An AI Trading experiment


An AI Trading experiment Visit (post n.7)

Never miss a news on Spiridione, subscribe now! Low frequency email, no share of your contact. You can unsubscribe at any time.

An AI trading experiment,  observed and improved by humans


Key assertions:

  • The combination of the concepts “collaboration is better than competition”, “skin in the game”, “Human and Artificial Intelligence”, “Freedom from the Ego”, has generated
  • is an experiment finalized at better understanding if AI Trading can perform consistently overtime;
  • In order to do so, it has been created an AI trading Engine that operates daily and a website which allow free subscribers to observe and comment on the experiment;
  • targets positive returns whatever the market regime is. The selected market is NASDAQ 100;
  • Market regimes combines period of low/high volatility with bullish/bearish/lateral trends; 
  • The Machine Learning algorithms select best criteria on a daily basis in order to make the best predictions (one month ahead)
  • For each stock (104 stocks at the moment) optimal one-month stop-loss and take profit levels are defined
  • The system elaborates long/short probabilities for each stock, both with a statistical and ML approach
  • The combination of methodologies generates a daily portfolio with 6 stocks (long and short, 10.000 $ to each stock) which will deliver its results after 30 days (a stock can early terminate its journey if meets its stop-loss or take-profit level)
  • Follow and free subscribe to to learn more

I’ve always been a mediocre tennis player. Despite this and so much money spent to improve my game, when I played against an opponent who was in trouble, I tried to give him advice to make his game more effective.

It often happened, maybe because I was more focused on giving suggestions to my opponent than to keep the ball in the field, that I would eventually lose the game.
The strong proposition of finding mutual way to solve problems and find win-win situations accompanied my entire professional life (definitively not my sport curriculum).

I never liked competing, I definitively prefer collaborative models.

Collaboration is better than competition


As a young teenager I had a Sinclair ZX Spectrum, one of the first home computers.
I learned to program it and to develop simple applications to predict the results of football.
I dreamed of discovering the way to foresee something meaningful and to make it available to others that could further improve the system.

Sinclair ZX Spectrum


My professional career led me to be a manager in the field of energy and public services, with commitments and stress that left little room for anything but work.
I have never forgotten the dream of predicting important things and of collaborating instead of competing.
Five years ago I came across some readings on artificial intelligence and its ability to learn from existing data and make better and better predictions.
I discovered Machine Learning and Python, a powerful language with rich libraries to program any application.
I gave up on managerial strategy books and relaxing TV series and I started programming with Python.
The enormous quality of information on the web, along with courses for beginners on Coursera, Udemy, Udacity, allowed me to create codes that worked, even if not in a fully pythonic way.

I focused on the stock market forecasts, a fascinating sector in which experts struggle to find scientific techniques to invest.
So if AI is superior to human intelligence in uncovering recursive patterns in the data, then this is the ideal field.
Some highly successful hedge funds claim they found an edge to beat market indices through the use of big data.
This capacity generates alpha, that is, the differential advantage of allocating money in a hedge fund rather than simply buying the reference index (SP500, Nasdaq, etc).
With the help of knowledge repositories, freelancers, experts, web developers, the first version of is online.

The system produces the results I wanted: daily portfolios built with the aim of obtaining positive monthly results in any market regimes. The system reacts to the data that is feeded with daily and tries to improve its forecasts. Benchmark vs Equity Line Portfolios generated in the current month

To view the complete results simply go to and register, in addition to what is available on the first page in the section reserved for free subscribers, many details will be available.
But in all this, how does the concept of collaboration (better than competition) come into play?
Those who register will be sent questionnaires to receive feedback on their vision on the experiment, on their suggestions for improving the tests, on further developments that could be planned to implement more performing algorithms.
I am convinced that further technical and programming developments will necessarily have to be carried out by professional programmers: the first version of will be soon forgotten and new features will be added if the engine is successful.
The 2018 development plan of is already in execution (see the PowerPoint presentation for further details).
The intent, if the solution works, (i.e. it has a real edge compared to the benchmark), is to launch a crowdfunded startup and build a professional platform for investment, based on artificial intelligence, but continuously improved by human intelligence.

Here’s what does at the present moment:

– updates the closing data of each security in the Nasdaq 100;
– calculate the technical indicators;
– update all descriptive statistics;
– apply Machine Learning classifiers to updated data;
– identify the best classifiers of the period;
– obtains the best long and short stocks for the next month;
– combines the selected securities into a daily portfolio that will last one month

After the portfolio has been generated, monitors the progress of share prices on a daily basis, stops actions exceeding the limits of profit and loss, and balances the results of the expiring portfolio.

The expired portfolio contributes positively or negatively to the performance of the system which is compared with the performance of the Nasdaq 100.

A series of statistics on the progress of the system are updated every day. Key Statistics

The system shows in the landing page of the portfolio generated the previous day.

The macro forecasts for the Nasdaq100 and the main statistics of each individual index component are also reported.

For free subscribers, within the reserved section, there are 30 active portfolios (those ranging from 1 to 29 days of maturity), as well as the complete slides of the presentation of

The systematic comparison of the 3 strategies that will be implemented in 2018 will be added in the coming months.

Artificial intelligence will be the new internet: a new industrial and social revolution.
Just as steam machines have brought development, but also the exploitation of natural resources and the consequent pollution, AI will also bring benefits and negative impacts on our way of life. is an experiment, in which the machines will do the work autonomously, human beings will observe and hopefully contribute to building a new way of investing.

Thank you for your attention and we stay in touch!

Tags: , , , , , , , , , , ,

About the author

Spiridione, my middle name, has strong Greek origins, and Greek philosophy is still current and pervasive, which is why I called the website

During my adolescence my preferred books were those that covered philosophical questions. At 19, I moved to Milan and then graduated from Bocconi University with a degree in Business Administration at 24. – Read more

Leave a Reply

Your email address will not be published. Required fields are marked *

In today's information society, our valuable ideas last the time of viewing a post. Wonderful concepts are drowned in tons of nothing. This blog is an attempt to resist and discuss things that can give meaning to existence.

Never miss a news on Spiridione, subscribe now!

2018 Copyright all rights reserved | VAT 01573380332
Privacy Policy