for investors and traders
Just the S&P 500 stock market index
with proprietary Artificial Intelligence.
No technical analysis, no venusian cycles, no esotherics.
It's just different.
robo-advisory for S&P 500
Do you need an unbiased approach to the market? It's not so different from the weather forecast. There is a model that reads the evolution of the present to give a vision of the near future based on a long recorded history.
It's artificial intelligence at work!
The spxbot.com website offers an insight into the development of a project that was born as a simple experiment and is now a fully working artificial intelligence (ai) that probes the S&P 500 (spx) index. The ai received a name, when considered stable enough, and its name is r.Virgeel, because you have to consider it has a guide through the ever changing sea of the stock market. It takes in account winds, weather, streams and all to provide an unbiased forecast of the S&P 500.
I'm Luca and I am based in Milan, Italy. I trained and practiced as an architect for thirty years, with a parallel activity in coding and in development of software tools with a variety of languages, plus a passion for investing, sailing and photography. I'm in the market since 1992. I begun the spxbot.com project in the Spring of 2013. I was lucky: the year before, I read the notes of Martin Armstrong, mainly composed by cyclical analysis and pattern recognition. I knew nothing about cyclical analysis, but I had a background with the use of neural networks since the early 90es. Before, I got in touch with the pattern concept by C. Alexander, growing as an architect. I put toghether all the pieces: code writing, database management, trading experience, pattern recognition, one by one, and in two years the ai (now r.Virgeel) begun awakening. This website is alive since September 2015. Since then, you may read the whole story through the blog. A selection of introductory posts can be find here.
Designing the model, I took inspiration from the weather forecast: there is a model that probes the present to give a prevision of the near future based on a long recorded history. Prediction comes out from the data, that form the database of past experiences. The larger the database, the better the forecast. This is very important, because macro trends appears at large scales and alter the behavior of the whole system.
Now you may wonder: this guy is kidding me, Mr. Market cannot be predicted.
thank you for the guy. Yes, you are correct, market is unpredictable, as an ever-changing-chaotic-bipolar system. Yes, but...
I came out with the conviction that a model for the forecast should cover most markets available, to give to the "brain" the possibility to build a very large set of correlations. Here lies the power of the r.Virgeel model, it does something that really no human may even imagine to try: correlate dozens and dozens of inputs and weight out a response based on the experience base provided. Without passion, without stress.
Market IS impredictable. Isn't it?
"Ok. But what if I want to know how the market performed in the past, under conditions similar to what it's happening right now?"
And then: "Instead of relating the price of the index to ALL markets available, why not have a reasonable wide selection of the most important?"
What sorted out, it is a large database of +200 historical quotes, a bit SPX centric, including forex, rates, commodities, stock indices, bond indices and other financial and econometric series. r.Virgeel brute force crunches a huge amount of numbers and it is trained to recognize how the market behaved in the past under current conditions. Since design, r.Virgeel is adaptive and responsive. It is higly deparametrized, so that it can generalize and see things that you people wouldn't believe.
r.Virgeel is trained to be a wise and prudent(*) investor and it bases its analysis on the correlation of the inputs without any human intervention. Data in, data out, and... whoop! you have the analysis every evening.
(*) The building of r.Virgeel undoubtedly reflects my own attitude and approach to trading. I'm mainly an investor, but not with the idea of holding forever. I'm aware that the market waves and we may take profit from this dynamic. I prefer to preserve capital than gain a fraction more. The training of r.Virgeel takes in account these attitudes.
As r.Virgeel was able to project and show the future path of the index, I verified that it is much more able to recognize the present, or see the SPX condition. Some indicators took life (and a number has died, as well), and, at the moment, the six basic r.Virgeel tools are:
Bars ahead - H/L/C is forecasted for the next 24 bars, neurally calculated
Target - where the current wave is heading, it is a confirmation tool, neurally calculated
Stop - a value that confirm the trend and generates alerts, a position following tool, neurally calculated
Position - a simple, but detailed, neural swing system, a signal generation tool, neurally calculated
ColorCode - easy green/red feedback of market status, neurally calculated
NextTurningPoint - evaluation of next coming turning point, neurally calculated
Signal - visual feedback on the chart, a signal generation tool, neurally calculated
Bars ahead and Target, they all snoop into the future, while Stop, Signal and Position are readings of the present status of the market. Forecast charts are updated daily, using closing values All signals are generated to be executed (when necessary) at the opening of the next bar. You may see a sample of r.Virgeel daily report here. (please note: as development goes on, there are some discrepancies between the shown charts and the present charting. I work for getting a simpler and more effective information for subscribers, with their contribute and support).
F.A.Q. Some frequently asked questions:
Does the model forecast only the S&P 500 index?
Yes. In a general view any of the components of the model may be forecasted, but the long development phase showed me that the quality of the inputs is crucial. In my opinion, data from many other markets or commodities or economics at the base of currencies are not so good to give a sufficient quality. Being the S&P 500 the most actively traded stock index in the world, I decided to concentrate on it, so that the challenge is at the highest level.
What is "the model"?
Numbers shown in charts are crunched by a set of backpropagation neural networks (BPNN or network or nn) fed by a model or data sheet that is basically composed of historical data from a sample of markets, financial instruments and economic indicators. Think to the model as a long term collection of experiences. Since 2017 its name is r.Virgeel.
What is "the network"?
The neural network is the code at the heart of the beast. It works a reverse logic: if I reduce errors to the minimum, I get the nearest desired result. It is subtle, because it does not garantee you that it is the real correct and precise result, but it is the only way to have numbers projected in the future. Networks must be "trained" or, in other words, the form of the surface that represent the solutions must be computed in advance. Training is a vital part of the whole process and the performance of the ai is largely affected by the quality of the training. R.Virgeel is instructed mainly by hand, and just some small minor task are automated. Then, when ready, the network is fed with never seen before data (today) and it shot its readings, incredibly fast. The neural network is a powerful diagnostic tool, because it can easily do jobs that other code would take an enormous amount of effort and time, or even be impossible.
How is the model composed?
The model contains main stock market indexes from around the world, major currency pairs related to the Dollar, commodities, precious metals, other financial indexes, financial indicators, economic statistics, rates, yields and bond market indexes. Three distinct model are maintained, on daily, weekly and monthly time frames. They are trained separately. As a sample, now (Summer 2018), the daily model is fed with +70 data sources, the weekly model is fed with +120 data sources and the monthly model sums us +70 data sources.
The resulting brain structure is in the scale of millions of neurons.
May the model produce a wrong forecast or miss a signal?
Yes, for sure. It is rare, but it happens. When the market moves under totally never-seen-before conditions, in respect to the experience incapsulated into the model, the network is forced to guess a result and so may generate a wronger forecast. It happens rarely, but a sort of blindness happens sometimes to r.Virgeel. You need to understand that neural network output is always wrong, depending on the measure scale you adopt. Due to the reverse logic applied, you always get the less improbable result available. The r.Virgeel model has a nicely long history, through many different market conditions and signals are revised constantly. Signals are different from forecasts, as the model doesn’t have to project something, but instead it has to read a stated present condition.
All signals works together, so that you can easily get the big picture.
Is SPXBOT a trading system or any sort of automated trading tool?
No (and Yes). Not automated, in any way. Again, think of it as the weather forecast: it doesn't tell you to go fishing, but if you want to go fishing you better get a look to the weather to get prepared for changing conditions. In trading activity, you decide, you act and your is the full responsibility. So many different trading style! On the other side, the signals output from r.Virgeel mimicks the optimal investment behavior, marking extreme market values and probable reversals and following trends. Signals are generate on all three time frames: daily, weekly and monthly bars. More than a trigger, the signal coming from r.Virgeel is a suggestion, as if it were saying: "hey, the market is in accumulating area, switch on your buying long strategies" or "we are entering a chaotic lateral move, better stay out". If you are a self-aware investor or trader you may get big advantage accessing the r.Virgeel numbers. I've wrote some notes in the blog about the psychological benefits of using a robo-advisor.
It happens that sometimes, from day to day, the model generates totally different views of the future development of the market. How to deal with it?
First: there is a lot of noise in the daily time frame to deal with, and this happens mainly in the daily time frame. Then: this is a very delicate point. Let me explain: under very similar conditions, market may have acted in the past differently: let's call them bifurcations. Dealing with long term data, there may be different alternative outputs that may be very "near" and sprout from present conditions. When passing through these events, we may observe output flipping in forecasts or repetitive pattern generation. The different time frames help to maintain a correct view of the market, during this usually short phases. The worst case I remember was the daily forecast stuck and blind for 5 or 6 bars, before resuming the usual responsiveness. It was about 1 year ago and some new improvements in the code seems to have greatly reduced the length of the defect. Globally, latest code revisions have reduced the noise of the daily forecast, that now compete with the weekly.
What do arrows mean in the chart? Are they a buy/sell signal?
Arrows are identifiers of inversion areas, either medium or long term. False signals or missing signals may be possible. Strictly, they are not buy or sell signals, as they are generated based on close of bars. Intraday action may override extremes. Arrows' calculation is anticipated by one bar, so that you may act at the opening of the next one.
The development of r.Virgeel and spxbot is fully accessible through its public blog, where you may find the documentation about almost all aspects of the tools, indicators, charts, signals, development, performance, psychological aspects and all.
To start, you may read a selection of introductory posts.
Thank you for your interest.
Subscribing to spxbot, you will support the project's development, you will get daily updates and you will get full access to r.Virgeel's data. The data you will see is the data I see, nothing is kept reserved.
Here follow the subscription plans available:
What do you get with the subscription?
You gain immediate access to all present and past materials published in the reserved area: the full history of r.Virgeel is available.
Every day, after market close, the daily r.Virgeel is updated. Every Saturday morning (CET), the weekly update is brewed and delivered. Every first of the month, it's the turn of the monthly forecast update. Weekly and monthly updates are delivered in your email box, all updates are available on the web site in the reserved area.