Hi Herschel O-Walker, you posted an interesting question. In brief, designing good algorithms for trading does not differ from designing good algorithms in general.

In principle there is no "Golden Recipe" for creating a good algorithm. Algorithmic design should always take into consideration many factors like the following

1) What are the most common cases in which the algorithm will execute? For example, sorting algorithms behave differently based on the type of input. So there is no globally superior sorting algorithm.

2) What are the available resources and at what cost? How much will you profit from trading for every ms gained and how much will it cost you in infrastructure and other costs?

3) What is the complexity of the algorithm? What are the possibilities of bugs and misbehaviors due to unavoidable complexity? Can the complexity be reduced?

and many more...

Ending these are questions that mostly fall into the software engineering profession and are not restricted to trading profession only

Thanks a lot Panagiotis Charalampous for the response - much appreciated!

Does a trader necessarily have to have programming experience in order to design an algorithm? Are there any kinds of trades which are not suited to this form of AI?

Hi Herschel O-Walker, a trader is the stakeholder of the algorithm and a programmer is the builder. The first one answers questions like what and why and the second one answers questions like how. If the two can combined in a single person, then that is good!

I believe that all trades are suitable for some automation but we are still far away from robots that execute trades unintended. Currently robots are used as tools in the hands of traders to execute automated and repetitive trades more efficiently.

Thanks again Panagiotis Charalampous - FINAL question from me: some people believe that in tandem with IOT, in future our washing machines and refrigerators will be the ones trading energy autonomously - do you believe this is the case?

Without wanting to sound like a Luddite, I think that this is still a far fetched scenario. If you think of the fact that we know all the rules of proper driving and proper aircraft flying but we still face issues with self driving cars and self flying planes, imagine how far we are from automating trading completely, at a time when technical analysis is considered by many a pseudoscience and trading seems to be more of an art rather than an engineered profession.

To start with your last question, everything goes towards automation - it is estimated > 70% of the trades are executed automatically, to trade in the future one will need the aid of a computer when it comes to intraday trading where fundamentals are not used. As to your first question, yes there is both an automation software and a non-programming software in the market for designing and managing algorithms.

As to robots automatically on their own develop automated trading strategies, and run them - this is today possible, and i fact available. However, such solution would not be for the masses, else we would all become billionaires - but such would rather be used in a closed forum or probably by the person/firm who invented the robot. Since we will never come to a point where any user can repeatedly push a green button Monday morning, and a red Friday afternoon having a profit of 1.000$. I say best scenario is the user always should have an influence on the outcome of any financial trading.

@Pangioitis wrote "trader is the stakeholder of the algorithm and a programmer is the builder." When the trader is using a non-programming development platform, we will be both the stakeholder and the developer. The trader is the best to develop the algorithm, since he knows about trading - the programmer does typically not know about trading or what is in the mind of the trader. Apart, we have the confidentiality problem where the trader has to disclose his idea, and the money and time problem. The trader-programmer setup is not operational, and very costly choise.

I agree somewhat with the other commenters. Algos can help to make decisions faster, but they can also cut through the noise. E.g. for news trading, 97% is noise. Natural Language Processing can filter the news and decide whether it is good or bad for a stock. You can see an example for this on my webpage abaca.ai. But also a combination of human knowhow (e.g. indicators) and algorithms that identify which one to combine to trading strategies can make money. You can also play around with this on my page. It is just for demonstration purposes, but if you want to setup something like this, I can help to get you started.

No offense but once highly efficient and profitable algorithms are created what on earth would get into someone to share them? Think about it. If I created an algorithm that could produce a 200% return (or greater) per day, why would I share it? What benefit would there be to sharing it? If you had a virtual ATM machine in your living room that you never had to refill, why on earth would you give anyone a key to your home. I understand the point of asking the question. Hopefully someone will be stupid enough to allow their ego to get in the way and give away their secrets.

Robots can help identify investment opportunities very fast. The question is if they are able to bring the information into the right context. Some simple investment strategies are still difficult to realize with robots. Automated detection of support and resistance levels can help to realize new products. Yes, I agree with you. Algorithms are the "future" for the trading sector.

There are many types of algos. Simple algos for order entry like icebergs, time and volume sliced orders. There are order management algos like adding logic to work an order with a tick or placing OCO logic around a resting order. There are of course round turn algos as well that open and close positions. Keep your mind open. There is much, much you can do in the algo space. If you can draw a flow chart of your algo idea on a piece of paper you can likely build that algo. You're welcome to create a free account at Trading Technologies and use our algo building tool ADL to give algo development a try. No coding is required. We use drag and drop functionality to piece together order logic. https://www.tradingtechnologies.com/try-tt/

Ok so the reason you would sell a strategy or a system generating 200% return would be because the fees you charge on the billions given to you is far more the 200% on your own capital or the profit on selling the system is far greater than returns on your capital. Initially you would do both invest using the system and sell it in some form.

I built one that does 200% every 6-7 months as long as you have a minimum of $1 million to invest. I'll share it, source code and all for $25 million. Msg me if interested.

William Schamp Thanks for the feedback, definitely see where you're coming from and it's a very valid standpoint. My goal in this instance however, was just to start a discussion - if it helps!

Mine is $8.5 Million plus 20% of profits in perpetuity or just subscribe before the Maximum Subscriber limit kicks in.

I cannot see everybody developing algorithms in the future simply cause people are lazy and want to get rich fast. I was in contact with a person who has used 4 years and a high amount of money to try find an EA which can make him rich. Even you told people they could develop non-programming wouldnt draw peoples attention. Very very few succeed developing, some spend a decade getting no where - visiting the forum now and then the story is the same, no one has got anywhere. They only change in this forum is people leave one by one and do something else. Which should be a warning before you spend a decade chasing the millions.

Joachim Arndt Peter Button Søren Lanng Thanks very much for your insight! One question I have: do you think, rather than seeing it as purely a way to set yourself up for "free money" you are instead diverting costs from day-to-day trading to more behind the scenes development and design of these programmes? Almost like how the ATM machine didn't spell the end for bank tellers, rather it led to them opening more banks so tellers could do more sales/customer service?

If you are making 200% return per year, no legitimate fund in the world will give you the time of day; you are seriously over leveraged or it is a scheme of some sort. Get 20% per year and you are in rock star territory.

Herschel O-Walker, your comparison with the bank teller seems right. For me, I use algos because they can trade a lot of securities for me while I could do only a few. But it is also because I can find new strategies quicker with algorithms than on my own. I am a believer in good technology is there to complement human beings, not to replace them.

Thanks Peter Button -- glad that I'm not barking up the wrong tree!

Herschel, it is the future rest assured, and it is here now. Make sure you can scale over a billion dollars. Control the volatility of the aggregate value of the portfolio. Here is your job you lose 10% any time you are fired. Fired if you lose more than 1.25% in a day.