Autochartist: Pattern Recognition Scanner
I know this instrument for a while to tell you the truth I have never considered it seriously. The software is called autochartist, and I think authochartis and PRS (Pattern Recognition Scanner are the same thing).
So what the Autochartis is stanging for in Forex?
It scans the entire Forex market, for chart patterns
After that the patterns are grouped by two basic categories for decision making :
-Completed patterns (the chart pattern has been completed and a break - out has occured)
-Emerging patterns (the chart pattern is currently emerging, no break - out yet)
Also each pattern is evaluated:
by quality, initial trend, uniformity, clarity: This matters because for example a reversal pattern like head and shoulders cannot exist without a previous trend.
A forecast is given for the price levels to be reached after a break - out. The price level achievement with technical analysis patterns is a difficult issue. One idea is that if the price target is not achieved quickly to lower the target or even to close the orders while it is positive.
 |
| От Поле за пускане |
I think that it is an interesting and valuable tool for a beginner wishing to learn the technical analysis patterns and to have a second opinion. For example you maybe know the excellent Encyclopedia of chart patterns of Bulkovski, this is the site of the author. However that was for the Stock market, the most interesting part of the book for me is the probability of success of each pattern. But there were two problems.
1. Those results were for the Stock market and I am in the Forex market.
2. Those results were for patterns identified by Bulkowski I would subjectively see maybe completely different patterns
Those problems do not exist with Autochartist.
1. The results are for Forex
2. The pattern reading is done by a machine so I cannot add my own subjectivity in detecting the pattern
3. What I can do is to filter the market conditions when a profitable pattern is more likely to emerge
4. To filter by my own procedures to take the pattern or not (every trader may have his own procedures but it is possible to miss that step too)
However today I reconsider my opinion. Really. Why?
The statistics is for a period 2011-07-04 to 2011-12-28
899 patterns were evaluated and 69 % from them were correct.
Then you can find a table of the results by pattern, by symbol, by time frame and by time of the day.
I can say to you that this is a lot of work. Really there is a serious work done here.
However I would like to add one precision. Those results are not for the patterns in general. Those results are specifically for the patterns identified by the PRS. However and we will never know.
Those results do they come from optimization of the PRS on past history or they are Out of sample results.
I think we would never know. As you can guess to derive patterns you need to have some hidden parameters there and those parameters can be changed and once changed they would give completely different results. As it is proprietary we would never know.
It is also possible that a team is dedicated on this job which is doing constant adjustments of the Pattern recognition scanner Autochartist.
If we ignore that information the results are pretty interesting indeed.
You need an opened account in order to have the full funtionality or you can use the limited functionality with a demo account. However if you install the Metatrader pluggin you can overcome this. In fact in the limited functionalyty they give you a plain picture on the more exotic paris and neveron EUR/USD, however if you install the metatrader indicator you will have it in Metatrader and you will see the chart in your Metatrader and you can use the separate software for scanning arround the market.
Comments
BFX made some major changes in their platform, so I had to switch my demo platform there. In the new platform there is a direct access to the Pattern Recognition Scanner. You can see how it looks. Anyway now this pattern may be correct but accorrding to the statistics this time of the day 12:00 GMT is not the best time to play a break-out pattern.
Today I did some demo trades using the Pattern Recognition Scanner to do some live testing. The basic idea is to see opportunities in forex pairs you usually do not trade. You know how boring may be to wait for the pattern on EURUSD that never comes. Here the scanner scans the entire Forex market for break - out patterns. The statistics show that there is not a significant change in profitability between different Forex pairs. That is why the use of PRS makes sense. You know as human you can make your own technical analysis, but you cannot scan all the time frames and all the pairs in the same time.
Here I just took the trades as they come without filtering them. The software gives a target level but does not give you a stop loss level, so you have to decide based on your experience and knowledge.
As this is demo account the pattern is delayed by 300 m, but I was looking at the scanner directly which was giving the trades in real time.
This is the last trade of the day. I was somewhat sceptic about this trade and I got out too early.
The interesting point about the pattern recognition scanner is that it is very, very good about timing detection. I am not sure about market direction, but that is perfectly OK, because it is about to detect the limit cases, and be right within the decision zone.
In this example it detected such a spot. However I do not believe it and I would take a long trade with the trend.
Her it is the corresponding trade. It is based on discretionary price action analysis.
The is an Channel Up.
We are at the bottom of the channel.
And this is a positions for an entry.
It can be classified using the Ross methodology as a Ross Hook. A Ross Hook is observed when in a trend there is a correction.
Here one pattern from today 25.01.2012 from the EURUSD. It looks nice.
However even if works nice at detection of opportunities this does not mean that the target would be reached. Here really nice timing of this chart pattern, really nice. The market goes down a little bit then it reverses.
I would like to add the Oanda open orders that give you even more precise data where the targets can go and how to define the risk levels based on the market data and the current situation.
Over all the approach would be.
1. Detecting an opportunity that exists NOW. There are many opportunities on daily basis with PRS. Here the PRS will help to get the appropriate timing.
2. The Oanda open orders give an estimate on the current market situation and would give support and resistance areas extracted from the deep market. The price is not the only thing, the open orders even not reflected by the price yet are a part of the market.
3. Additional studies: Here we can add more things: entropy, fractal dimension, optimized directionnal systems, neural nets, correlations whatever you like etc.
OK I would add a shot of the oanda open orders. Here We can see that it Euro overshoots a little bit the resitance.
The idea here is that the basic principles are shared. The question is about the practical implementation.
I ask the question to myself what is the best can do a part time non professional retail trader (the professionals know very well what they can do otherwise they won't be professionals).
What I did here is to expose the basic guidelines of this syncretic approach of the modern technical analysis.
Why modern? It is modern because we want some kind of statistical validation and mathematical back testing and validation of the strategies, we use mechanical pattern recongition and scanning abilities. It is classical because here are the classical patterns that are involved (for example if you use a Neural Net extrapolation (let's day BPNN) I would advise to make a prediction when you would be away from those support and resistance areas and especially if you have a clear technical pattern).
The idea is to do the best we can do.
However futher testing is necessary and some kind of group that would test the ideas, sound like work indeed.
I would continue the discussion here.
There is an upgrade with the autochartist software. Now the basic classification is between expired and actual patterns. That allows you to make back testing of the tool.
And the actual patterns are devided as usual in emerging and completed. However you swith between those patterns by using the filter tab. And the tab appears when you click on the right Where is the text "results filter".
Is there any success story in the Forex market with Autochartist?
Also Pattern Monitor
http://en.ampmonitor.com/dashboard
Is it a free service? Thanks for the link.
7 days free. I think this best filter for autochartist.
Russian Pattern tool for Metatrader 5
http://www.alexstal.ru/hwafm
What? 7 days free!!!
Autochartis is 100 %. And even you do not need a real account for using the basic functionality, which is scanning and detection.
But we can create a new account with fast-mail tool. What filters you can reccomend for autochartist, concrete tools?
What really matters with the price pattern recognition is the statistics. In the best book for patterns is the Bulkowski book about patterns: Encyclopedia of Chart patterns. So it this book there is an estimation of the probability of each pattern, but it is for stocks.
And here with PRS Autochartist it is about Forex.
So make sure you check the Autochartis results:
You can data mine those results as the sample is really big.
Filter 1: Time of the day
What is obvious is that the most important filter is the time of the day.
Between 18 and 21 o'clock to play the break - out from a pattern is having real drop in the performence.
So logically you can even fade the break-out during those periods.
(However just make sure between the ibfx time and yours, it is probably not the sime time zone).
Filter 2 iVAR/ FGDI
The seconf filter is of course the iVAR, if you have a break - out you would be more sure if you have an iVAR reading below 0.5, or at least a dropping line (the indicator line should be going DOWN - that means an increase in the predictability).
If we compare with the FGDI it is below 1.5 level.
Filter 3 It is up to you
I do not know basically it is not a good idea to have too many filters. It is up to you what works for you?
Recentrly I was pointing that changing very often the indicators and their settings is not a very good idea for the discretionary trader. The reason is that your brain is trained to respond in the indicators in the same way an AI is trained to respond to indicators.
Thanks, I never played on real money, thats why for me scary, and want more and more filters.
I checked the current statistics of this tool and they are still great. I am wondering if somebody uses this.
The statistics are updated on weekly basis and they are from the type: too good to be true.