The ideal timeseries to trade - page 5
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
I actually think it is a very fair question considering in what he is trying to achieve. He's trying to statistically find an edge, there is nothing wrong with that. In fact, it is very good to deal with numbers and statistics because that is how patterns are mined with a computer. I'm just not smart enough to do it, but my hat is off to the OP to take this on. There are different edges, a lot of quant funds will probably hire him to do just what he is trying to do and get paid very well.
=================================================
well, yeah, maybe !
but i realize its of no interest to me, sol i go back to the threads that are interested in making money the old fashioned way !
STEALING IT !!!
mp
As of now, nobody has posted anything that is even remotely close to answerring my question: what is your ideal timeseries to trade from a statistical point of view. Any takers?
Don't know if this may be any help but, I can tell you that personally am not trading any timeframes at all.
Instead, I built a new "period" on price only
If the market moves up X pips (plus spread) from the open then a new period is created.
If the market moves down X pips (plus spread) from the open then a new period is created.
This provides a price line which has a "constant" gradient and all I need do is "predict" the number of data periods before the next turn.
This approach is also adaptive - when the market move is fast, many new data periods is created in a short period of time, the opposite in a slow market.
While nothing is perfect, for me this provides a price curve which is much easier to work with.
Just my ideas.
Correlation etc.
Personally, I have been trying for some long time to ensure a correlation between price and indicators but, without much success.
The reason is that different indicators respond differently to price - as such correlation goes in and out of phase.
I have been trying to trade the "best" indicator (with the best correlation) but because the system constantly swapped between the different indicators, the whiplash makes it not really worthwhile.
For me, this held true - except in the longest of cycles.
As such - what is needed is a single, highly adaptive indicator capable of "following" different market conditions. (Staying within a high correlation figure.)
I do use a correlation test between my datasets (for a simple NN) and price - but this is merely some sort of watchdog/cutout in the case my building processes (to create the datasets) loose the plot completely.
As for Hurst - I have tried for so long to make something worthwhile out of it but, till now, without any success.
Any help would be appreciated.
Time compression
Don't know if this may be any help but, I can tell you that personally am not trading any timeframes at all.
Instead, I built a new "period" on price only
If the market moves up X pips (plus spread) from the open then a new period is created.
If the market moves down X pips (plus spread) from the open then a new period is created.
This provides a price line which has a "constant" gradient and all I need do is "predict" the number of data periods before the next turn.
This approach is also adaptive - when the market move is fast, many new data periods is created in a short period of time, the opposite in a slow market.
While nothing is perfect, for me this provides a price curve which is much easier to work with.
Just my ideas.Interesting,interesting..
This concept has been called time compression by an exceptional trader,trader x,that also happens to have written a more serious book under his own ,and well known among Chicago traders,name.
If you want to explore it further,I suggest that you read "Dancing with Lions",grab the course "Eating the loser"..and when finished,enjoy "trading rules that work-28...".
Your approach being similar to point and figure charting,with a twist,you could be interested to explore that area too.
I have found that when somebody has already walked my path,learning from his mistakes and successes can speed the process of reaching destination.
BTW,the more experience I have ,the more I believe that any market is just suport and resistance with volume and open interest..plus some momentum analysis AT THE CRITICAL POINTS..at the purple lines to paraphrase mp.
To trade,you just get enough experience until it feels natural,then you just wait for the right setup and execute it.
Of course,there are many ways to look at S/R and V/OI...you can just draw lines or decide to model the market with nonlinear equations in 3 dimensions(just check the Aleksey thread at forex factory)..just choose the one that fits better with both your personality and your capabilities
Hope this helped
Simba
theta time
Hi, I think this is really interesting concept and it reminds me of theta time, the concept introduced in Dacorogna et al. (1993b): in this paper timescale is expressed as a function of volatility, and also corrected for seasonal patterns (no, not winter and summer
). This business time concept could also be called "Dynamical Deseasonalization" (how's that for a titel of your next post heh).
Something else: I strongly believe in the fact that nonlinearity increases when the timeframe goes down, and that the really small timeframes show aspects of chaos (which in theory would make it more predictable to a certain number of periods away(lyapunov time) than the longer timeframes..).
(and this was confirmed by "Unveiling Non Linearities Through Time Scale.. - Guillaume, Pictet.. (1995)").
I think this is also a valid reason for the lack of performance of most EA's on short timeframes: they are usually based on -lagging- linear indicators, which don't "know" how to transfer nonlinear price action into an accurate signal.
for the definition of theta time: see image below from this paper which contains a follow-up of Dacorogna et al.
The a(t) stands for activity rate (volatility: this should be revised standard deviation: it should be transformed to take the excess kurtosis into account: high peaks and heavy tails in the probability distribution.
Imagine we could use the period converter (look it up) and use the 1 minute chart -ticks would be better but you know..- to make a volatility timeframe chart: now that would be something new!
In theory it would also improve all your linear indicators!
Now my question to all of you would be:
1) Do you have a corrected version of standard deviation (stdev+excess kurtosis)
2)Can you make the periodconverter convert its timescale according to the revised stdev from 1)?
Good Day
i don't believe in complicated stuff. H4 or D1 gives me the green light (SR, direction and strength of the trend) and M5 gives me the entry point (m1 is instable); i always wait for the green light from H4/D1 because it gives me also a "security stop loss" (far from the entry point) but i exit on a signal never higher than H1. After trading for two years (still a newb) i'am not able today to trde on one tf only.
Hurst / factal dimension seems to be useful. I played around a bit with it and coded an indicator, and from my first impression it gives promissing results. The best is, that the signals come BEFORE something happens on the money market. I will attach some images to show what I mean. It doesn't tell you in which direction it will go but it will (unfortunatelly not always) alert you before a significant change will happen.
As a rule of thumb: don't go in a trade when the Fractal dimension is constant and around 1.5, look when the FD is at a bottom or at a top or when it suddenly changes, then probably also the market direction will change. As FD is constrained in the range 1..2 and it can never be a long time at 1.0 or 2.0, you could already make a trading decision when the FD reaches a very high or low value and dont have to wait until it turns around. FD is sensitive to the time frame and period, so check out what works best for you. The FD could be used together with other indicators to get the signals earlier, and for confirmation of trading decisions and/or lotsize adjustment. Good luck!
Perhaps I still don't really understand the problem with regard to time series so, let me try explain what I think - and perhaps we will see some more (interesting) comments.
First I'd like to say that my broker is perfectly happy if I say price is going up (buy) or, price is going down (sell)
In the above statement there is no mention of time so, why then should I include time in my analysis of the markets?
I can understand that, in ancient times (before computers) traders had to make do with daily price etc. - surely, we moved on from that?
Why do we take a one dimensional problem (price going up/down) and insist that it should be 2 dimensional? (price and time)
Personally, I do not care about time, neither about trends for that matter because, ideally, I only place trades on turning points - whenever, wherever they may occur.
As such I fully agree with Simba - perhaps the market should be seen as a map of (possible) turning points only. In the purest sense, this map would become a vertical line, with certain (support/resistance) points marked as possible turning points.
After this then perhaps (maybe) things like momentum etc. may help us go from a "possible" turning point to "real" turning point.
But - most importantly - before we can make sense of any market move, indicator or what else we need to draw the above map/line with the possible turning points clearly defined. (Support/restance map)
In all the above - I fail to see why/where the time aspect is important at all?
post your indicator pls
Hi,
could someone post/quickly create an indicator that draws a bar every ten pips up(green) or down (red)? Instead of the regular timeframes?
That should get us started. I've only recently begun to think in this direction so I'll need some time to test the above to make some usefull comments