Thursday, February 23, 2017

Next generation Trading software

Introduction

The next generation of trading software has already arrived. The software takes advantage of heavy operating computer force from the cloud, as well as artificial intelligence in order to get an edge over the market. Read more in this blog about the constant development of the advanced software from ASA Trading Software.

Welcome to ASA Trading Software

Welcome to the first article about ASA Trading Software. The objective of this blog, is to provide a deeper understanding of how we will be working and developing ASA Trading Software ongoing.


EA, also known as Expert Advisor, is and ordinary expression for so called trading robots. I would like like to emphasize that ASA Trading Software has a completely different objective, and a much more complicated that those so called EAs. The EAs are limited because they often are limited to e.g. programming in a piece of software called Metatrader. The programming language MQL4 is very limited, required a lot of heavy work and does not include a database. In the event that there are many calculations, the MQL4 can not handle the code. Therefore, we have been forced to develop our own system.


Moors law has made a very powerful cloud based computer force available for a far greater audience. With API technologies such as Google, Amazon, IBM and Microsoft, it is now possible to work with artificial intelligence in the cloud. And why exactly is that interesting?


Information has always been a power factor, however, today we are practically flooded with data. So in this modern day, the art is to work the data as fast as possible. Preferably in real time. This is now a possibility by performing a real time probability assessment,which is completely correct probabilistic.


Visual recognition technology, is being trained to be able to recognize all the patterns that a trader uses. This can be characterized as a technical analysis. The technical analysis compares historical data with real time data, in a manner where the result is probabilistic correct. Furthermore, every fundamental data in the calculations are added, thereby mapping out all the parameters that could influence the price development.


ASA Trading Software is therefore a probability calculations machine. The more horsepower added in the form of cloud based computer force, the closer we will get to Real time.


The future and next generation software will be a competition on being the first with probability calculations. Because in the future there will more and more people taking advantage of deep learning, and artificial intelligence in order to get the most probabilistic correct assessment.

In the next article, I will be talking about a few examples on how we train the artificial intelligence in order for it to make its assessments as good or better than humans.

Friday, February 17, 2017

How do we work with artificial intelligence?


Introduction

Artificial intelligence can be a tool to understand and comparing an uncomplete world. A tree can be drawn in infinite ways and likewise a pattern on the market can take a lot of different variable into account. Finding structure in imperfection is the main objective of artificial intelligence.


ASA Trading Software and artificial intelligence

To find structure and being able to compare patterns, crisscrossing in an imperfect world is a very central part of the next generation of trading software. With good reason, a lot of people will claim that well known robots or EAs (Expert Advisor) can not compare to humans. However, that is only because EAs are rule based system with no flexibility.

Let us take an example in a descending triangle, which is a pattern being traded upon by a lot of traders. In the figure below you can see an example of the pattern.






The assumption is that activity and volume will fall at some point in which the market will break down toward the horizontal resistance range. The resistance range is an area where the price has a hard time breaking through, which can be seen when the price reaches a certain point, it will be returned.


There are no rules in regards to how long a period of time, a descending triangle can be built up over. Furthermore, there are no rules for how many times it has to reach the horizontal line. There are no rules for how big fluctuations will be or how much they will have to fall towards in order to break through. However, that is a pattern which a human can learn to see relatively fast, but it is not as easy to identify with the help of traditional code.


Now, let make the example a little more complex. The third time its stops upwards it will break through the diagonal line and generate another known pattern known as a hammer. Here, a trader will say there is an obvious price deviation, and thereby get another confirmation that the market will go down. How a hammer looks can vary a lot. The more patterns involved, the greater advantage humans get compared to traditional EAs. Because EAs, as previously mentioned, can not handle too complex code because the system will break down.


Artificial intelligence is much different. As a starting point, it knows nothing. It can be compared to a child, or a completely new trader, who has never seen a descending triangle or a hammer. We will e.g. go through hundreds of different variants of descending triangles, and hundreds of different variants of what looks descending triangles but actually is not. Furthermore, we will ask our software to go through historic charts in order to find all the available descending triangles. By itself, it can go through 60.000 pictures a day. That means in a single currency pair such as e.g. EUR compared to USD, you will be able to check multiple descending triangles from the previous 15 years. So not enough with it quickly being able to identify descending triangles from the last 15 years extremely fast, it will also get an exact picture of the patterns historic success rate.


The next step is training it to know all other known and traded patterns. Any pattern is trained separately. So if it knows e.g. 10 patterns, it will be able to identify those patterns from a picture instantaneous. Furthermore, it will be able to go through and find success rate across all the different patterns. There will very quickly arise infinite possibilities for combinations, which required heavy processing power.


Luckily, this processing power is available on cloud based systems from the big IT mastodons such as IBM, Microsoft, Google, Amazon and such.


So every time a pattern is added, the required processing power increases exponentially. Therefore ASA Trading Software will quickly learn to perform better than humans. Because no human can work with the exponential curves while simultaneously reflecting on it historically. The man with years of experience can be surpassed in a matter of minutes on a supercomputer. E.g. Jeopardy, Chess, Quick-Draw, Google-Search and Poker has all been beat by artificial intelligence.


ASA Trading Software is constantly being fed more and more patterns whereafter it starts teaching itself meanwhile building a database. The database uses the same access as search engines, so the results are accessible fast and with a relatively low use of computing power. The lower the use of computing power, the closer to real time you will get.


Search engine technology actually uses artificial intelligence. Therefore, ASA Trading Software actually contains a search engine. So we are using the artificial intelligence to build a structured search engine in a imperfect world, that can deliver probability calculations in real time.

It costs ressources, knowhow, time and money. This is why we sell our trading software, so all expenses can be divided amongst its users.

Friday, February 10, 2017

How do we work with risk?

Introduction

Murphy’s law says that if something can go wrong, it will sooner or later. Because we work systematically and probabilistically correct, that even the most improbable event will occur at some point. In order to prevent the most improbable thing from becoming a nightmare, ASA Trading Software has been built up in a way, that takes every scenario into account that we have deemed improbable.

How do we take the improbable into account?

We know that e.g. disasters come ongoing on the stockmarket. In the last twenty years, we have had one in 1999-2002, 2007-2009 and a bigger fall in 2015. These falls on more than 50% are inevitable. We also see flash crashes over a few minutes. The last one was in 2016 with the british pound in october. They are very unpleasant and have cost a lot of people a fortune.
When you invest in a stock, the risk is in principle 100%, because there will always be a chance that a company goes bankrupt. When you invest in a stock index, you risk will not be a 100% because that would be the equivalent to all companies going bankrupt. However, the risk is still more than 70%.
Bigger falls can be fool proofed by parting the curve in intervals. In the event only allow one trade pr interval, it can be calculated how much you need in your account in case of a fall of e.g. 90% amongst all companies.
It is actually much worse concerning Flash crash. One of the worst cases was when the Swiss franc (CHF) was released across from the euro (EUR). Again, there is only one way to fool proof against such a fall. Only have a few open trades in regards to your balance. 
Correlation analysis can minimize the general risk, however, it will also take a hit in the case of CHF.
ASA Trading Software is not immune for these instances, but the loss is always in control. Software will be selling ongoing in order to keep the risk down. Furthermore, we are keeping very strict tabs on the intervals, so any negative fluctuations only will give a temporary unrealized loss, much like investing in stocks.

When a Flash crash takes place, lengthy trades are taken into account by only being able to open a new trade once every four hours. Which means, that if the improbable should happen, and the stock market falls 40% over night, we will only have one extra trade open. So a unrealized loss is lesser.
ASA Trading Software also trades on the background of news, in this part we are only in the market for a few minutes to an hour. Thereby holding the amount of open trades down to a controlled amount.
We can not foresee the unpleasant situations, but we can come through them with limited damages.
Quite another thing is the technical. You can get hacked. We can lose our connection to the server. These things have also been taken into account. Even though the instructions come from the cloud, the system has been instructed to which parameters needs to be instated when the connection is gone.
Because everything is controlled mathematically, we even know exactly how the system will react. So if there is suspicious activity on a customer's account, we will be notified immediately.
We have people sitting in Denmark and Bangkok. Thereby having 24/7 human coverage.
Basically we have a plan for everything we have predefined as a risk, but of course there is always the probability of a scenario we have not thought about.

Thursday, February 9, 2017

Why do we search for customers but only in a limited extent?

Introduction 

 Customers are a condition because the system is so complex and advanced, it requires ongoing cost heavy development. Furthermore, there are many advantages by being directly connected to the market, which is quite expensive. So sharing the ongoing expenses is a condition for being able to develop the best possible system on the market.

Without customers we would not be where we are today 

 You could call it win-win. The customers give economic possibilities. Furthermore, we have also developed a concept which we call 97%. Here we share our knowledge with our customers. Teach them up with the hopes of them one day finding new ways of approaching the matter, that we would not have thought of otherwise. More elements from ASA Trading Software comes from this network, and here we are talking world wide. That is amongst one of the reasons that our development lies in Bangkok.

Things are moving really fast in Asia and developments can be kept down. One very big expense is to the cloud based computing power and usage as well as the development of artificial intelligence.

Furthermore, there are expenses to Reuters, which gives access to news in a few microseconds after the news are out. Likewise, we also work on trading directly on the exchange in order for the reaction speed to be brought down. All of this will never be realistic as a private person. The customers needs us because we give them opportunities that they would never themselves reach, likewise we need them due to the ambitions of the system.

Tuesday, February 7, 2017

How do we work with historic data?

Introduction

A very central concept in ASA Trading Software is historic. There is always a probability for a once performed action will duplicate, and thereby you can substantiate different possibilities. There is never a guarantee the history will repeat, however, as long as it does it significantly more than 50% of the time, history will be the best indicator for getting a picture of the future. 

The more history the better

History can be economic news. It can also be technical patterns or data about Spread, volume ect.. Basically anything that in principle can affect a price is interestingly enough in regards to history.

As previously mentioned, we use deep learning and artificial intelligence in order to get historical picture of the technical patterns.

It is a bit easier working with news, also known as a fundamental analyser, because all it requires is a number. But again these data can be crossed with the technical analysis. Once again search engine technology works in order to get real time results.

The challenge is, the more data you connect to every search result, the lower will the population result become, and a result become less valid. This is where nominal distribution comes in, in order to rate the precision and validity of the result. The less spread of the variable, the better we can trade on background of the result. Therefore, it is not enough having only one causality and a probability of 90%.

Luckily, all of this can be automated with software, so a correct probabilistic assessment can be performed. Statistics is therefore, just as important in order to get the most optimal out of the history.

Thursday, February 2, 2017

Probabilistic calculations in real time will always be updated

Introduction

The whole point of the above standing approach, is that ASA Trading Software will always be updated. Today, people are rating the different strategies. Some work on some pairs and certain periods of time. This issue is not relevant anymore. ASA Trading Software will always chose the correct strategy, no matter how times change.

The perfect probability calculator

A probability calculator indexed and optimized by search engine technology, will be as fast as any possible strategy that will be delivered in real time.

The indexing for search engines can be performed on powerful servers ongoing, like Google does, and when the market moves, the artificial intelligence will recognize it and come up with suggestions before the market takes form. So when a descending triangle is 50% done, we will already have an idea of all the possible strategies that have a high probability. Equivalent to instant search. You start off by writing a search term. From here a matter of alternatives will be provided before the words and sentences have been written.

All strategies are based on different patterns and parameters. They never have all of the parameters. Furthermore, they do not have the entire history either. Moreover, they do not have the associated statistics to rate the validity. All of this is put together in ASA Trading Software. ASA is the future, and the next generation trading software.