Volume Splitter

[BlowFish 10]

AgentKay,

 

there is nothing wrong with 'sampling' price data to help make sense of it. How you choose to sample that data is of course somewhat arbitrary. Firstly the variable you hold constant (time, range, volume) Secondly the sample size you use for that arbitrary constant (e.g. 5min 4points range 1000contracts). Same holds true for point and figure, renko, market profile, volume profile and any other charting method I can think of. (I'd be delighted to find a truly unique way of sampling data if you have one )

 

At its very simplest sampling allows you to determine trend. If the price now is higher than it was when you sampled it an hour ago (or perhaps the close yesterday) you might say the hourly (or daily) trend is up. How can you determine tend without sampling data?

 

Agent Kay how do you make any sense of the price data without sampling it somehow? Maybe just by watching the tape stream by (though you will still be constantly sampling and comparing in your head). Or are you saying the only meaningful sample is comparing price now to yesterdays high/low/close (I guess you mean an actual 'session'). Using that argument a weekly sample would make much more sense with 24 hour trading and markets globally correlated.

 

I guess you are being provocative to get people to think but you have 'over stated the case' imo. Unless you make decisions based on ...I dunno lets say planetary movements for example? You are absolutely going to need to sample price data and when you do that you are going to absolutely need to make arbitrary decisions on how how you sample it and what you hold constant.

[AgeKay 10]

  pimind said:

By your comments I assume that you only use the tape ie, time and sales. Although some may be able to trade without bars or indicators, others myself included need a graphical representation of what is going on in the market. I'm sure that we would all appreciate any insight that you may have that will add to this thread. Slamming an approach is easy trying to add to it and increase its value by maybe taking it in another direction has value.

 

I do use a graphical representation of trades but one that makes sense to me (which I have not seen anywhere). My intend is not to slam an approach, just to point out its flaws which is not being done enough in my opinion. It seems that anyone can come up with an approach, give it a fancy name and say Jesus used it and no one will question it at all. I know I might sound rude at times, but hope it will make people really think about trading.

 

  pimind said:

Again if you have any positive contributions that will assist in developing this indicator, "the purpose of this tread" then it would be greatly appreciated, otherwise i'm sure there are other threads that discuss the legitimacy of indicators and bars.

 

Ok, I'll suggest you use it to predict the next inside market before it changes and use it as confirmation to stay in a trade or exit.

 

  RichardTodd said:

Imagine how much more helpful it would be if you simply posted an example of what you'd consider good analysis. Since you seem very sure of yourself, I'm sure it would be of interest to all. More interest than essentially calling everyone an idiot, at any rate.

 

I don't want to post my charts, sorry.

 

  Tasuki said:

Out of curiousity, AgeKay, how do you take money out of the markets? That's not a facetious question, but a serious one. What is your methodology? In an earlier post, I was encouraging an interdisciplinary approach to the markets. So, what has meaning for you?

 

I am not going to tell you. I think everyone has to find it for himself. I know, it's a cliche, but it's true. I've posted in other threads what I think makes sense. If you don't think you would have come up with an approach, then I don't think you should use it. Observing the market, would you have transformed a series of individual trades which contain 4 pieces of information (price, time, size, side) into bars, candle sticks, MACD, VSA or other indicators/approaches? If not (which is highly likely) then there must be a reason (yes, we are naturally not stupid, but we like to believe other stupid ideas). No one is going to tell you there method if it makes them shitloads of money which is the reason you don't read about top trader's methods.

 

  BlowFish said:

there is nothing wrong with 'sampling' price data to help make sense of it. How you choose to sample that data is of course somewhat arbitrary. Firstly the variable you hold constant (time, range, volume) Secondly the sample size you use for that arbitrary constant (e.g. 5min 4points range 1000contracts). Same holds true for point and figure, renko, market profile, volume profile and any other charting method I can think of. (I'd be delighted to find a truly unique way of sampling data if you have one )

 

I do not sample data and none of my indicators use parameters/variables (this is actually one of my criterias, I do not use it if it requires a parameter). Your jaw would drop at how obvious/natural it is and how much easier it makes profitable trading decisions.

 

  BlowFish said:

Agent Kay how do you make any sense of the price data without sampling it somehow? Maybe just by watching the tape stream by (though you will still be constantly sampling and comparing in your head). Or are you saying the only meaningful sample is comparing price now to yesterdays high/low/close (I guess you mean an actual 'session'). Using that argument a weekly sample would make much more sense with 24 hour trading and markets globally correlated.

 

My price charts do not use bars or candles and my time charts have no resolution or you could say they have infinite resolution. I trade very short term and do not hold positions overnight.

 

  BlowFish said:

I guess you are being provocative to get people to think but you have 'over stated the case' imo. Unless you make decisions based on ...I dunno lets say planetary movements for example? You are absolutely going to need to sample price data and when you do that you are going to absolutely need to make arbitrary decisions on how how you sample it and what you hold constant.

 

Yes, I try to be provocative because it's a very sad state the trading education and software is in right now. I have wasted more than 2 years reading about traditional trading advice before I started to really think about it and come up with my own approach which took a lot of research, observation and thinking.

 

This is why I loved UrmaBlume's contributions. They are original and inspirational. He does not spoon-feed you and of course I wished he divulged more information, but there is a reason he does not. He is making a lot of money from it. Why give away your hard work?

 

And no, I do not see what effect planetory movements could have on the markets.

Edited June 23, 2009 by AgeKay

[BlowFish 10]

Well you have piqued my interest (again). How about a tiny hard to make out screen shot

[BlowFish 10]

As you know I also enjoyed UB's stuff but that is all about sampling data (with a bit of post and maybe pre processing).

[pimind 10]

I have a trading method that is superior to any other idiot method that everyone uses but i'm not going to share it i am just going to let everyone know that theirs are wrong and meaningless, now knowing that you are doing things wrong because i said so, come up with the right way of doing it. I'm sorry for the sarcasm but is that what you are saying.

[taotree 10]

A little harsh there. I understand how you feel, but... I'm very grateful that these people are here sharing some info, even if they don't tell us all.

 

It does reinforce some of my ideas of where I'm heading with my own work. I have been developing my own software since I haven't seen any current software that does anything near what I think is ideal.

 

I was going to include a visualization I did some time ago but it's not executable at the moment. Each transaction is plotted as a circle with radius depending on size of transaction. The overall window is zoomable in/out infinitely (using mouse wheel).

 

Perhaps by saying you have to come up with it yourself, AgeKay is saying you have to think creatively? Or maybe... to understand how to use something effectively, you have to have the level of thinking that can come up with it yourself.

 

Most recently, I've been considering: what about defining a "bar" as a movement of bid/ask. The current value within the bar equals the average transaction price (weighted by volume) in the bar--so as more transactions occur at ask, the plot approaches the ask. Or plot each trade as a line following that. This fits with what AgeKay was saying about not having any parameter at all.

 

Anyway... a little off topic, but I think pertinent to the way the discussion has turned. And could be specifically applicable to "volume splitter" by considering... what are some effective visualizations that could be applied to the concept of volume splitting? I find it interesting that instead of considering, what is the best way to consider this information, it has been converted into a MACD--a very familiar tool indeed. Do we have a tendency to take any new thing and fit it into the familiar? That may be ok initially, but do we let that hinder us from expanding to new ideas? And how much is that limitation dictated by charting software capabilities?

 

Ok, now let's start discussing Sapir-Worf as it applies to traders thinking being limited by the software/tools they are familiar with just kidding

[pimind 10]

taotree i agree that i may have been a little harsh but it's how they came across, i suggested in a previous post that instead of trying to replicate something we should try to develop a volume indicator that can shed some light on o what the market may do next, I think that volume is extremely important.

 

I agree with coming up with a way to show this information in a way that is not currently. and although the macd version seems to give signals at times, i think it take us a little away from pure volume.

 

As small traders we don't have the resources that are available to larger traders and firms , both monetary and technical. So i think that the premise is to take the path of least resistance and go with the larger traders. and not trade against them, especially in the short term a 1-5 minutes anyway.

 

So for the purpose o developing such a tool which , if it not the intent of this thread i would be happy to start one. We could define what is important.

 

personally I would like the following.

 

Trades at bid and ask only by large traders, size of trades, # of trades, a way of showing velocity or thrust, (how fast) if possible.

[mrhomsi 10]

hi all , it is really an amazing thread

 

can anyone put the code of vs_macd2 so i can translate to e-signal language?

 

hint : in esignal you can get the last volume traded separately if it was on bid or ask which i think will help to much

 

getMostRecentAsk()

 

getMostRecentAskSize()

 

getMostRecentBid()

 

getMostRecentBidSize()

 

getMostRecentTrade()

 

getMostRecentTradeSize()

 

this function you can get from efs reference source for time and sale functions

[pimind 10]

Here you go:

 

{Phall VS_MACD
volsplitter version 2, using close compare
fix for Block reset implemented 6/17/09}

{ PLOTS VERY SIMPLE MACD} 


inputs: 	
Length1(6),
Length2(10),
MACD_Length(3),
SmallBLockFilter(99),
LargeBlockFilter(9999),
UpColor(Darkgreen),
DownColor(Darkred);


variables: 	
	MyVol(0), 
color(yellow), 
avg(0),
intrabarpersist MyCurrentBar(0), 
  	intrabarpersist VolTmp(0), 	
intrabarpersist DeltaCu (0), 	
intrabarpersist DeltaH (0), 
intrabarpersist DeltaL (0), 
intrabarpersist DeltaO (0),  
intrabarpersist Block(0),
intrabarpersist LastClose(0);	

if LastBarOnChart then 
begin    	
	MyVol = Iff(BarType < 2, Ticks, Volume); 

	if CurrentBar > MyCurrentBar then 
		begin 		
			VolTmp = 0; 	
			MyCurrentBar = CurrentBar; 	
			DeltaO = DeltaCu; 	
			DeltaH = DeltaCu; 	
			DeltaL = DeltaCu; 
		end; 	

	Block=absvalue(MyVol-VolTmp);
 		VolTmp = MyVol ; {rest block}

	if Block >= smallBlockFilter and Block <= LargeBlockFilter then 
		begin 		
			if Close < LastClose then	DeltaCu  = DeltaCu - Block
			else if Close > LastClose then DeltaCu = DeltaCu + Block;
			LastClose=Close;
		end; 

	end ;   


DeltaH = maxlist(DeltaH, DeltaCu); 
DeltaL = minlist(DeltaL, DeltaCu);    
if DeltaCu <= DeltaO then color = DownColor 
	else color = UpColor;  	




avg=(DeltaH+DeltaL+DeltaCu)/3;

value1=xaverage(MACD(avg,length1,length2),MACD_Length);


Plot1(value1,"MACD");
Plot2(0,"ref",Yellow);

[AgeKay 10]

  taotree said:

Each transaction is plotted as a circle with radius depending on size of transaction. The overall window is zoomable in/out infinitely (using mouse wheel).

 

You can see this being done in FutureScalper without the zooming though (there are a few videos on YouTube).

 

  taotree said:

Perhaps by saying you have to come up with it yourself, AgeKay is saying you have to think creatively? Or maybe... to understand how to use something effectively, you have to have the level of thinking that can come up with it yourself.

 

Exactly.

 

  taotree said:

Most recently, I've been considering: what about defining a "bar" as a movement of bid/ask. The current value within the bar equals the average transaction price (weighted by volume) in the bar--so as more transactions occur at ask, the plot approaches the ask. Or plot each trade as a line following that. This fits with what AgeKay was saying about not having any parameter at all.

 

This is what I am talking about. Be creative. Think outside of the box. Discard everything you know and start from scratch.

 

  taotree said:

Do we have a tendency to take any new thing and fit it into the familiar? That may be ok initially, but do we let that hinder us from expanding to new ideas? And how much is that limitation dictated by charting software capabilities?

 

This is what I think too.

[pimind 10]

I ran this today on a 5 minute chart just to get an idea of the blocks of traders.

 

Look at the relationship of the increase in price with the bid ask volume and the block traders.

 

 

 

This is 2 long a time frame to scalp, but i just wanted to see the data.

[RichardTodd 10]

  pimind said:

Here you go:

 

 ... omitted ...

 

hmmm.. if I'm reading this correctly then this code will ignore the second of two block trades in a row at the same price? And it decides up vs down by comparing each block trade to the last block trade... I would think if you are going down this road that it'd be better to guess upticks vs downticks on every trade, and then only count the ones that happen to be big blocks...

 

something along these (pseudo-code) lines would be my suggestion... hopefully it is clear enough:

 

intrabarpersist bool uptick(true);

// guess that close above last close is 
// an uptick, and vice-versa
if (close > lastclose) then uptick = true;
if (close < lastclose) then uptick = false;

// ... and implicitly if close == lastclose then uptick should still 
// be whatever it was last time price changed

lastclose = close;  // remember for next time

if(trade is large) then begin
if(uptick) count as upblock
else count as downblock
end

 

You might want to try something like that, anyway. Actually at least on the e-minis it's probably a pretty good heuristic.

[phall 10]

  RichardTodd said:

hmmm.. if I'm reading this correctly then this code will ignore the second of two block trades in a row at the same price? And it decides up vs down by comparing each block trade to the last block trade... I would think if you are going down this road that it'd be better to guess upticks vs downticks on every trade, and then only count the ones that happen to be big blocks...

 

something along these (pseudo-code) lines would be my suggestion... hopefully it is clear enough:

 

intrabarpersist bool uptick(true);

// guess that close above last close is 
// an uptick, and vice-versa
if (close > lastclose) then uptick = true;
if (close < lastclose) then uptick = false;

// ... and implicitly if close == lastclose then uptick should still 
// be whatever it was last time price changed

lastclose = close;  // remember for next time

if(trade is large) then begin
if(uptick) count as upblock
else count as downblock
end

 

You might want to try something like that, anyway. Actually at least on the e-minis it's probably a pretty good heuristic.

 

RT:

yes you are reading correctly; the same blocks are ignored in this original version because we were interested in the delta's like Blowfish4. if you want the absolute totals, absolutely it must written as you say.

BTW, this is why the "Simple MACD" works best on share bars, as it forces the delta onto a fixed block of shares every bar for averaging in the MACD. on time charts with highly variant volume, things don't work as well...

 

phall

[RichardTodd 10]

  phall said:

RT:

yes you are reading correctly; the same blocks are ignored in this original version because we were interested in the delta's like Blowfish4.

 

What's the advantage of ignoring the subsequent block trades? If someone hammers away at a price level 100's of contracts at a time, I can't think of a reason off-hand that I wouldn't want to count them all.

[Tasuki 10]

  RichardTodd said:

What's the advantage of ignoring the subsequent block trades? If someone hammers away at a price level 100's of contracts at a time, I can't think of a reason off-hand that I wouldn't want to count them all.

 

Richard's right, I think. Ben Lichtenstein of Squawkbox has mentioned many times that the floor traders in the S&P pit will hedge their positions in the big contract by taking 5X positions in the ES (since the big contract is five times larger than the emini). From what I gather from Ben, the floor traders don't place these trades themselves, but rather they have headsets and call over to an employee to place the ES emini trades for them, who place multiple large-lot contracts in a row, until the position is sufficiently hedged.

 

If there is some way to change this feature of the VS_MACD2, it would be a good idea to go ahead and do so (wish I knew how to do it myself, but I'm just a newbie greenhorn at Easylanguage).

[BlowFish 10]

Yes not a bad Heuristic with the great advantage of working with historical data. A reasonable compromise.

 

With the direction of recent talk people might want to revisit (or visit) UmaBlumes trade intensity thread. Worth considering aggregating (into a block) seperate ticks that arrive within a certain definable delta time (delta referring to t1..tn ). I have been meaning to combine something like that to the large block concept in Ninja but was kind of waiting for V7.0 and have been rather busy playing MMORPGs

Edited June 24, 2009 by BlowFish

[AgeKay 10]

  BlowFish said:

As you know I also enjoyed UB's stuff but that is all about sampling data (with a bit of post and maybe pre processing).

 

No, he does not sample data, he processes every trade. In his most recent post on the "Trade Intensity" thread he explicitely says

  UrmaBlume said:

BTW - it makes no difference how TS or anybody else breaks up the volume bars as while the chart is presented on a 1k contract price bar chart the data for the indicator comes from elsewhere.

[BlowFish 10]

No he samples into 100 (or greater) volume chunks. Whether your background is stats or signal processing that is sampling. He goes to some length to point out why sampling with respect to a constant volume is his prefered approach. (All the studies in this thread process every tick and every contract whilst they are running).

 

Maybe it is just a question of semantics. You can still process all the data but sample it. Unless you use every tick of the ES since the day of its inception and look at that as a whole you are sampling. Also he is sampling by looking at V@Bid and V@Ask and dividing the population into 2 samples by that particular property.

 

Anytime you divide the population (of every single ES tick) into 2 or more 'chunks' (be it by V@bid/ask, X contracts, Y-Z range, t minutes, n ticks, d days, y years) you are sampling.

 

Cheers.

[pimind 10]

I was looking around and found something that may be of interest and help us in our development. I found fxadvanced.com and they have a trade intensity and volumestrength indicator, i just found it so im not familiar with it. I also found a site volumementum.com that has a volume momentum indicator that may be worth looking at.

[BlowFish 10]

As an aside I posted a couple of charts here http://www.traderslaboratory.com/forums/104/open-free-discussion-volume-6180-4.html#post68997 as they are using pure volume seemed like a better place for them. There is a point to liking that post..... you can get by with pretty simple tools. Having said that I am still committed to finding better order flow / buying selling pressure/ intraday 'open interest' tools. The thing is you can get pretty comparable levels of clarity with 1st generation tools if you know how to use them. (perhaps more importantly where to use them)

[RichardTodd 10]

  BlowFish said:

(All the studies in this thread process every tick and every contract whilst they are running).

 

Exactly, though to be fair all the indicators I've seen here have grouped the data into bar-sized groups and then averaged/macd'd/etc them. Still, I'm not sure why this particular thread was hijacked to talk about the usefulness of statistical sampling.

 

I've done some indicators recently that operate on the last x ticks regardless of what bars I'm looking at. It's nice on some indicators to avoid the discontinuity at each fresh bar. Of course it makes the programming more tedious, and then the charting platform snapshots it at bar intervals as the chart progresses, so you are back to a sampled signal in the end.

 

Anyway, I'd love to see threads about nonstandard chart types and ways to avoid sampling. I'm not sure what's so novel about nonparametric studies but even that might be interesting. Just not sure why it's here in a thread about a volume study.

[BlowFish 10]

  RichardTodd said:

 

I've done some indicators recently that operate on the last x ticks regardless of what bars I'm looking at. It's nice on some indicators to avoid the discontinuity at each fresh bar. Of course it makes the programming more tedious, and then the charting platform snapshots it at bar intervals as the chart progresses, so you are back to a sampled signal in the end.

 

 

'Sliding Windows' are another interesting way of looking at data. Mind you this is how some traditional indicators work, a simple moving average adds a new datum to the front and throws one away at the end.

[RichardTodd 10]

  BlowFish said:

'Sliding Windows' are another interesting way of looking at data. Mind you this is how some traditional indicators work, a simple moving average adds a new datum to the front and throws one away at the end.

 

Yeah, I wasn't trying to be novel or "provocative"

[phall 10]

  RichardTodd said:

What's the advantage of ignoring the subsequent block trades? If someone hammers away at a price level 100's of contracts at a time, I can't think of a reason off-hand that I wouldn't want to count them all.

 

yes; but how do you know the direction of the subsequent trades at the same price? are they biased up (buys), biased down(sells)? this is the age old problem with these methods; even when using bid/ask you don't truly know whether to bookkeep the trades as buys or sells....

 

after we realized that the bid/ask was asynchronous, i went back to closes and made the assumption that what we needed was something that looked like Pressure up/Pressure down.

 

i concluded that the most probable(but certainly not guaranteed) of all the options in bookkeeping a trade as a buy or sell was to assume that if the price moves up on a big trade, that's a buy (and vice versa). I ignored the rest as it's hard to assume that they are anything more than randomly distributed buys and sells....

 

this basic "engine" generates a OHLC information for the period of this "close pressure". the idea is/was to apply this delta "pressure" to whatever technique the user thinks is appropriate.

 

As i'm sure you appreciate, beauty is in the eye of the beholder; everyone will like to modify it to what they think it describes... this was just my take.

 

phall

[RichardTodd 10]

  phall said:

yes; but how do you know the direction of the subsequent trades at the same price?

 

You guess. I would suggest you assume that an uptick followed by a zero tick is two buys, because when you watch the tape this is statistically far more common. If you update your guess on every trade as in the pseudocode I gave, and not just the big trades, I expect that you will rarely be wrong on the eminis.

 

The biggest problem with only guessing the up vs. downtick on big trades is that for certain settings, you'll only get a big trade every few minutes. Price can move a lot in between... Say someone buys 800 contracts at 903. Then five minutes later someone buys 900 contracts at 902.50. Well... if I read it correctly, the algorithm in the macd2 indicator I saw would call the second trade a sell because just because it was lower than the last block trade. I feel it's much more accurate to keep track of the best guess upticks vs. downticks on every trade, and then only count the big ones.

 

It's only a suggestion, of course. I'm quite happy with the splitter I have, so...