in trend. Market extremes and periodicity can also be observed on the
> %F charts.
>
> DEVELOPING A TRADING PLAN
>
> You can use the forecasts to develop a specific trading plan to suit your
> trading style. I use the forecasts
>
> in several ways.
>
> Forecasts as stops. I use the high and the low as action points. If the
market
> exceeds the forecast high, it
>
> wants to go up. To trade with the trend, I put a buy stop a few ticks above
> the high. If the market falls
>
> below the forecast low, it wants to go down. Hence, I set a sell stop a few
> ticks below the forecast low. If
>
> you want to trade against the trend, sell short near the forecast high and
> buy near the forecast low.
>
> Forecasts as intraday range scale. The forecasts provide a scale for
> evaluating the trading day. The
>
> market can stay within the expected range or go outside. On a down day, the
> intraday high is well below
>
> the forecast high and may be below the forecast close. On an up day, the
> market stays well above the
>
> forecast low and often above the forecast close.
>
> General rules for trading with forecasts. Here are some general rules:
>
> " Use the forecasts only if r 2 is greater than 0.1. Higher the value of r 2
> , the greater the confidence in the
>
> forecasts.
>
> " A trend change is imminent when r 2 falls below 0.1. Prepare to close
> longs.
>
> " A trend is in place if r 2 is greater than 0.6. As a trend follower, you
> could wait for this value to be
>
> exceeded before opening positions. This would keep you out of short-term
> fluctuations.
>
> " An early warning of a trend change is provided by a zero-crossing of %F,
> the forecast oscillator.
>
> Prepare to tighten stops and look for changes in slope and coefficient of
> determination for
>
> confirmation.
>
> "A change in trend is confirmed by a change in slope of the regression. Open
> positions in direction of
>
> trend change. To trade against the trend, look for peaks in slope and
> strength of the linear trend.
>
> "The trend will usually change in the direction of %F.
>
> "Always be prepared for a market move against the forecast. Use stops!
>
> A SAMPLE TRADING PLAN
>
> I have developed a forecast for the high, low and close for January 20,
1992,
> from the previous five
>
> trading days, seen in Figure 1. The market was making new highs the previous
> week. Was a downward
>
> movement imminent? Let's look at the data from Friday, January 17, 1992:
>
> The market was trending moderately (0.4<= r 2 <0.6), but the forecast
> oscillator %F was negative for
>
> high, low and close, warning of a possible change in trend. The relatively
> small slope of the regression
>
> for the high meant the market was meeting resistance. The slope of the
> regression for the close had turned
>
> down from the high values during the recent strong uptrend. The forecast,
> however, called for a strong
>
> close near the highs of the day, but that seemed doubtful, given the low
> slopes in a moderating trend. The
>
> plan was to watch for a change in trend. If the market opened weak, a
bearish
> strategy was called for. For
>
> example, I would consider buying the Standard & Poor's 100 Index OEX January
> 390 puts, or selling
>
> short the S&P 500 March futures contract.
>
> The high daily volume of OEX index options traded makes the S&P
>
> 100 index an interesting application of me regression forecast
>
> approach.
>
> The market opened at the Friday close and weakness was evident at the open,
> as the S&P 500 futures
>
> opened lower. It was clear in early trading that the trend would be down, as
> the market traded well below
>
> the forecast high and close. Clearly, the forecast range provided a good
> scale, since it reinforced the
>
> concept that the market was weaker than the trend of the prior five days. A
> bearish stance would have
>
> been profitable.
>
> THE NATURE OF REGRESSION FORECASTS
>
> The high daily volume of OEX index options traded makes the S&P 100 index an
> interesting application
>
> of the regression forecast approach. I have examined a time period from
early
> October 1991 to
>
> mid-January 1992. The OEX close and its forecast are in Figure 2; the r 2
> values in Figure 3; %F in Figure
>
> 4, and Figure 5 has %F around the mid-November plunge.
>
> Several observations can be made from the OEX analysis. First, the forecast
> lags the OEX in an uptrend
>
> or in a downtrend. Second, the close and the forecast cross over several
days
> before a trend change. This
>
> crossover can be seen as a zero crossing in the %F chart. Significant trend
> changes are preceded by
>
> trendless periods with values of r 2 near zero. Strong trends are
accompanied
> by high values of r 2 and
>
> regression slope. These observations support the general rules of
> interpretation noted above. As Figure 5
>
> shows, %F provided a timely warning of an impending trend change just before
> the OEX fell 15.68 points.
>
> I have included data for wheat (cash) from 1989 to indicate the use of this
> approach with commodities.
>
> The market showed significant trends during this period with good
> periodicity, as shown in Figures 6, 7
>
> and 8. The %F zero crossings were timely indicators of trend change.
Features
> observed with OEX charts
>
> are also seen here; note in particular how %F can be used to identify
> extremes in the market from Figures
>
> 4 and 8.
>
> Simple linear regression yields forecasts of the high, low and close for
> stocks, indices or commodities.
>
> these forecasts can be used to develop a trading plan. You can trade with
the
> trend, against the trend,
>
> intraday or interday. The forecast oscillator, %F, provides early warning of
> trend changes taken together
>
> with the regression slope and coefficient of determination. This approach
> works best in trending markets
>
> or trading range markets; it is only moderately useful in volatile markets
> with choppy price action. These
>
> objective forecasts will let you trade less emotionally and more
> mechanically. Profits will look up when
>
> you can look ahead.
>
> Tushar Chande holds a doctorate in engineering from the University of
> Illinois and a master's degree in
>
> business administration from the University of Pittsburgh.
>
> REFERENCES
>
> Lafferty, Patrick [ 1991 ]. "A regression-based oscillator," Technical
> Analysis of STOCKS & COMMODITIES,
>
> Volume 9: September.
>
> Merrill, Arthur [1991]. "Fitting a trendline by least squares," Technical
> Analysis of STOCKS & COMMODITIES,
>
> Volume 9: December.
>
> Pfaffenberger, Roger, and James Patterson [1987]. Statistical Methods for
> Business and Economics,
>
> Irwin.
>>
>> -----Original Message-----
>> From: owner-metastock@xxxxxxxxxxxxx
[mailto:owner-metastock@xxxxxxxxxxxxx]On
>> Behalf Of Steve Karnish
>> Sent: Thursday, January 10, 2002 10:34 AM
>> To: metastock@xxxxxxxxxxxxx
>> Subject: Forecast Oscillator
>>
>> List,
>>
>> Does anyone have the math formula for Chande's Forecast Oscillator?
>>
>> Thanks,
>>
>> Steve
>
>
>
|