Price and Time data set

[030985]

Quote from forrestang:
Unfortunately I am still working on the path of just using my tables as a historical database of sorts, and finding patterns of higher than less-likely outcomes, which I doubt is 'the way.' But we'll see?!?!?! [/B]

I am more and more thinking the same, but who knows ? ^^


Frequently, I saw repeated that this was a zero sum game; Fudgestain said the following sentences, which are similar to what we can hear here and there :

"But the market (Dow) is a zero sum game with gyrations around a theoretical zero. This means there is not and can never be a very great difference between the sum of the upmoves and the sum of the down moves."

"This is a zero sum game; it ain't going anywhere that it won't come back from"

I tried to stop and think what could be the implications of such a thing, and some ideas came to mind, though, maybe it is nonsense, and I am actually not able to verify them for now.

It could mean that for ANY up/down move, there is a corresponding opposite down/up move.

Also, I read about Price being maybe not a 'fixed' thing but more like a multi dimensional living, and so that sometimes a certain timeframe is in command, sometimes another, etc... which could for example mean that the corresponding moves, if they exist, can sometimes be a dozen of days before, or sometimes few minutes before.

So what i thought is that maybe, a corresponding move, could be the last opposite move of the same size ( but can be contained in a move of bigger size ) with no retracement during the move longer than the maximal retracement observed during the move of reference. Sorry for my English, maybe it does not sound clear, so I will give an example :

Let's take, for no particular reason, the Main Move of 7th October, looking at the EURUSD.
Considering the start and close of the day I use, this Main Move is for me a Down move, size is 48 pips and duration is 336 minutes. The maximal up-retracement observed during this move is 13 pips.
So, according to this 'corresponding' opposite move theory ( which I repeat may be just nonsense ^^ ) we may look for the last Up move of at least 48 pips with a maximal down-retracement of 13 pips, which leads to looking for the last Up move of at least 48 pips without 14 pips down-retracement inside it.
The last Up move respecting these conditons occured on 2nd October, with a move of 48 pips, with no down-retracements of more than 13 pips, in 2 minutes.
So, now we have an Up move of 2 minute duration, a Down move of 336 minutes duration, and in-between the end of the Up move and the start of the Down move ~4085 minutes ( measured with ruler on 1 minute bars, during usual Market Open, weekend data hidden ).
Can something be discovered thanks to these time durations and price ranges ? I don't know.

And, well, I am sorry but that's it, I just wanted to share what I came up with in case if it gives ideas, but I cannot conclude anything from that.

But I think that maybe taking note of the maximal retracement observed when observing a particular move can be of interest.

Also, I thought about this, but I won't be able to verify it neither because lack of programming skill, and the inability to tabulate so many datas :
If we consider again the Main Move of the 7th october, and we would like to anticipate what comes next, maybe we could observe what came after :
- every down swings of a size of 48 pips
- every down swings of 336 minutes
- every down swings with no up-retracement of more than 13 pips
And then try to see some correlation in these datas, by averaging or something else.

Just thinking out loud.

Though it is maybe not the good path to follow because we actually may not be able know for 100% when a move is finished or not in real time ?

A lot to say not much

030985

 

[030985]

Thinking through this more, I noticed something :

Cheese often speaks of gyrations, not just swings. I think that if he insists on using this term, and defining it as an upmove followed by a downmove, or a downmove followed by an upmove, it means that it might be an important element of the datas required to try to predict what will come next.

And I noticed that the few statistics I have are about the Main Move and swings ( Main Move being just one swing ).

So maybe it would help to derive statistics from gyrations first, instead of swings (moves) only :

- Its direction type (Up followed by Down, or Down followed by Up )
- Its range
- What move has the highest range, down one or up one
- Number of points available by adding the range of both moves
- Its total duration
- ( and maybe duration of each move )

And according to something I read, we may consider that the last gyration ( which means, the last two swings ) would help to predict the next swing.
Though there is no evidence of this, just something to think of.



030985

 

[mushinseeker]

Hi. My Sierrachart writes OHLC data into a csv text file and I import this file into Amibroker using javascript loaded in Windows Task Scheduler to fire every 1 minute . The csv look like...

2014/02/21, 16:27:00, 0.89743, 0.89715, 0.89698, 0.89705, 60, 60, 33, 27
2014/02/21, 16:28:00, 0.89713, 0.69715, 0.89698, 0.89705, 60, 60, 43, 27
2014/02/21, 16:29:00, 0.89713, 0.89715, 0.89698, 0.89705, 60, 60, 33, 27
2014/02/21, 16:30:00, 0.89704, 0.89704, 0.89690, 0.89701, 39, 39, 24, 15

Javascript reads the entire dataset per ticker . All I am interested in is the last 1 minute since the other minutes have already been loaded previously . Can anyone please suggest an automated way to combine old csv and new csv file so that only the last minute is imported to speed up process? Thanks.

 

[forrestang]

With regards to PRICE and TIME.......

Has anyone given any thought to the difference between calendar time and market time?

Calendar Time being just the difference of time between two points, whether the market is open or not.

Market Time being the time between two points while the market is in session.

 

[forrestang]

Another thing to think about...... That is regarding the way you can measure something.

You can measure something relative to something that already happened.... sort of KNOWING what happened. For example, you could measure a swing relative to the Main Move of the day. At the end of the day. So you'd have somewhat of a big picture of how that swing fit in.

You could measure that same thing, but measure it relative to the Main Move, as it CURRENTLY sits now. So that Main Move could currently be of a different magnitude and direction at the time you measure it.

Just another possible query with regards to p/t.

 

[forrestang]

Working with a database of prior prices (about 5 years of recent ES data), there are some moves that seem to happen all over the day, and then some that seem to only happen at certain times (timescale wise). I.e. likely hood of a longer move from the AH session vs. a smaller/similar size move (time wise), during RTH.

For example, looking at a move of 47 tics down(red).... if it has a larger duration, the proceeding legs appear to be a bit more uniform in behavior. Shorter duration moves appear to be a bit more random, and spread out over the trading day (the shown moves are all relative to the OPEN price and time when they occurred). They also appear to have swings that follow them that are a bit more sporadic as well.

Not sure if this is just coincidence due to the smaller number of samples as larger durations are rarer?

Just a note these are also filtered relative to the current and prior day's high/low prices relative to each other(e.g. today current low is < than prior low and prior high is > current high).

 

[030985]

With regards to PRICE and TIME.......

Has anyone given any thought to the difference between calendar time and market time?

Calendar Time being just the difference of time between two points, whether the market is open or not.

Market Time being the time between two points while the market is in session.

Indeed, maybe it makes some senses to look at datas and time only by refering to RTH and ignoring after hours. I tried to see if there was something regarding High/Low order and time between them on 2,3,4 and 5 continuous RTH sessions, and so ignoring the 'after hours time', with some other parameters. But as usual, nothing interesting and always 50/50 probabilities about the next day being a BS or SB.

You could measure that same thing, but measure it relative to the Main Move, as it CURRENTLY sits now. So that Main Move could currently be of a different magnitude and direction at the time you measure it.

I defnitly think this is important in order to be more precise and pinpoint probabilities and entries/exits. But if we try to anticipate the next day at the close of today's RTH, then if I'm not mistaken everything seems 'temporarily' settled.

For example, looking at a move of 47 tics down(red).... if it has a larger duration, the proceeding legs appear to be a bit more uniform in behavior. Shorter duration moves appear to be a bit more random, and spread out over the trading day (the shown moves are all relative to the OPEN price and time when they occurred). They also appear to have swings that follow them that are a bit more sporadic as well.

Not sure if this is just coincidence due to the smaller number of samples as larger durations are rarer?

Just to verify if I understand : could it mean that, maybe, the longer a move last, the more accurate could be the anticipation of what may happen next ?



030985

 

[forrestang]

So... for the last question, what you say may be true, I dunno?

But what I was saying, was that certain moves just don't happen during the middle of the day, or VERY rarely, while some moves happen everywhere.

For example, to just pull an example from the ES. Using the last 5 years of data, a minimum swing size of 16 tics, and all ETH.

If I pick from that database a swing of -47Tics and 868 minutes with a 10% tolerance on both the price and time. This is a relatively large 'stretch' for that span requirement. I'll attach a distribution of the 16 tic spans. You can see where the bulk of the samples occur. So you can see how -47/868 is a bit of the edge of how they occur.

If I look at WHERE all those swings occur, this type of swing doesn't occur in the middle of the day. They're all pretty much at the beginning(red lines).

Not sure how this could lead to prediction, as you just wind up with what happened in the past, and not sure if that's useful. But, it does somewhat suggest 'maybe' that certain things only happen at certain times? I dunno, but was just an observation.

 

[forrestang]

Anyone following.....

Have you guys done any detailed study into:

Days where the Low comes BEFORE the high...

-VS-

Days where the High comes BEFORE the Low....

And found very precise timing with regards to major swing points on the former, and much less so with the latter?

It seems the overall structure, and how the Daily Extremes are made are waaaay different between the two?

 

[030985]

But what I was saying, was that certain moves just don't happen during the middle of the day, or VERY rarely, while some moves happen everywhere.

I see, thank you for your clarification.

Anyone following.....

Have you guys done any detailed study into:

Days where the Low comes BEFORE the high...

-VS-

Days where the High comes BEFORE the Low....

And found very precise timing with regards to major swing points on the former, and much less so with the latter?

It seems the overall structure, and how the Daily Extremes are made are waaaay different between the two?

As far as I am concerned, I cannot help much about this, but thinking of it, it makes sense to me : If we consider Low before high as "up", and High before Low as "down", it could be as any two opposite principles like yin and yang for example, and one could be more 'ordered' than the other one, if we may say so (the order exists for the "down" too but is less apparent to us ? ). A bit like, rational and unrationnal...
What i just said is of no help, but just to say that your question makes sense to me.


Lately I thought about this Price and Time things, and I came to conclusion that maybe the accent was not put enough on time. It was often more about Price, as far as I am concerned, as I actually am not sure how to study the Time part. So I wondered :

- What if we consider an up or down trend regarding only time, could it help to see things differently ?
What I mean here is basically : If we consider RTH of the Dax for example which is 510 minutes, if the main move lasts more than 255 minutes, the main move direction is the trend for the RTH day. If it lasts less than 255 minutes, the trend is opposite to the main move direction.
( what made me think of this, is trying to put accent on time, but also because often some traders who seem to be a bit more enlightened spoke about things being not as what they look, and 'optical illusion' or things like this...And truth is, for example, when a spike occurs, we are somehow attracted by the spike, but they are often happening on short time periods...)
But from what I have observed I did not see anything special for now...

- Another thing is, if we are observing swings from a minimum particular size :
If we observe 3 consecutive swings of minimum 30 points ( so either UP/DOWN/UP or DOWN/UP/DOWN ). It seems that maybe the duration of the middle swing compared to the added durations of the previous and next swing, may have some implications. As in : Second swing duration > first swing duration + third swing duration or Second swing duration < first swing duration + third swing duration.
I am sending you a PM for some more details.

030985