Getting in was the easy bit - Now what do I do?

[TheBramble]

I’m currently doing an analysis of GOOG on another thread and part of that required me to codify a trailing stops strategy. I did that based on the specific dynamics of the instrument in question (GOOG) - how it has tended to perform in the past and where it is at this point in time.

I maintain this is the optimum way to construct all aspects of a trading position for any instrument. However, it is more time intensive than an automated or indicator based approach and while I’m going to stick to my guns on the way I now do it, there is room for considering a close approximation to a one-size-fits-all if you’re not obsessive about fine tuning your accuracy. So I’ll share with you for what it’s worth, and for want of a better term, an Inverse Volatility stop system I worked with on my way to getting to where I got to before I got to where I am now...

I’ve mentioned elsewhere that most tyro traders’ approach to using volatility as part of their stop constructions, if they’re not trading volatility specifically (and most tyro traders don’t) is quite opposite to what it should be. When volatility expands, most take this as a sign to exit their positions when in fact, they should be hanging back and allowing the price some room. The time to come on in hard on the action is when volatility slackens after a bout of excitement. You only have to think about the basis of volatility and volume for this to become obvious.

When you’re looking to fine tune your exits based on dynamic development of price and volume, you have to put yourself in the position of those who actually do move the markets (those who ARE Volatility and Volume). They have a view of ‘both sides’ of the market. Something you and I do not have. So when they are seen to be active (and as importantly, when they are seen to have withdrawn their active involvement) it is a sign that you need to be paying attention.

The contexts you need to be looking at are recent price action (flat, rising, falling); where the close is occurring on the bar/candle in your timeframe of choice; what the volume has been and is now and; what the high-low range has been and is now. All of these give vital clues to professional involvement and as importantly, lack of professional involvement.

I personally use this rather more dynamic and pro-centric approach to my final exit determination rather than genuinely fixed trailing stops which if I can find a way to describe that which doesn’t take another War & Peace type post, I will do so.

I started off this line of enquiry and research a while ago and a post of mine on this site somewhere went into some detail defining what factors and data I was starting to work with, but I haven’t the foggiest where it is now. However, I’ll update you on how I have developed these lines of thought in further posts as and when I get the time.

 

[TheBramble]

Taster for Inverse Volatility

Look, before I give an algebraic rendition of the above, and as I know working stuff through for yourself works a million percent better than just having it delivered to your door (ever wondered why even the most meagre self-cooked meal tastes so much better than a Domino’s?) consider the following.

Just volume and range.

What does it mean to you when volume is relatively large and range is large? Or for that matter, volume is relatively light and range is relatively small?

And in contrast, consider what it means when volume is large and the range is tight. Or range is wide and volume is tight.

Imagine those two as two separate dials or gauges and as you press something, one goes up or down and you press something else and the other goes up or down.

What of those two primary categories above do you believe represent the greatest and the smallest in terms of real volatility?

We’re looking for synchronous movement in both. Big volume and big range. Small volume and small range. What do they mean.

When you get small volume with big range or big volume with small range - what does that mean?

I’ve just seem Charlton’s post on the other thread, but I’m way too knackered to respond tonight.

 

[Charlton]

I maintain this is the optimum way to construct all aspects of a trading position for any instrument. However, it is more time intensive than an automated or indicator based approach and while I’m going to stick to my guns on the way I now do it, there is room for considering a close approximation to a one-size-fits-all if you’re not obsessive about fine tuning your accuracy.

This is an approach I employ. It is possible, if you have the time and minor programming skills, to make it less intensive. It's possible to define specific combinations of price e.g. relative price spread, direction, where the close lies etc. and volume e.g. relative size and the inter-relationship between the two, with the aim of identifying areas of strength or weakness in the market to focus on or for use in scanning.

The contexts you need to be looking at are recent price action (flat, rising, falling); where the close is occurring on the bar/candle in your timeframe of choice; what the volume has been and is now and; what the high-low range has been and is now. All of these give vital clues to professional involvement and as importantly, lack of professional involvement.

Indeed

I started off this line of enquiry and research a while ago and a post of mine on this site somewhere went into some detail defining what factors and data I was starting to work with, but I haven’t the foggiest where it is now. However, I’ll update you on how I have developed these lines of thought in further posts as and when I get the time.

Codification of a lot of this is not especially difficult, but (as per your other thread) the assignment of factors it the trickier bit. For example: what constitutes a wide spread, narrow spread, abnormally wide spread; what consitutes high volume, low volume, normal volume. I have adopted the approach of using an MA of price spread and volume. This is the only "indicator" I use, but not really as an indicator, but as a means of determining the norm.

Have you attempted to codify your analysis and how have you handled the variables ?

Charlton

 

[Spam Man]

I have adopted the approach of using an MA of price spread and volume.

Alternatively - start getting your head round percentiles and the PERCENTRANK feature in Excel.

For example - knock up a spreadsheet showing a few years worth of daily OHLCV data, and use PERCENTRANK to rank each day's volume within the context of itself and the previous 99 days' data. A "1" means the highest in the last 100 days, a "0" means the lowest, and everything in between is somewhere in between.

Example attached (the PERCENTRANK formulas in col 7 do not start until row 100, obviously).

You can then, if you choose, create a formula on each line for the daily volatility, or the movement in price, or whatever, and repeat the process.

At the top of the sheet is a table which uses the volume PercentRanks and a CHITEST (Chi-Squared Test - google it) or two to prove that high volume on any given day, suprise-not-very-surprise, is likely to lead to high volume the following day. Anyone with an interest in this type of stuff will probably be able to see that clearly, but I can explain further if necessary. The code which created the tables is in Module 1, and is extremely simple, so feel free to mess around with it....a familiarity with CHITEST in Excel can take you a surprisingly long way....

 

[Charlton]

Alternatively - start getting your head round percentiles and the PERCENTRANK feature in Excel.

For example - knock up a spreadsheet showing a few years worth of daily OHLCV data, and use PERCENTRANK to rank each day's volume within the context of itself and the previous 99 days' data. A "1" means the highest in the last 100 days, a "0" means the lowest, and everything in between is somewhere in between.

Example attached (the PERCENTRANK formulas in col 7 do not start until row 100, obviously).

You can then, if you choose, create a formula on each line for the daily volatility, or the movement in price, or whatever, and repeat the process.

At the top of the sheet is a table which uses the volume PercentRanks and a CHITEST (Chi-Squared Test - google it) or two to prove that high volume on any given day, suprise-not-very-surprise, is likely to lead to high volume the following day. Anyone with an interest in this type of stuff will probably be able to see that clearly, but I can explain further if necessary. The code which created the tables is in Module 1, and is extremely simple, so feel free to mess around with it....a familiarity with CHITEST in Excel can take you a surprisingly long way....

Thanks - I will take a look at this approach and compare it to my averaging method.
Certainly generating a ranking number would prove extremely useful to me.
I use Amibroker for charting and will check if this function exists there.

Charlton

 

[firewalker99]

I am really looking forward to seeing this thread develop as the title of it describes a part of trading that imo a lot of people overlook. "Getting in" is indeed the easy bit (well, not always...) but knowing what to do next is a different matter. Especially since there are so many different variables that come into play. After getting in a trade, your stop will determine the point where you are wrong, but how about the target? A small scalp-size target might prove you are right on a 3-minute chart but wrong on a 15-min chart. A fixed target based on the current or historic volatility can give a certain indication but what about those days when the average daily range suddenly expands into something as big as three or four times the normal range? Trailing stops haven't been one of my favourite methods as they usually only work in clean moves. Either way I can think of several situations where you'd be leaving lots of points on the table, nomatter what method you implement.

I've been researching and studying several ways to improve on exits (cfr. the thread I started recently http://www.trade2win.com/boards/showthread.php?t=25049) and am slowly succeeding in taking more profits out a good trade. However, catching the exact turn and exiting at exactly the best price is still a whole different matter. So as always there is room for improvement and I'm anxious to hear what you have to say about this matter TheBramble.

I personally use this rather more dynamic and pro-centric approach to my final exit determination rather than genuinely fixed trailing stops which if I can find a way to describe that which doesn’t take another War & Peace type post, I will do so.

I started off this line of enquiry and research a while ago and a post of mine on this site somewhere went into some detail defining what factors and data I was starting to work with, but I haven’t the foggiest where it is now. However, I’ll update you on how I have developed these lines of thought in further posts as and when I get the time.

 

[firewalker99]

"Just volume and range"

Interesting questions you raise there. I'll try and give my view on the matter. Anybody who thinks or feels like interpreting the signals otherwise, please share your views.

What does it mean to you when volume is relatively large and range is large? Or for that matter, volume is relatively light and range is relatively small?

When volume is large and the range too, the move is very obvious and therefore the public often gets lured into thinking this is the direction the market will continue. But often the volume will peak and so will the range near selling or buying climaxes, tricking the majority while the smart money will instead already be doing the opposite of what price is doing at that particular moment.

Or for that matter, volume is relatively light and range is relatively small?

When volume is light, and the range is small it's worth noticing the volume peaks. In my experience when you have a slim range and volume is relatively light a larger move is bound to follow. Usually the professional players are silently accumulating or distributing. Because of the money they are playing with they are unable to unload or buy all at once, so they'll try and spread it. Preferably buying at or selling from the same price level, so they prefer to keep price in the range.

And in contrast, consider what it means when volume is large and the range is tight. Or range is wide and volume is tight.

When volume is large, but the range tight, there's a tough struggle going on between buyers and sellers. Eventually buying or selling pressure will overcome the other and price will move outside of the range. I'd like to think of it as a "tug of war", a rope pulling contest where there's a lot of pressure going on from each side, but the rope doesn't move a lot. When eventually the one side overcomes the other, there's a sudden release of energy and the rope will shoot to one direction. A similar thing happens to price after it's been released of the range.

Or range is wide and volume is tight.

On the other hand when the range is wide but volume tight, I would be careful of not getting eaten up by the spread. When trading is thin and price is jumping around I prefer to stay out. It seems like a risky situation. Just my 2c here.

 

[TheBramble]

All indicator based systems end up generating average results. When you start building an indicator you’re typically doing so for one instrument. At one point in time within its cycle phase. And within the then specific context of the overall market dynamics – at that time. You tweak and tweak and tweak until you get it just right, which it is, for then. And they you use it on other instruments and across other market types and phases. And it ends up averaging out result-wise over time. It can’t not.

So you constantly review your indicators for each instrument and as the markets move and you tweak and tweak and tweak some more. All the time working with the base price, volume, action and time data.

My point is, why go to all the additional trouble after doing all the tweaks for that particular instrument at that particular point in time within the current market phase in coding up indicators when you’ve already acquired more than enough fundamental technical information to make a trading construct based on the price and volume action alone anyway? Within a matter of days, weeks or months, an indicator’s validity will diminish and if you carry on using this indicator, you’ll end up with average results. All indicators end up by definition, if you use them across a number of instruments and in all market conditions, producing average results.

You’ve gone to more effort to make a one-size-fits-all indicator that never actually does for more than one instrument at one time and eventually doesn’t even do that, than you would have needed to, to have made high quality, base data decisions on each instrument individually, each and every day.

Look, at the week-end generally (beginning of the week this week for me) it takes me 10 minutes to get a fix on the various inter-market relationships (Currencies, Interest Rates, Commodities, Bonds and the currently in my view the very anomalous stocks sectors). It’s an update of my world view on the ‘nature’ of the markets. I then look through the instruments from which I’m going to be selecting potential trades in the coming week. I trade intraday. There are 126 selected for this week. They are selected from a very large number of base instruments that currently meet my trading criteria in terms of price, activity, volatility, behaviour and volume at this current time. Although they are selected from a very large database, by and large, it’s the same ‘names’ over any 3 month period which is the period of my major review. And even then, same as everyone else I’m sure, you have stuff which you’ve been happily trading since Day-1. Old friends

Each day before trading opens it takes me maybe an hour to go through all of these instruments I have hot-listed, about 30 seconds for each one to note areas of S/R that are key for today, areas of congestion and ranging in readiness for breakouts and note where each sits within it’s own cycle phase. That’s it.

They sit in a scanner based on EXCEL which monitors the volatility issues I’ve been discussing elsewhere. When the scanner throws me a bone, I’ll give the instrument in question my full attention for a whole 10 seconds to establish as best I can what the meaning is of the current action and what is the likely intent of the professional interest or disinterest at that point (something I may get into in a later post). It only generally takes 10 seconds or so as you tend to get to know your instruments quite well and when you get the ‘beep’ you’re almost already aware of what it’s about and which of your plans are likely to be developing.

If it’s something that is cued-up for entry I’ll move it to my active screen and wait for the entry criteria to develop, or not. Although I can monitor a number of potential entries at a time, the scanner will also let me know when something has gone off the boil if I don’t clock it first – either way, I’ll move it out of active and back into the alert heap.

It’s less effort and more accurate for me to do a physical manual analysis of each stock in this way than it would be to try and build indicators which don’t get half as close. Plus indicators tend to take you away from educating yourself on the current true nature of the instrument you’re trading, especially if you rely on them solely for your signals.

Anyhow, here is the Inverse Volatility Stop I used to use which worked on the premise that when volume and range were relatively in tune, you were more likely to be able to leave your trade on than not and if volume and range went out of tune, you may want to consider pulling the plug. Volatility for me may not be your version or the standard definition. Low volatility (my definition and usage in this context) is when there is unusually (relatively) large range action with relatively low volume (why would that be and how would that be accomplished and for what end?) or when there is unusually (relatively) low range with relatively high volume (where is all the volume going to and from not to add any impetus to the price action or range and why would that be and how would that be accomplished and for what end?). High volatility is where the range is appropriate (given recent history) to the range developed by that volume being transacted.

This indicator doesn’t accurately factor in the position of the close on any bar, which is very important in my manual analyses, other than the exit bar. It sometimes gets you out too early. Sometimes too late. Sometimes it doesn’t work at all. But Mostly is gets you out within 2-3 bars of the top/bottom. Interestingly, I haven’t modified or even used this indicator for some time and if anyone bothers to use it, I guess I’d be intrigued as to how it has stood the test of time.

It doesn’t take into account the bias you need to have in place between SHORT and LONG trades. There is a bias, and it is timeframe sensitive, but this simple indicator doesn’t have one.

It is a very general purpose codification in that it doesn’t adapt the moving average factor dependent upon the timescale of the bars its used on – it should if you were looking for even greater accuracy, but this one doesn’t. Which doesn’t matter, as the whole point of my post (this entire thread actually) is to underline the second class nature of using indicators in place of the base data itself you have to thoroughly understand and use anyway to engineer the indicator in the first place! And the (paradoxically verbose when attempting to describe) simplicity of looking just at price action, volume and time.

One other thing, it’s not an ENTRY system, it’s an EXIT system. You’re already in a trade or you aren’t interested. Now there will be some who will seize upon this indicator and attempt to turn it into a ENTRY based system. I know I always used to... but don’t. Just because you get a signal to quit a LONG does not mean you are getting a signal to enter a SHORT. Very much not the case. And especially so when working with raw volatility.

Finally, the indicator, its use and interpretation.

It’s a STOP indicator in the sense you are going to use it to exit your position. But it might more accurately be termed a TARGET indicator as you use the indicators to STOP you out of your position while it’s still moving in the direction you want to be headed. It’s not going back against your price or direction like just about every other STOP system you’ve ever used. Some will not feel comfortable with this (surely there’s much more to be had out of this trade if it’s still heading my way?), others will much prefer getting out without having to suffer the indignity of being told enough is enough by having it take a snip out of their profits.

You will have a line above your price action and one below. Mostly. As it’s a target system rather than reversed-direction stop based system, you don’t have to worry about fitting any crank mechanism to prevent your stop moving back against your position (in a normal stop system you wouldn’t want the stop getting any lower than any previous stop level in a long position). Just let it meander like a badly behaved Bollinger.

When you get a bar with a close above your top line when you’re in a LONG position or a close below the bottom line when you’re in a SHORT position, you exit on the next bar. The price action will normally still be moving in your direction, but you get out because the volatility (my definition, my context) has shrunk.

You base your upper and lower bands around the mid-pint of the range.

You take the value of previous bar’s Inverse Volatility value. I’ve ripped this directly from my old Metastock formula and realised I forgot to rip the actual Inverse Volatility function. LOL! I’ll either edit this post later or slap in a new post with the formula but going from memory, the Inverse Volatility function was the current range divided by the 5 bar simple moving average of range DIVIDED by the current volume divided by the 5-bar simple moving average of volume. Pretty simple stuff.

(H+L)/2 + Ref(((((H+L)/2)*(Fml( "Inverse Volatility")/100))),-1) UPPER BAND

(H+L)/2 - Ref(((((H+L)/2)*(Fml( "Inverse Volatility")/100))),-1) LOWER BAND

 

[TheBramble]

When volume is large, but the range tight, there's a tough struggle going on between buyers and sellers. Eventually buying or selling pressure will overcome the other and price will move outside of the range. I'd like to think of it as a "tug of war", a rope pulling contest where there's a lot of pressure going on from each side, but the rope doesn't move a lot. When eventually the one side overcomes the other, there's a sudden release of energy and the rope will shoot to one direction. A similar thing happens to price after it's been released of the range.

Another possibility is that the buying/selling is being absorbed by the pros. Depending on where the current price is in relation to historical high volume price levels and local S/R this might be that they are willing to soak up all floating stock in readiness for the advance or they are unlaoding all stock to weak hands prior to the development of a reversal.

On the other hand when the range is wide but volume tight, I would be careful of not getting eaten up by the spread. When trading is thin and price is jumping around I prefer to stay out. It seems like a risky situation. Just my 2c here.

A wide range does not necessarily indicate anything about sthe spread. Sprerad could be wide too, but it doesn't have to be. This is all possibly secondary when you consider what mechanics have to come into to play to allow this scenrio to develop.

If range is wide what is it that is making the price move out of proportion to volume? If there's no volume or insufficent volume to 'justify' the wide price range you have to ask yourself who is in a positiuon to manufacture this situation, why and what do you think they are going to do next. Part of the answer of course lies in the context within which the low volume high range bar occurs.

 

[firewalker99]

A wide range does not necessarily indicate anything about the spread. Spread could be wide too, but it doesn't have to be. This is all possibly secondary when you consider what mechanics have to come into to play to allow this scenario to develop.

If range is wide what is it that is making the price move out of proportion to volume? If there's no volume or insufficent volume to 'justify' the wide price range you have to ask yourself who is in a position to manufacture this situation, why and what do you think they are going to do next. Part of the answer of course lies in the context within which the low volume high range bar occurs.

You're obviously right, context is always important to take into account.
It's just that in my (limited) experience I have seen little of situations where the volume is thin and the range is wide. It could help clarify this scenario if you (or anybody else) could perhaps give an example of such a situation.

Thanks for the reply already and a lot of useful posts, I'm having trouble keeping up with the excellent posts you've been spreading around lately great stuff!

 

[TheBramble]

Inverse Volatility Function:-

(VOLUME/Mov(VOLUME,5,S))*((H-L)/Mov((H-L),5,S))

OK my earlier post had this wrong way round which is pretty much for the course really when doing anything from memory. It's the current volume divided by the 5 bar simple moving average of volume DIVIDED by the current range divided by the 5-bar simple moving average of range.

 

[egro1egro]

Hi!

Excellent posts, Bramble!

I've tried your indicator with stocks and here is my take on it.

1. There are 2 types of stocks: with volatile volume and with quiet volume. Big caps usually have similar volume from day to day (i.e. "quiet volume"). A little less a litle more - peaks and troughs are random. Perhaps, accumulations/distributions are less visible on them, company's activities are not surprise, or wrong time-frame (smaller time frames can exhibit bigger differences in volume at different candles). Thinking of price mechanics for big caps/blue chips,
I must admit, I don't understand how accumulation/distribution/manipulation can happen very often on them for several reasons:
a) one needs _VERY_ big capital to do manipulations, and very clear intentions. But big caps are traded programmatically, for balancing positions.
b) accumulation can be achieved before release of some important info. For big companies it is impossible not to leak info. Less opportunity for sharp movements. Only on global news sharp movements are possible, that affect all market, but in this case volume will be very unspecific.
c) real accumulation can happen on a smaller scale - smaller money, smaller TF, etc.. Reading from daily charts might be difficult.

So, trying to express difference between "quiet volume" and "volatile volume" in precise terms I can suggest either variance/stdev of volume over some period as an additional factor to the formula. Perhaps, to spread lower band from upper band further if the stdev is not big enough. Otherwise, the bands force you to exit earlier in many cases. Another approach is to measure volume outliers - they are more outstanding for small/mid caps. Someone suggested to put BB on volume and pay attention only to those days that are above/below stdev=2.0 bands.

2. There are zero volume days or low volume days for some reasons. For example, Fridays can always be quiet etc.. If we look at smaller TF, then trading ceases at lunch time. Somehow you need to exclude these periods from the formula or it becomes choppy.

3. Different TFs need different factors. On weekly charts the bands are to close to the median price. On hourly - too far. Even for daily the bands can be too close/too far. Perhaps, you are using in on a certain type of stocks? With certain technical characteristics (say, certain volatility, average daily range above certain level, in a special stage etc..).

 

[egro1egro]

Remember as well - an important factor when talking about big caps is the free float (i.e. how many of the total number of issued shares are actually freely available and trading in the market (as opposed to sitting in a pension fund because it's inclusion is demanded by the fund's benchmark for example).

GJ

Yes, very appropriate remark!
Indeed, one needs to pay attention on free float and other parameters. I would also look at average (over a month or a week) volume traded per day * 360 / free float. For most stocks, this number is under/near 1.0 (i.e. yearly volume is less than total number of freely traded shares). Also, different volume reporting standards on Nasdaq, NYSE, AMEX mean that 2-3 times more reported volume is required on NYSE in order to bring about the same level of liquidity as on Nasdaq (just look at NYSE-stocks charts with their gaps and jumps). And, therefore, more liquidity is required to prevent price manipulations.
I've just chosen a term "big caps" to say in short what I meant when talking about lazy volume.