if tvpe = ti or type = 7 then begin

5.0V = @stt]cJeu(i<val ,period); if Ventry = i; tnen beqm

if (Tiadiff > a oTiii TadifffU 0) o- i<vol <= evcv rpsdsJ

izx" tstT:-rt rrarket;

if (tii-Gcliff < J snO i-t[: j >= 0) cr ."Jl dvqv rpd-J eMtlnrq aL Market,

maaiff 0 and TadirfLl] 0 then LiU. at reikei, if nadW < and ridtiiffCl] >= 0 Liien sel. 3t -rarKCt;

rise begin

if nridi-f > : arc iriodif[i] <= px-t-,nor: -ji-.;

li Tcd-if c;-! naa tfLU >- 0 f-itn extlcrq d narkLl;

if nadirf > ( maclif[n cid <viji cvcv rp5-!*5:v b.iv at rr.drket:

if iTi:ir.iff < 0 and TiiadiTlfl] Z r.-tz .vo - .jv:;v-psd"c-. sel1 at Dorset;

end;

!Low vo LiTP -! slew anr. fact cveroge

if type = 1:1 tlici beain

jvgv - iaveragoCvo . lod,;

t Ltnliv -= *fen beam

1 inad > 0 cjnd rrediTf!;i] -j * " .-u" - .-vi-l-sc ticn t.x cn.-rt [iJ-Ket;

if !!-! < 0 and -ticLliff:!! >= jv,::2 <= -.gv-o:. i-her eit at maiKf.

• r,j-.iff eid nacifrrr, •- ) ii-e-i Duy c iir.-<fct.

if . <= 3 arG naa-f[:i >- the- sl . a; F J----

rise begir

1 na:!! n > m:4ff[l- <= 0 .- :>r :t>i-rt --. - .

if r.aliff ! ar.-: TirsiffTll 0 :->!4.;ij ct .p.(.-.

ir riaclf > 5a naif+CP (* j\-r,Z .- ".e

if cjd-f; 0 sfd r;idJ.lr[l >- ; --o a-oZ j .Ov " " -e

Is OneVolume/Breadth Indicator Better Than Another?

The final test of a successful indicator is whether it adds value to your trading decisions. This can be determined by programming the indicator into the total strategj and testing the rules over a reasonable historic time period. Before doing that, we should decide what the increase or decrease in volume or breadth is telling us. if performance testing does not confirm our expectations, than we must rethink the approach because the chance of an error has increased. What appears to be a profitable discovery may only be an unlikely coincidence.

Looking at the selection of indicators, we see that small net changes in price can result in all volume being designated to one market direction. This is the case with Granvilles On-Balance Volume and Bolton-Tremblay. Is it reasonable to add all volume to the accumulated index if the S&P gained only a fraction of one point? On a single daj! this may seem to be arbitrarj, but just as so many of the simple sjstems. over a long period of time the net effect of this

approach is sound. In addition, we should not forget that these indicators fail to account for a range of noise that is similar to the erratic movement of prices: that is,

there is a normal variation in volume that does not mean anything special. On the other side, these indicators fail to isolate outliers, the significant events or price shocks that are associated with high volatility

The changes made by Mait Chaiken, which take a percentage of volume based on the relative close of prices within the daily range, have a very sensible basis and avoid the allor-nothing technique used in On-Balance Volume. Sibbetts Demand Index, by using the sum of 10 days volume, avoids this problem completely and smooths out results, a situation often needed when using the highly variable values of volume and breadth.

On the more confiising side, it has been suggested that the Demand Index, which simply accumulates the net of advances minus declines, be used with the absolute value of the daily calculation. Advances - Declines. That would give the same value when advances outnumbered declines by the same amount as declines outnumbering advances, which appears to lose information and fails to distinguish between maitet direction. Unless you clearly understand why taking the absolute value of any advance-decline formula makes sense, they should be avoided, even if historical testing shows a better result than other methods.

By seeing these indicators close together, they appear to be a collection of minor manipulations of data. The resulting series still requires interpretation using trendlines, diveigence, or new highs and lows, instead of having to interpret only the difficult movements of prices, we have added an equally difficult series of volume. While it is not clear that volume is alwajs usefiil in analyzing price movement, it can impart valuable information in certain situations.

INTERPRETING VOLUME SYSTEMATICALLY

Most sjstematic approaches to volume apply a long-term smoothing method to the data, then identifj trend changes to confirm price direction. This can he implemented with any of the accumulation indices, but not with single-day momentum or oscillatir values. For the oscillatcrs, most analjsis have taken the approach that high volume confirms a new price direction; therefore, they look for a reversal signal at the same time as peak or nearpeak volume. If they do not occur at the same time, a volume peak should precede a trend change. A decline in volume has also been used to confirm direction, but it is more likely to indicate that prices have reached equilibrium, and that a fiirther advance or decline requires additional confirmation.

Moving Average Approaches

A Straightforward way of using volume is to calculate a 10-day moving average of the volume to be used as a confirmation of a 20-day trend position. By simply requiring the current volume to be greater than the average volume over that past 10 dajs, you introduce the idea of greater participation associated with the new trend. An additional important benefit is that this volume condition acts as a filter, eliminating a substantial number of trades, if the nel retums are the same, the volume-filtered approach is far better because you are out of the market more, and are not reversing your position every time there is a new signal.

A similar method was proposed by Waxenbei. A 10-day moving average of the volume is calculated as the normal level, and a change in trend must be confirmed by a 20%

Alex Saitt.%APnce and Volunie-Based cyiem," Teclmical Analysis - .-& - iM.ircli 1991 H-ff.nIK Waxenbei "Teclmical Analysis Volume," Teclmical Anal s .-& Cinniodities iM.nch 1 TCI

increase in volume above this norm. (The 20° band acts as an additional smoothing filter, but may be replaced by a longer trend and smaller band.) Rxtremes in a trending move can be found at points that exceed approximately a 40° volume increase. Applied to the stock maitet, Waxenbei used the extreme volumes to indicate the end of a sell-off. To add more fiexibility for longer test periods, and to adjust for volatility, Bollinger bands (based on about 2 standard deviations, or 95o probability) can be substituted for the fixed percentage bands.

Alternately, using 13 dajs of volume, subtract the total down volume from the total up volume. A plot of the results will serve as a momentum indicator fran which overbought and oversold levels can be identified, if these values are unstable due to lact of liquidity, they may be smoothed using a short-term moving average.

Advance-Decline Sjstem

Advance and decline values, as with most volume figures, can be more useful if they are smoothed. By combining peak values of the net of smoothed advancing and declining shares with a directional move in price, Conners and Hajward have created a basic sjstem structure that they named CHADTp6 (Conners-Hajward Advance-Decline Trading Pattems ). This sjstem tries to identify reversal patterns by appljing the following steps:

1. Add the past 5 dajs of advancing issues (New York Stock Exchange)

2. Add the past 5 dajs of declining issues (New York Stock Exchange)

3. Subtract the 5-day sum of dechning issues (,1) from the advancing issues (2)

4. Divide by 5 to get the average daily value

CHADTP = ((«sum(advNYSE,5) - («sum(dec 1 NYSE,5)) /5

To trade using this oscillator, Conners and Hajward have determined that ±400 are the extreme levels where the values have been overbought and oversold. Based on this, we can apply the following rules to the S&P futures:

1 . Sell when CHADTP := +400 and the SP frades. 10 below the low of the previous day; buy when CHADTP =: -400 and the SP trades . 10 above the high of the previous day

2. Note that the oscillator does not have to exceed its recent exhemes on the day of the buy or sell signal.

3. Timing is best if the signal corresponds to a newspaper commentarjindicating "depressed volume" (which is seen as an excess of cash waiting to enter).

AN INTEGRATED PROBABILITY MODEL

If there is a noticeable relationship between price, volume, and open interest (or market breadth), then a probability model can be constructed to test its importance. To do this it is necessarj to constmct a one-dayahead forecast using a simple linear regression model, then back-test weighting factors for each element. Because of the powerful software products, this has become a very manageable process. Using Omegas EasjLanguage, each oneday-ahead forecast is determined using n past dajs:

1. Price forecast: Pf = Price + LinearRegSlope(close,r)

2. Volume forecast: Vf = Volume + L1rearRegSlope{volume,n)

3 Open interest forecast: Of = Dpint + LtnearReqSlope{opirit,r) 4. On-BalanceVolume forecast: oeVf = OBV + LinearRegSlope{DBV,n}