Home › Forums › Trading Systems Discussion › True Bars/Candles and Market Sentiment
This topic contains 257 replies, has 18 voices, and was last updated by Anti 1 year, 5 months ago.

AuthorPosts

Seems that the quote function doesn’t work again …
Take you time to do whatever you want ;) I know that this forum isn’t as attracting as it was two years ago since most users are very silent.
Regarding 2*(HighLow)/abs(OpenClose): I’ve tested different versions for the prize path. I think I’ve tested a similar formula, too. However, 2*(HighLow)abs(OpenClose) explained the most amount of variance under these formulas.
I think we’ve discussed my understanding of the prize path earlier. My understanding was that the minimum prize travelling distance is twice the wicks and the openclose range only once. I think @gg53 disputed that way and suggested the formula highlow. Without looking back to my R script I remember that highlow only explained 60 % in the total variance but 2*(highlow)/abs(openclose) explained a 64 %. Nevertheless, both linear regressions (I wonder if a linear regression is really the best function) where highly significant (p<2.2e16). Thus, I’d prefer that way. But as soon as I’ve finished my day job I’ll test 2*(highlow)/abs(openclose).
Hmm, rethinking this, my latest idea will lead to a certain number of ‘divide by zero’ errors. So a different method of normalization would be nice.
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top and bottomfishing to people on an ego trip. (Dr. Alexander Elder)
Yes, it does. Even if I ignore these cases, your idea only explains less than 3 % of total variance …
@simplex: This evening I had some time and thus I followed your suggestion and calculated prize paths and regressions between prize path and volume for EURUD. I obtained the results listed below:
I calculated the results beginning with the smallest TF. Thus, I was amazed to see that there’s a tendency up to the H1 that for higher TFs the prize path explains more variance in volume. I expected the opposite interrelation. However, this relationship collapses when it comes to the H4 and D1.
Another finding is that there’s something strange with the M15. If you look at the slope and intercept you’d usually expect that it increases from M1 to D1. And that’s indeed the case. However, M15 deviates from it – at least my broker’s data.
Here are the graphs (right click on it and choose “show image” to see the higher resulted picture):
In the H4 you can see that the point cloud widens and there may be a bi or trifurcation. The same bifucation can be seen in the D1 much better. What could cause that bifurcation? The D1 data dates back to August 2010, H4 to December 2009. Was there any big intervention?
 This reply was modified 1 year, 5 months ago by Anti.
And here is the prize pathvolumestructure of EURUSD in terms of the intercepts and slopes …
From the slope relation the outstanding M15 behaviour becomes even more visible.
The question is, what we now could do with these numbers …
 This reply was modified 1 year, 5 months ago by Anti.
Oh – I found indeed a shift in behaviour – even in my H1. The indicator window shows the H1 excess activity for a typical week in 2014 (left) and in 2017 (right) … The shift seems to happened somewhere in the beginning of Octobre 2014.
Interesting! Thanks for your work and posting it!
In the H4 you can see that the point cloud widens and there may be a bi or trifurcation. The same bifucation can be seen in the D1 much better. What could cause that bifurcation?
In H4 and D1 you appear to analyze a time period of about 7 – 8 years. In order to find out more about that bi or trifurcation, I think it would be interesting to split the data in (possibly overlapping) chunks, perform the same analysis on each of them and see whether there’s a trend, or maybe a sudden change in the appearance of your data clouds along the timeline. Analyzing the variation of slope, intercept, and R2 along those chunks may add some insight.
One idea re. the dramatic decrease of R2 in H4 and D1 as compared to shorter TFs: those longer TFs consist of candles which reflect trading activity throughout more than only one session, while H1 and shorter candles reflect data from only one specific session each. This fact may be important for all volume specific statistics.
Writing that, one more thought comes to my mind: not only dividing the 8 year timeline in several chunks could be interesting, also looking for daily cycles by analyzing (H1 and shorter) hourly chunks of data separately. My guess would be, that in such an analysis ‘Atlantic’ pairs like EURUSD and GBPUSD will whow different behaviour than ‘Pacific’ pairs like CADJPY or AUDJPY.
Enough ideas for homework over the weekend???
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top and bottomfishing to people on an ego trip. (Dr. Alexander Elder)
… and one more idea: the D1 trifurcation could possibly be caused by weekly cycles. So analyzing five weekdays separately may bring insight.
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top and bottomfishing to people on an ego trip. (Dr. Alexander Elder)

AuthorPosts
You must be logged in to reply to this topic.