Quantitative Technical Analysis | Page 2 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Fundamental Analysis, Technical Analysis Or Graph Analysis? 28 replies

Quantitative analysis 8 replies

Is there anyone using behaviour+quantitati ve Analysis ? 0 replies

The Shark's Trading Book - Quantitative Trading 2 replies

Quantitative Qualitative Estimation for MT4? 0 replies

Attachments: Quantitative Technical Analysis
Exit Attachments

Quantitative Technical Analysis

  • #21
  • Jun 23, 2014 8:05am Jun 23, 2014 8:05am
Quoting PipMeUp
Thanks for posting this paper. This is an example of what I consider a questionnable definition. The authors don't explain the reason why they used a 120 day window. They don't explain why a 10x10 grid is used instead of 9x13 for example. They don't explain the reason for the weights in the grid cells. They don't explain why the pattern they try to match is a broken bullish flag as opposed to a forming flag. They don't explain why they choose to match a bullish flag in a bullish market. I would have found it way more interesting to match bearish...
I dont think you will find any paper that will describe everything precisely, if they did that would be the holy grail for you i guess.It's like the trading books, they teach you how to trade but they dont teach you how to make money, and trading most of the time is not equivalent to making money.There is the catch, so there are always missing blanks which you need to fill.Hundred years ago there wasnt any trading book, yet there were many profitable traders then too.

Its all about the opportunists, who can always find a way to make money, not about the herd who always needs justification and detailed explanation about the money making process.

Nobody will teach you how to make money, you have to figure that out yourself
"There's a sucker born every minute" - P.T. Barnum
  • #22
  • Jun 23, 2014 8:38am Jun 23, 2014 8:38am
Quoting Proximus
{quote} I dont think you will find any paper that will describe everything precisely, if they did that would be the holy grail for you i guess.It's like the trading books, they teach you how to trade but they dont teach you how to make money, and trading most of the time is not equivalent to making money.There is the catch, so there are always missing blanks which you need to fill.Hundred years ago there wasnt any trading book, yet there were many profitable traders then too. Its all about the opportunists, who can always find a way to make money,...
Please just stop!

This thread is NOT about building a holy grail, not about any trading system, not even about opening a single trade.

It is about patterns matching and artifical intelligence used to quantify the construction of the usual TA elements.
The paper posted by olsen-yersen shows an interesting method. Unfortunately it contains shortcomings.

Please keep your advise for you or post them somewhere else. I want to keep this thread clean and I will.
No greed. No fear. Just maths.
  • #23
  • Jun 23, 2014 8:45am Jun 23, 2014 8:45am
I've coded my idea. The black dots are drawn at the price of the extreme and at the time of the detection.
Attached Image (click to enlarge)
Click to Enlarge Name: pivots.png Size: 37 KB

EDIT: Note the last pullback/congestion has not been matched.
No greed. No fear. Just maths.
  • #24
  • Jun 23, 2014 8:53am Jun 23, 2014 8:53am
Quoting PipMeUp
{quote} Please just stop! This thread is NOT about building a holy grail, not about any trading system, not even about opening a single trade. It is about patterns matching and artifical intelligence used to quantify the construction of the usual TA elements. The paper posted by olsen-yersen shows an interesting method. Unfortunately it contains shortcomings. Please keep your advise for you or post them somewhere else. I want to keep this thread clean and I will.
Ok, it was only an advice .

Ok here is a paper which describes acutally Adal's approach, by measuring the ticks you can sort of quantify big price movements and predict them ahead of time, be that trend or breakout doesnt matter.But you need to get some tick data first.

Attached File(s)
File Type: pdf 0809.1040.pdf   812 KB | 774 downloads
"There's a sucker born every minute" - P.T. Barnum
  • #25
  • Jun 23, 2014 8:54am Jun 23, 2014 8:54am
For each country owning his own currency (or group of countries with the same currency) the value only draw the economy health (and/or its supposed evolution).
So is it sensitive to minerals, cereals price exchange ? is it importing/exporting ? (often sensitive to USD delta)
Or is it sensitive to weather troubles ? (cereals, ....)
If each country have some political calendar (with possible economy trouble), all have sensitivity with regional troubles (war, strike, embargo) but with different sensitivity levels.

And of course the speculation on minarals, cereals, .... is also playing around.

Thats said, the fx value is the whole result.
  • #26
  • Jun 23, 2014 9:04am Jun 23, 2014 9:04am
Quoting dinofx
For each country owning his own currency (or group of countries with the same currency) the value only draw the economy health (and/or its supposed evolution). So is it sensitive to minerals, cereals price exchange ? is it importing/exporting ? (often sensitive to USD delta) Or is it sensitive to weather troubles ? (cereals, ....) If each country have some political calendar (with possible economy trouble), all have sensitivity with regional troubles (war, strike, embargo) but with different sensitivity levels. And of course the speculation on minarals,...
Yes, i also think it is good to view each currency on its own not in pairs, so its better to find the USD strength from EUR/USD ,USD/JPY and EUR/JPY.

AND compare the USD with the S&P 500 and the other major indices, they are a good benchmark, and you can spot trends early there.

And then when you found out the USD strength, find another currency , weaker or stronger, for example if the S&P 500 is moving up, then find a weak currency, the EUR is the weakest at the moment.

So because of this trade the EUR/USD, as going short.This is probably the fundamental definition of a downtrend on E/U.

As for how to quantify it simple:

1) Extract each currency's strength from 3 pairs: USD and EUR from EUR/USD,USD/JPY and EUR/JPY
2) Correlate each individual currency with its major stock index: USD with S&P500, NASDAQ, DOW 30, 10 Year T bond and EUR with EURO STOXX 50, and 10 year German bonds , JPY with Nikkei 225 and Japanese 10 year bond,etc
3) Find the weakest currency and the strongest
4) Trade that currency pair in the strong direction

Its simple.
"There's a sucker born every minute" - P.T. Barnum
  • #27
  • Jun 23, 2014 9:27am Jun 23, 2014 9:27am
How are those last two posts related to the thread's topic?
No greed. No fear. Just maths.
  • #28
  • Jun 23, 2014 9:51am Jun 23, 2014 9:51am
Quoting PipMeUp
How are those last two posts related to the thread's topic?
Because in my opinion its easier to spot a forming trend if you look at other currencies too, and their correlation.By looking at a single pair you might miss some important data, if somebody sells big cash on JPY that will affect the EUR/USD, and if you look only at EUR/USD past movement, it will be hard to predict accurately its movement.

Here is my correlation indicator which i`ve used for this research, i will share it with you because i no longer trade trends:

Attached Image (click to enlarge)
Click to Enlarge Name: eurusdm5.png Size: 54 KB


Green = EURUSD
Orange = USDJPY
Red=GBPUSD
Pink =AUDUSD
Blue =USDCHF
Yellow =USDCAD
White =NZDUSD

All movements are in percentages, because these pairs are denominated in different digits so its easier to quantify their movement in %.
The percentage move then its saved in a file, in MQL4/Files, the first row is Bar[0] and the last row is Bar[COUNTED_BARS]
Good Luck analysing the correlations in the movements, i bet you will spot some trends with this analysis

Attached File(s)
File Type: ex4 LINETABLE_CURRENCIES.ex4   12 KB | 416 downloads
"There's a sucker born every minute" - P.T. Barnum
  • #29
  • Jun 23, 2014 10:22am Jun 23, 2014 10:22am
Quoting Proximus
{quote} Because in my opinion its easier to spot a forming trend if you look at other currencies too, and their correlation.By looking at a single pair you might miss some important data, if somebody sells big cash on JPY that will affect the EUR/USD, and if you look only at EUR/USD past movement, it will be hard to predict accurately its movement. Here is my correlation indicator which i`ve used for this research, i will share it with you because i no longer trade trends: {image} Green = EURUSD Orange = USDJPY Red=GBPUSD Pink =AUDUSD Blue =USDCHF...
Yes of course it is a good idea to work at the currency level and pair the two opposite in strength. But you just replaced EUR/USD with some index of EUR.
How to quantitatively estimate the trend of this index? Not to answer "is EUR strong?" but "how strong is EUR?"
=> http://www.forexfactory.com/showthread.php?t=435706

"if somebody sells big cash on JPY that will affect the EUR/USD, and if you look only at EUR/USD past movement, it will be hard to predict accurately its movement."
This is why I don't try to forecast anything. Only measuring. I'm not trying to guess the next trend but to follow the current one.

"All movements are in percentages, because these pairs are denominated in different digits"
No. movements are in percentages because percentages are unitless. 1 pip of USD/CAD is 0.0001 Canadian Dollars and can't be added to pips of EUR/USD which are valued in pips of US Dollar.
No greed. No fear. Just maths.
  • #30
  • Jun 23, 2014 10:31am Jun 23, 2014 10:31am
This is a comparison between the previous pivot detector and another ZigZag-like indicator I wrote (mine is not repainting and uses no parameters!).
Here the dotted black lines mark the price and time of the detection. Both are quite often in agrement. The ZZ version flips earlier but more often.

Is one objectively better than the other?
Attached Image (click to enlarge)
Click to Enlarge Name: ZigzagVsKalman.png Size: 48 KB
No greed. No fear. Just maths.
  • #31
  • Edited 10:55am Jun 23, 2014 10:35am | Edited 10:55am
Quoting PipMeUp
{quote} Yes of course it is a good idea to work at the currency level and pair the two opposite in strength. But you just replaced EUR/USD with some index of EUR. How to quantitatively estimate the trend of this index? Not to answer "is EUR strong?" but "how strong is EUR?" => http://www.forexfactory.com/showthread.php?t=435706
Well this indicator wasnt actually designed for this sort of analysis.But you can try and extract the currency from 3 major pairs EUR/USD, USD/JPY and EUR/JPY.

How to quantify the overbought/oversold conditions, well you need to measure the average upmove % and downmove %, and anything above that will be the 2 extremes, once the price reaches any of them, it will have a higher bias of reversing from there.

You can also measure the distribution of the changes too and see how it distributes over many bars.

Quote
"if somebody sells big cash on JPY that will affect the EUR/USD, and if you look only at EUR/USD past movement, it will be hard to predict accurately its movement." This is why I don't try to forecast anything. Only measuring. I'm not trying to guess the next trend but to follow the current one.

You can also predict them not only measure, as you see from the chart some currencies react later to sharp moves than others, you have to find which pair moves before the others.

I would be a risk free inter-currency arbitrage

Quote
"All movements are in percentages, because these pairs are denominated in different digits" No. movements are in percentages because percentages are unitless. 1 pip of USD/CAD is 0.0001 Canadian Dollars and can't be added to pips of EUR/USD which are valued in pips of US Dollar.

Yes.
"There's a sucker born every minute" - P.T. Barnum
  • #32
  • Jun 24, 2014 10:46am Jun 24, 2014 10:46am
Quoting PipMeUp
After one has defined that the trend is a smooth component around which the price oscillates one has to build a mean to find this component, in real time, from the price.
I never found the "smooth" trend definition to be useful. I don't think it is a particularly good match for real price moves. For me, smooth = delayed = useless.

Quoting PipMeUp
I'd like to automatically build the trend channels, that is to closely bracket the PA and see if the estimation can be improved. So I need to build a mean to find this other component of the TA.
I now do it the other way. I remove the oscillations to find the trend.
  • #33
  • Jun 24, 2014 12:27pm Jun 24, 2014 12:27pm
Quote
I never found the "smooth" trend definition to be useful. I don't think it is a particularly good match for real price moves. For
me, smooth = delayed = useless.
The smooth component is the trend, per definition. I mean my definition. It does not lag. You just cannot estimate it without intruducing lag. This is why I use a Kalman filter to get a best estimate of this hidden component. This estimate does lag and is less smooth. Which is the reason why I'd like to add some phenomenological elements like trend channel to improve this estimation.

What other definition would you substitute for the trend?

Also I disagree that delayed means useless. Everything is delayed anyway even a 1 min candle. I did show in another thread that the trend is stable. Nothing requires to enter the trend at its very beginning.

Quote
I now do it the other way. I remove the oscillations to find the trend.
Hmm I'm perplex. Almost everybody does that. It is what any MA does. If you use a high pass filter that you substract from the signal you get the same lag and the same filtering power. Perhaps you could explain better?
No greed. No fear. Just maths.
  • #34
  • Jun 24, 2014 9:06pm Jun 24, 2014 9:06pm
Quoting PipMeUp
{quote} The smooth component is the trend, per definition. I mean my definition. It does not lag. You just cannot estimate it without intruducing lag. This is why I use a Kalman filter to get a best estimate of this hidden component. This estimate does lag and is less smooth. Which is the reason why I'd like to add some phenomenological elements like trend channel to improve this estimation. What other definition would you substitute for the trend? Also I disagree that delayed means useless. Everything is delayed anyway even a 1 min candle. I did...
You're right. I misspoke. Measurement, detection, adaptation, it all adds delay however you do it. My model isn't necessarily better, just different.
I bracket the price oscillations with a channel. Any price movement not bound by the channel I count as the trend/range signal. Successive steps in the same direction indicate a trend. Alternating steps (or no steps) indicate a range.
Attached Image (click to enlarge)
Click to Enlarge Name: channel3.PNG Size: 36 KB

Sometimes the trend moves or turns quickly which is why I prefer not to use any low-pass filters.
Attached Image (click to enlarge)
Click to Enlarge Name: Channel.PNG Size: 24 KB
  • #35
  • Jun 25, 2014 2:45am Jun 25, 2014 2:45am
Very interesting indicator. It looks like a median filter but without the lag. From the screenshot I think I understand how the white line is updated. But how do you get the width of the channel? Is it based on an estimation of the variance around the white line?

Did you try to draw it over a constant range bar chart?
No greed. No fear. Just maths.
  • #36
  • Jun 27, 2014 6:52am Jun 27, 2014 6:52am
I want to estiamte 4 market state with zig zag indicator and machine learning :
1-Stable and quite 2-Trending and quiet 3-Stable and volatile 4-Trending and volatile

http://i.hizliresim.com/oQB8AX.jpg

I am using a zig zag indicator..The zig zag indicators' each leg giving the number of pips number of bars and the ratio between the length of the leg and the previous length of the leg.Which machine algo and/or stats would be suitable for the estimation of 4 market state with zig zag indicator ?How many legs would be optimal to estimate the market state?Any feedbacks appreciated.




http://i.hizliresim.com/0VOLMW.png
  • #37
  • Jul 1, 2014 7:21am Jul 1, 2014 7:21am
I've built a zig zag indicator which generates statistically meaningful swings. Grouping these zigs and zags is the next step to build the usual TA patterns. They can also be used to build not so usual patterns like the beastiary of the harmonics or olsen-yersen's market states. But how to do that? Sorry olsen-yersen I don't know much about neural networks.

A swing can be of same size, bigger or smaller than the previous one. Swings always alternate in direction. If we want to take the fibo ratios into account we can name them using their ratios. So we have:
{ swing-very-small, swing-23, swing-38, swing-50, swing-61, swing-76, swing-100, swing-127, swing-138, swing-161, swing-very-big }

Any swing can be labelled and we can build a context free language from this alphabet. The word "X swing-61 swing-127", where X is any swing, describes an ABCD pattern. The pattern is bearish or bullish depending only on X being a up swing or a down swing. My problem is that this language is ambiguous and can't be parsed trivially:
Attached Image (click to enlarge)
Click to Enlarge Name: variousCases.png Size: 29 KB
No greed. No fear. Just maths.
  • #38
  • Jul 4, 2014 6:24am Jul 4, 2014 6:24am
I try to automatically build a trend channel around the recent PA. For now I create two channels. One made from the highs and the second made from the lows. There is certainly room for improvement...

The zip file contains a video. Sorry for the quality I had to keep the size of the file under control.
The trend I consider is the curve made of the little blue cirlcles. To big black dots are on the middle of each channels.
Attached File(s)
File Type: zip trend_channels.zip   7.2 MB | 754 downloads
No greed. No fear. Just maths.
  • #39
  • Jul 4, 2014 10:59am Jul 4, 2014 10:59am
Some stats for these channels:

Top row = the average price (blue circles) vs the big black dots.
Bottom row = the slope of the trend vs the slope of the channels.
Quite good and... normally distributed!
Attached Image (click to enlarge)
Click to Enlarge Name: channelsStats.png Size: 15 KB
No greed. No fear. Just maths.
  • #40
  • Jul 19, 2015 6:03am Jul 19, 2015 6:03am
recently, i attended a talk on Quant Analysis. May i ask if it is good? i heard most hedge funds are using it. please advise, thanks!
0 traders viewing now