Price and Time data set

Quote from Blotto:

Yes there were. Concerning the state of supply at that point in the market and how certain categories of participant would predictably trade next.

I'll give some pointers on this. First, it would be helpful to know what has happened in the past in a similar condition. I would suggest creating a historical database of intraday market data for the instrument you trade, and sorting it according to various parameters. You need to obtain the data, and create a way of sorting and organising it logically. You can then start to see similar conditions, and compare like with like.

So I could ask what does the ES typically do in the last 30 minutes of the session when we are X from a high (volatility dependent), down on the day, and some other supporting conditions are in force (to do with both the prior action say from lunchtime EST onwards and the location in the bigger cycle). I'll pick an example at random from my historical database for this sort of query. I can produce a 1 minute chart for you. The date is October 15th 2007. Visually alone you will see the similarities. Going back on a daily chart, you will see similarities also with the condition in force and the market continuing below the low the following day.

This isn't about visuals, it is about underlying conditions and the prices / quantities players are willing and able to transact at any given time. Prices and times of course play a factor, and I'd suggest specialising in one market and getting familiar with the historical conditions, repeating occurrences, try to logically deduce and reason why X is more likely to happen then not happen given supporting parameters. This knowledge can later be applied to other markets, using the same approach.

So there are a few questions we can ask which might be of assistance:

-is price action alone sufficient to predict future prices to a high degree of accuracy?
-how do we consider time, if at all?
-do we consider volume to be relevant? why? what does it represent?
-could there be benefit in organising and categorising prior market data to look for repeating conditions?
-do we need more information than just what we can see visually - time and sales data either in raw form or presented in a chart - ie should we be doing some sort of analysis based on how traders are positioned etc, and how can we quantify that?
-how players are positioned - to what degree does this influence how the market will progress next? is there predictive value here?
-are some players "more equal" than others due to size, access to information, etc and if so should we respect their participation (to follow perhaps but not trade against)?
-do some players make predictable decisions and can we exploit this for profit?


Some starting points. I just wanted to give an alternative viewpoint to the one which says "price is random" and "I trade as if price cannot be predicted". It is possible under certain conditions to predict the next N (minutes|seconds|microseconds) with a high degree of accuracy, or the next N ticks if you prefer. We have HFT guys predicting inside of seconds and very short term trades inside of minutes, and traders like me playing the next 10-30 minutes, and sometimes longer. Horses for courses, but it all involves prediction of some degree. I would argue that deciding that price is random AND attempting to trade profitably is somewhat of a contradiction, and hope that my posts have showed some alternative possibilities.

There are reasons for the time I mentioned, and the 8 point rally. Thorough testing of these conditions and prior price movement may reveal the appropriate parameters. It would not be beyond an independent trader to discover similar theories and implementations which yield similar results. I'd suggest analysing durations and magnitutes of price movement, ratios, degrees of movement within prior ranges, frequency and timing of turning points, market microstructure around turning points (with high quality accurate time and sales data), etc. Plenty to be investigated and discovered. This talk of "higher highs" "lower highs" etc does not do justice to the full range of possibilities to be explored and is frankly a one dimensional dead end which will not lead to consistently accurate trading.

Happy to discuss further if there is interest and some feedback / interesting questions. Of course cannot give too much away. Just providing some starting points, there are many ways to trade, what I propose is just one.
More...
Link : http://www.elitetrader.com/vb/showthread.php?s=&postid=3807643#post3807643

I copy/past this post of Blotto here as I think the questions raised in his message are interesting and may help to go forward, as "starting points" as he says.
But then I guess the difficulty will be also in the organisation of the datas gathered and how apply filters to obtain the desired information.
Valuable posts like these ones are rare. :)

030985
 
030985,

Like I few others, I have detailed catalogs of a few member's posts, one starting with the letter 'c' and another starting with the letter 'b.'

I have spent quite a considerable amount of time tracking my own price/time data in attempts to make sense of how to use it. I have many years of this data.

I've created various forms of indicators to capture this information, also attempting to parse/sort it in useful ways.

So a few things that I am doing......

I am using this on FX data, as data is easily available, and gathering data for several years is no problem. I am unsure if this is a problem for currencies, as most of the members who speak about the time/price data are trading futures.

My indicators have a bit of intelligence built in as to be able to track the gyrations in a way the price actually moved. To that effect, I think I have some of the best zigzags in use. Usually big outside bars skew what price actually did to most zigzags, mine sorts that out close to 100% of the time.

Next, other members often speak of gyrations..... I spent quite a bit of time with this, but I have been unable to really sort that out. So instead, I would say I mainly look at HHs and LLs... and the continuations that occur within. Particularly the continuations that occur withing the Main Move of the day. This is where I attempt to put most of my focus. In other words, I'm looking to try to trade price as it is going to make a HH or LL, and attempt to capture a percentage of that swing, particularly if price is less than the average 3-Day-Range.

For example.... I am attaching a picture, "3dr," in this picture, what you are seeing are years of data for how OFTEN price moves to the 3DR. What you can see is the where the majority of days lie, in that the days will usually attempt to reach their 3DayRange...... i.e. if I'm looking to trade, and we're beginning to bump against that, I usually am less excited about it.

Another idea, is about timing windows. With Time/Price data, its easy to sort this information. Here I will attach a picture called "Timings," where you can see how the Main Move of the day is 'usually' created. I've put it into an easy to read format, and labeled it which I think should be pretty self explanatory.

Another thought is about templates, and HOW price will make its Main Move..... for example, if you look at today on the EU, price made its reversal right around the 950 minute mark. By that I mean relative to the open. For currencies I like to set the open at the start of the Sydney session, and track clock time up to 24 hours after that. A common template I have observed is that price will often be making daily lows up until around the 900-1100 minute mark after the open.... and move swiftly up to make new daily highs. At that point, there will usually be a sharp drop to make new lows. I liek these days and have been getting better at trading those. Today was such a day, another example of this day on this currency was the 5th of this month.

These are just a few thoughts, as it seems like most people are laughing at the idea of tracking price/time data. Understandable as it surely is not easy to make sense of, and I am still waaaaaaay long away from figuring it out, but I can see the benefits of it, and am a bit weary about posting too much of my information here.

Here is an example chart of what I track live, this is on a 15minute chart (posting 15 minute so I can fit more days into a screenshot), not what I use, but it should give an idea of what I look at throughout the day. Similar concept to what member "C" spoke of, with various lengths, Tics on offer, etc..... you can see here mine. Take note of the swings.... various datas I like to have displayed in real-time on my chart above the swings, and also the Main Move as it is forming throughout the day to know where I am.
 

Attachments

Interesting work, forrestang.

I came a way myself in organizing this type of data and actually developed an application to help me better sort out the statistics, but I was not able to find any advantage from it yet.

Sure, I could use it to track market metrics, develop statistics and calculate odds for certain scenarios, but nothing close to the kind of accuracy Cheese suggests, i.e., predicting the day in advance all the the way down to a micro level.

To be frank, I'm not sure if that's even possible.
 
Just just a bit more on the price/Time idea......

A current picture of today, until recently, the picture for the day was down, as the primary move was down, and looked like it was going to be a typical type of down move (other than the lack of range), and the day would peter out.

But then we have a new high on the day, and that makes the prior LOD come REALLY late in the day. For reasoning as to why I call it late, reference the prior picture I posted labeled timings to see why.

Now... since that has happened, the likely next turning window... or end of the day should be around 1300 minutes beyond the open... or very near the end of the day. But that is just according to the "Timing" chart.

To get more specific, will attach a picture in the next post, that will be a tad more specific. Something I have been working on, is showing visually the likely next place of stopping. In the pic, you can see several clusters of points where the day in the past has had a similar late timing for the lOD. Typically they finish anywhere from 30-75 tics beyond the open, and you can get a typical idea of the timing as well from this picture. As is now we are currently only 22 tics off the open.... which could be the end, but based on the timing aspect, I think there might be another impulse up?

Due to my lack of ability to incorporate clean gyrations into my trading, I still unfortunately rely on abcs/123s whatever you want to call them to trigger an entry, as I have stats on those.... for example I like to wait for minimum of 70% retraceemnts..... if you ask why I will attach a distribution of 123s that make HHs or LLs and you can see that the majorty of them fall in that zone. Unfortunately this means I miss opportunities too :( But it's a work in process.
 

Attachments

Associated clustering... the zone I pick is relatively arbitrary and I pick it by hand through my little data engine I've created in Matlab. There is likely an optimal way of comparing/sorting like days, but this is just one way of associated similar events in the past. But should give an idea.
 

Attachments

Quote from forrestang:

Now... since that has happened, the likely next turning window... or end of the day should be around 1300 minutes beyond the open... or very near the end of the day. But that is just according to the "Timing" chart.

.......Typically they finish anywhere from 30-75 tics beyond the open, and you can get a typical idea of the timing as well from this picture. As is now we are currently only 22 tics off the open.... which could be the end, but based on the timing aspect, I think there might be another impulse up?
[/B]
More...

LOL.... Holy shit!!!

So... Before we had that upmove, I suspected there would be another, and the typical objective I thought we'd accomplish. I mentioned 30-75...... but I didn't mention that larger data point you see on the picture..... I didn't mention it b/c I figured it was a bit of an outlier and unlikely. That data point was 1308 minutes and 150 tics beyond. But look where we finished on the day.... 1319 minutes and 154 tics! So wow.

Also I mentioned 1300 minutes.... new highs came in at 1268, 1304 and the max at 1319.... So not a bad guess I suppose?

Since using price/time data I haven't looked at news in quite a while, as it seems the anticipated pattern plays out regardless.... if anything the news announcement seems like it just makes your price objectives reach outliers instead of what is typical.

So had I been monitoiring the news, I may have thought of going for the more aggressive target. This is something I JUST thought of today with outliers and how they may be created. That data point I am calling an outlier was on the 17Dec2003, if anyone has access to back news or FOMC announcements..... I'd be curious to know if that date was also an FOMC announcement or other major event like it was today?
 

Attachments

Quote from forrestang:

Since using price/time data I haven't looked at news in quite a while, as it seems the anticipated pattern plays out regardless.... if anything the news announcement seems like it just makes your price objectives reach outliers instead of what is typical.
More...

That's what Cheese says.

Thanks for sharing.

Quote from Cheese:

Generally agree with your approach on news.
The 10am major economic news items I take account of but only as a distortion to the predicted days pattern (eg YM). FOMC too tends to make price swing big in the direction of the predicted pattern.

As you say keep it technical.
:)
More...
 
Thank you very much for your comment and sharing forrestang. It is really much appreciated. :)

I can see you went further with this Price and Time relationship. As for myself, I put it a bit on the side, trying to come back at the very basis of reading charts, as I did not achieve to see anything interesting with what I did on my spreadsheets. Now I may try 'again' to go further with these Price and Time thing, but certainly my lack of coding skills make me slow.

I am also now more interested in EURUSD as it was my main instrument before but I did not know how to manage the "no open/close" time issue, that is why I began to do researches with Dax. ( I think I said the same thing in a post on the first page of the thread ).

Do you think your choice of Sydney open as the starting of the day makes a great difference, or anything could do ?
With no evidence of it, I always thought that any 'starting' time can do for daily charts or whatever timeframe as long as you don't change every two weeks.
I am tempted to choose 0:00 EST, just for convenience regarding the hour changes ( DST ?) as EST is the time of the datas I can get.

It is funny you talk about your ZigZag indicator as I requested to make develop one ( as I could not do it myself ^^" ) in order to overcome the problem you evoke with standard ZigZag. 'Mine' do not miss some moves as certain ZigZag do, and can restart each day at the desired time. I have very basic information on each turn points, but a bit less than you have as I did not think about it when I requested the development.
My major problem is that I am required to enter these swings datas manually, as I can't export it directly, and that is what was a bit rebarbative and sometimes discourage me.
Also even if I could export datas easily in spreadsheets, I am not sure how to organize the swings datas relative to the minutes/time in order for them to be useful and to get the maximum information.

When you compare current day with past similar days, are you aware only of the 'intraday' behaviour of the price ( by what I mean, difference from open, high or low first, range, etc... ), or do you take into account for example the month we are in, previous day absolute and relative close and such ?

Regarding the 17th december 2003, it seems not to be FOMC statement as it was on 9th December. You can see here :
http://www.federalreserve.gov/newsevents/press/monetary/2003monetary.htm

:)

030985
 
You must log in or register to reply here.
Back
Top