Price and Time data set

Just to add ,

On another trading board there is someone who seems to trade with same principles than Cheese, but he is trading EURUSD, so it seems not to be a problem.
See an extract of one post :

"Today we said the main trade was the upswing ,and we were looking for 75 points.

We had an opposite swing of smaller size first. At the "opposite swing end", the price and time was:

1.4367 @ 10.43
and we had a secondary low at 1.4367 at time 13.34

We are looking for 75 points. 75 points added to the end of the opposite swing gives 1.4442. At 15.34 we had a price of 1.4442, which was the exit for the upswing. An upswing of 75 points from either of the two lows to the 15.34 high.

Earlier today, there was a signal for an upswing, closed at breakeven as the opposite swing took over. Trading the opposite (smaller) swing down to 1.4367 yielded 33 points. Trading the opposite swings is not necessary to profit, but an expert trader will trade all swings."

I am not sure I can put a link, but you may ask me by PM if needed, or google

030985

 

Faire,
Thanks for the comments. I used to be a regular poster in a few other forums, but took a few years off just to work on research etc.... I might be posting a bit more often, but I dunno as I don't think my best ideas really every came from posting or reading forums?

030985,
With your thoughts......

First, about the session's opening, initially I would have said that it wouldn't matter, as long as it is consistent. But now, I somewhat think it matters, as people work/trade in cycles. But I dunno for sure?

I think the idea, at least as a good starting point, is to pick something that is universally, or at least largely considered a starting point for most people. I suspect that for maybe some US futures, maybe using 00EST would be ok, as it is possible that a lot of people trading such instruments would respect the instruments time zone of origin?

But with something like currencies.... everybody is trading that, and I would guess that people look to 00GMT as a common starting point for it, hence why I choose to start my sessions when I do.

Onto the coding problem with the sessions and start times...... I had that issue too. As even with my platform and the ability to create whatever session breaks desired, it still has the dates run across different days, which is annoying. What you have to do is sort of create your own 'synthetic' timing counter. I.e., for every minute that passes, you literally just assign it a clock time. So, everything I do is essentially time-stamped via minutes beyond the open. Something in my script might look like:
upClkTime = (int)(Time[CurrentBar-highBar].Subtract(Time[CurrentBar-openBar]).TotalMinutes);

Where 'upClkTime' is just a variable that will hold the integer of the number of minutes for whatever I desire to timestamp. The dates are a bit trickier and require a little bit of messing around with, but usually thats just a matter of checking for a condition, and adding 1 day to it. If none of that makes sense, the point is that its not too hard to extract time however you need it.

As to the standard zigZags available, yes, all the ones I've ever tried are terrible! At least for extracting precise, repeatable swing data. I can't tell you how many hours and how many dozens of iterations of ZZs I've created over the past few years. Usually, the off the shelf ones don't let you start it when you want, and they don't really have any intelligence built into them to track how price actually moved when large outside bars occur. It's unlikely that you'll have access to precise tic data for 10 years, but minute data is no problem. And if you're tracking swings of larger than 10 tics, this isn't a problem anyway. My theory is that if you can build a zz that works over 95% of the time on a 15 minute chart, it will be just great on a 1 minute chart, as the average range of the bars is smaller, and unlikely to penetrate both sides of your bounds. Even if they do, you can get creative with bar directions to nail down those circumstances. I've actually run statistics by hand on my indicators over a decade to be sure they work properly. And the number was over 99.9% on minute data, so I feel pretty comfortable saying that its definitely possible to build one that works well.

As to what I'm doing with the comparison of past days... I think you are asking about sort of a macro view of data. And the answer is Yes in that I have provisions to collect it for say the week or the month. My ZZs will plot across days, weeks or months if desired.... I.e. if I want to not restart it at each day but a week, its no problem. But I figure that's really not the key. Some of those numbers are visible on the charts ive posted. But.... for the moment I am only focusing on intraday, so any comparisons I make aren't really a macro view, but soley based on the day. I am sure this is a huge problem with my analysis, but I guess I'm working on this 1 step at a time so to speak.

About organization/sorting, you are not alone, I don't know what the hell to do with most of my data either?!?!?!

 

Yes, I would say that most of my time here is probably wasted, but the fact is that I would probably be even more clueless without it, since there is a lot of great content in the archives here. Still, the solitary nature of trading and not knowing anyone in the real world who understands the markets, pulls me back here.

Regarding prediction, would you agree that it most likely not possible to predict the trading day in detail in advance, but to have an expectation of the day's main template that is updated (with live data?) as the day folds out?

My impression from reading Cheese's posts, is that his method initially was based on a macro approach, i.e., catching the major move of the day, but later on he evolved to micro trading the day, exploiting the smaller gyrations as well.

Initially, it seems like he was opposed to charts, but later on he talked warmly about charts and how learning to read them were crucial.

What I get from that is this: predict the day's 'major' moves using market statistics/metrics (the predictive model) and micro-trade the day using price action and proper chart settings.

I still have not been able to predict the next day's template, but I look forward to resume my work and start sorting again and asking questions.

 

Interesting queries. No doubt anyone following along has wondered these things as well.

I dunno for sure of course, but what I expect is that one can at least predict/guess/hypothosize, whatever semantic verb desired..... with surprising accuracy, both the timing and approximate length of the next day in advance, with regards to the main move and its direction. I think once you have that, the approximate expected micro moves within that would be more apparent or obvious.

Right now, admittedly that hasn't been my focus for quite some time. It probably should be something that I think about more? Mainly, I am trading with my own statistics I create from my data. But I know this is a tiny piece of the pie.

With the parameters one needs to sort out that portion of the day in advance, I am unsure of? I'm not sure if it must come from several days prior, if it is just an aggregation and statistical breakdown of similar events/occurrences in the past or if it is a matter of creating a sort of premarket delimiter OF THE CURRENT DAY that will tell you what the rest of the day will look like?

I have read references made to "@ the open or before I know in advance," so that's another consideration to think about.

At any rate, I'm pretty sure what is ACTUALLY possible is a level of trading that most people just wouldn't accept or believe possible. So just my thoughts are that there is really no reason to think it is unattainable or out of reach just b/c most people would prefer to 'grind it out,' eeking by and never really really killing it.

 

Yes.

I collect a lot of metrics, possibly more than needed, but I'm not sure which one to include when sorting the days and attempting to predict.

Does one include only the prior day? Or more days as well? More days limit the number of occurrences, but may not make for more accurate prediction.

Regardless, from what I gather, you may have an expectation for the next day by the close, but the open needs to confirm. It may go like this:

"Based on the predictive model, tomorrow will be either template A,B,C.

If the market opens gap down, it is template A. Wait for the price action and condition X to occur for confirmation or possibly the first 30 minutes of the day to be able to predict with accuracy what kind of day it will be.

If the market opens gap up, template B. Flat open; template C.

When you know; trade the day."

Anyway, I think I've said enough already. Would be fun to bounce some ideas on PM later on if you're interested.

 

Interesting thread. Thanks!

IMHO, the approaches described above seek to record empirical observations (about characteristics of price, or the market) which are then used as a basis (“signals”) for forecasting future price/market behavior. i.e., systematic trading.

A key element (again, IMHO) that might be missing is to relate these empirical observations to a “model” of how the market works (and I’m not pretending that’s easy!). The model can help to emphasize the importance of some observations, de-emphasize others, and keep matters simple (i.e. KISS … to avoid curve-fitting).

 

Trying out some different sorting of various metrics of my tables. This isn't done on much data as I'm still working some things out before running on larger sets of data, then going back to visually eyeball similar conditions on the chart..... but for example wondering if for example this sort in particular presents similar sets of data?

Some similarities for sure, but still not there......

 

Another example:

 

I don't know if it is on purpose, but the pictures are small size, so that we cannot see clearly. Are the similarities we are meant to see are in the way the main moves developped ? I can see it on first picture, but I am not sure for the second

 

Yeah, I was just thinking about the 'general' look of how it unfolded, or comparisons made on sets of data that have similar historical conditions in force PRIOR to their occurrence.

The parameters were based mostly on the data PRIOR to the first day shown in the picture. The 2nd, 3rd, and so on probably become a bit less coordinated. I'm sure this makes a bit of sense.

I still have a lot of work to do, just some early ideas.

I think this still isn't 'the proper path,' as this has nothing to do with 'prediction,' or much to do with the numbers in the swings themselves. Also, I still will be looking to slice the data up in different sizes once I get my ideas in place. So maybe some more discoveries can be found when that takes place.

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?!?!?!