Price and Time data set

Quote from forrestang:

So one thing I am wondering, is the key simply the proper aggregation and compilation of various statistics? I.e. is it just a matter of "8 out of 10 times price did this?"

-OR- Is it that once the data is organized in a specific matter, the path becomes obvious?

Then of course there is the matter of actually having useful metrics to track in the first place which I suppose is part of the task.
More...

It has been said that it certainly would not be that easy to 'solve' for an amateur, but that it would not necessarily require anything 'higher' than basic algebre/mathematics.

So I would tend to think it is a matter of 'playing' with different kinds of statistics BUT it will be somehow useless as long as you are not asking the right questions to progess in the solving of the puzzle as it seems the reasoning works then for any volatile enough markets.

And this is the really hard part. Because if I am an amateur struggling, it may be due precisely to my mind being not sharp enough for now to see the right things or ask the right questions, so then how can I make myself go on the right path in order to access at first to the right questions ?
I guess you know that you were asking the right questions only once you 'solved' it ? :D

Just some thinking :) .
 
Quote from 030985:

It has been said that it certainly would not be that easy to 'solve' for an amateur, but that it would not necessarily require anything 'higher' than basic algebre/mathematics.

So I would tend to think it is a matter of 'playing' with different kinds of statistics BUT it will be somehow useless as long as you are not asking the right questions to progess in the solving of the puzzle as it seems the reasoning works then for any volatile enough markets.
More...

I agree, I don't think any of it is going to be something that requires advanced math to accomplish.

What I was thinking with my question above, with organization vs. statistics.......

You can imagine how some Gann principles work, where it is essentially taking data on a properly 'squared' chart to look for turning points in time and price traveled. Using things like squares, or triangles. That would be an example of 'organization' of data. Just an example.

Using statistics would be more like taking averages, or distributions of various groups of data, and aggregating that into percentages of occurrence, or likely distance traveled.

As with the examples above, I could see either of these being properly displayed in a tabular format. I'm not saying either of these two examples are related to the material discussed here and in the threads related to this topic..... but just something I was thinking about as to figure out what the likely path or route is.
 
I concur that it will likely be less a matter of advanced math, but possibly pattern recognition (swing time of day, length, duration, etc.) and corresponding probabilities.

As an example, I had noticed that when TF (Russell 2000 futures) and corresponding ETFs experience a rapid sell off between the open and 11:30AM it very often puts in the low for the day. This doesn't occur with the frequency that it used to, and I suspect wasn't/isn't unique to that index, but it is the Russell 2K is the one with which I am most familiar..

Given your data set, you may be able to in/validate this and similar repeating patterns, and if validated calculate corresponding probabilities upon which to trade. Hence:

If YM declines P points (it's a sell off) within M minutes (it's a rapid sell off) before T time of day, go long after it retraces P percent (reversal) because there is R probability that the day's low was put in.

Your data set affords you an infinite number of similar probability calculations:

If you define a "swing" as being a minimum of 45 points, what is the probability of it becoming 100, 150, 200, 300, etc? Barring outliers such as a flash crash, there is some maximum for which it would be nice to know the probability.

Can the duration of the swing provide any meaningful insights? Does a 100 point swing occurring within 2 minutes provide anything predictable about the probability it will become 200 points and when, versus a 100 point swing that takes 2 hours to develop?

So, does swing length, time of day, and/or duration provide any predictability about future prices?

Considering the chart posted by Zen Student, it appears he has a method for predicting when a swing has come to an end and a new one in the opposite direction is beginning, and therefore one would want to reverse their position to take advantage of the next swing. Whether he is calculating the probability of a new swing based on percentage retracement, swing duration, or something else, I don't know. But obviously that would be very valuable to predict with reasonable accuracy and might be hidden somewhere in your data just waiting to be discovered.

Just my (amateur) thoughts, for what they're worth.


:)
 
Quote from csjohnson65:

I concur that it will likely be less a matter of advanced math, but possibly pattern recognition (swing time of day, length, duration, etc.) and corresponding probabilities.

As an example, I had noticed that when TF (Russell 2000 futures) and corresponding ETFs experience a rapid sell off between the open and 11:30AM it very often puts in the low for the day. This doesn't occur with the frequency that it used to, and I suspect wasn't/isn't unique to that index, but it is the Russell 2K is the one with which I am most familiar..

Given your data set, you may be able to in/validate this and similar repeating patterns, and if validated calculate corresponding probabilities upon which to trade. Hence:

If YM declines P points (it's a sell off) within M minutes (it's a rapid sell off) before T time of day, go long after it retraces P percent (reversal) because there is R probability that the day's low was put in.

Your data set affords you an infinite number of similar probability calculations:

If you define a "swing" as being a minimum of 45 points, what is the probability of it becoming 100, 150, 200, 300, etc? Barring outliers such as a flash crash, there is some maximum for which it would be nice to know the probability.

Can the duration of the swing provide any meaningful insights? Does a 100 point swing occurring within 2 minutes provide anything predictable about the probability it will become 200 points and when, versus a 100 point swing that takes 2 hours to develop?

So, does swing length, time of day, and/or duration provide any predictability about future prices?

Considering the chart posted by Zen Student, it appears he has a method for predicting when a swing has come to an end and a new one in the opposite direction is beginning, and therefore one would want to reverse their position to take advantage of the next swing. Whether he is calculating the probability of a new swing based on percentage retracement, swing duration, or something else, I don't know. But obviously that would be very valuable to predict with reasonable accuracy and might be hidden somewhere in your data just waiting to be discovered.

Just my (amateur) thoughts, for what they're worth.


:)
More...

Interesting post, thank you :)

One thing which has intrigued me and tried to work on it for now more than finding the turns of swings and their potential sizes, it is the faculty to know if the day ahead will have its low coming before its high ( Buy Sell day in Zen Student's or Cheese's terms ) or its high coming before the low ( Sell Buy day ).
I tried different combinations and filters but did not come across something reliable.
Maybe I do not enough datas as some refinements give only one occurence. I managed to have nearly 6 years for now.

It also seems that it can be predicted the day before ( at the end of the session ), meaning that a gap may not be meaningful to determine this parameter.
 
I have also spent quite a bit of time looking at premarket, London and other parameters to be able to somewhat forcast the day ahead in some way as well..... wether it just be the type of day, or a range of sorts.

What instrument are you looking at with 6 years of data?
 
Quote from forrestang:

I have also spent quite a bit of time looking at premarket, London and other parameters to be able to somewhat forcast the day ahead in some way as well..... wether it just be the type of day, or a range of sorts.

What instrument are you looking at with 6 years of data?
More...

I am focusing on FDAX datas for now. I also looked a bit at CL before that ( with 'european session' and 'RTH' US session ).
I would have looked at EUR/USD pair too as it was the main instrument I have traded so far, but I was not sure of how interpret datas as there is no "RTH" though I guess it does not change anything much.
 
Quote from 030985:

...the faculty to know if the day ahead will have its low coming before its high ( Buy Sell day in Zen Student's or Cheese's terms ) or its high coming before the low ( Sell Buy day )...
More...

I've also read their posts, though have not yet begun constructing the necessary data set. The folks in this thread are ahead of me in that respect, but I can provide my what I have gleaned from their posts.

In one of his posts Cheese highlighted an observation that the present day was a gap down day, and that on the previous four Mondays where the market closed down the session prior, and gapped down at the open, the first swing was down. I also noted in a number of his posts that while he was trading YM it appeared that he used the INDU as his point of reference (possibly because there is a 15 minute disparity between the close of the index versus the futures and the index close is receives the most attention, or maybe he's using YM and INDU interchangeably and I'm over-analyzing.)

So his data set includes the Day of Week, and he is able to sort/filter on it. Another obvious filter might is the size of the gap, as I would presume that the larger the gap the more probable the market will continue in the direction of the gap.

I also noted, however, that the gap is but one determinant, and often not the most important, so a key is knowing when it is predictive and when it is not. He actually emphasizes using the previous close to predict the next sessions open, and sometimes there isn't a significant gap.

It still seems to me that he is observing recent market patterns by sorting and filtering his data (easy to do in Excel), and once those patterns establish themselves, they tend to recur for some period of time during which we can take advantage. So maybe in addition to whether the previous session closed up or down, it is also important whether the last swing was up or down. The last swing of the last session could be a strong swing higher, even though the overall market closes lower on the day. Does that matter?

Do combinations of Prior Session Close (Up or Down), Prior Session Last Swing Direction, Current Session Gap Direction and/or the strength of each in points, or even Day of Week provide any predictability as to the direction of the first swing of the current session, and/or whether it is a Buy then Sell (BS) or Sell then Buy (SB) day?

It sure is easy to ask questions! But as Cheese suggests, are the the RIGHT questions? And then there's the harder part of collecting and crunching the data to find the answers.

Cheese, right questions or wrong questions? Warmer or colder? :D
 
Do combinations of Prior Session Close (Up or Down), Prior Session Last Swing Direction, Current Session Gap Direction and/or the strength of each in points, or even Day of Week provide any predictability as to the direction of the first swing of the current session, and/or whether it is a Buy then Sell (BS) or Sell then Buy (SB) day?
More...

I have done a bit of work on all the metrics you list here, save for the one I have highlighted there. Still looking into it, but so far I haven't found anything that tends to reliably predict a direction, or distance traversed at the open.
 
Maybe there isn't an "algorithm" that provides consistent predictability, but the ability to sort and filter the data set in search of recent patterns, and that is what provides useful insights on a consistent basis.

To quote Cheese, "On the last 4 occasions of Mondays being gap down with the previous close down, the Close on the day has been between 20 to 165 points down on the Opening Price (9.30am YM)."

Maybe it's knowing what questions to ask each day and having a data set that you can sort and filter to get the answers, rather than developing a static function. That approach would better account for changing market dynamics.

I recall a time where the Russell 2000 (TNA and TF) seemed to experience a strong and steady sell-off at the open, ultimately resulting in a sharp reversal back to around the opening price, then drifting sideways to somewhat lower during the mid-day, gave the appearance that an afternoon sell-off was developing, but after the head-fake rallied strong through the close and into post-market trading. Despite being an amateur I do quite well when I recognize and trade those sorts of patterns. I often wonder why they repeat as certainly everyone else can see what I'm seeing and would want to take advantage, but I suspect there are structural reasons why the institutions driving the action have to do what they do, and so I just hitch a ride.

My observations were/are on a day-to-day basis, but maybe there are similar occurrences based on day of week, or something else, and with a data set similar to what others on this thread are constructing, one can discover and exploit those patterns, to include having an expectation regarding an opening gap or whether it is going to be a BS or SB day.
 
I was reading some posts of a user named 'spike500' and I came across these 2 posts : it is a matter of statistical probabilities, and entries based on time, not on price, by selling in the top and buying in the bottom, with thin stop loss and high probability win/loss.
Which made me think to traders like Cheese or Blotto, even though the basis of the model per se might be different.

Maybe it can be of some values for people reading this thread :) .

Here are the links to the 2 posts :

http://www.elitetrader.com/vb/showthread.php?s=&postid=747578#post747419
http://www.elitetrader.com/vb/showthread.php?s=&postid=747578#post747578

Just to add something as I think it is not written in these two posts, it seems scenarios are based on 8 possible trend strenght from -4 to +4, where -4 is strong short and +4 strong long.
 
You must log in or register to reply here.
Back
Top