Luis Fernando Molina


Superalgos, Part One:  The Trading Singularity

Image © GarryKillian, Shutterstock.

One day in the future, a trading intelligence capable of outperforming every other entity at the markets will emerge. Both humans and current algorithms will be surpassed by Superalgos.

By Luis Molina & Julian Molina

AI is a hot topic these days.

Elon Musk is still worried about the risks of a Technological Singularity happening as a result of advances in AI.

You know, that moment in time when an Artificial Intelligence starts improving itself 24/7, non-stop, producing an exponential growth of intelligence that — in a short period of time — would render the most clever people on earth stupid.

Such worries, of course, are justified…

We know what happens when a species becomes so much more intelligent: the rest end up in zoos, circuses, grown as food or extinct. The lucky ones may end up surviving in the last pockets of nature out of reach of the dominant species — for a while.

The reason why the Technological Singularity has not happened yet is because machines are extremely good at what is called Specialized Intelligence (dominating a specific task, like playing chess, for instance), but are still pretty bad at General Intelligence (interconnecting thoughts about multiple different specialized topics at the same time, in an intelligent manner).

Put in other words, machines can outperform humans in specialized intelligence but still lose miserably in terms of general intelligence.


A supermind is a group of individuals acting together in ways that seem intelligent | Image © GarryKillian, Shutterstock.

Scientists have found that computers and people thinking together can become more intelligent — as a group — than computers alone.

Professor Thomas W. Malone and his colleagues at the MIT Center for Collective Intelligence spent years studying the collective intelligence of groups of people, machines, and groups of people and machines working together. Extensive research and conclusions are summarized in his book Superminds.

The Superalgos project is framed around the perspective of Superminds and Collective Intelligence. | Image © Thomas W. Malone.

A key takeaway is that collective intelligence can be measured and that there are factors correlated with the intelligence of a group. In essence, existing science provides clear guidelines to maximizing a group’s collective intelligence.

Another worthy conclusion distilled from research and actual empirical evidence from collective projects such as Wikipedia is that hyperconnectivity, that is the ability of vast numbers of people to communicate with one another from anywhere in the world in real time and at nearly zero cost, makes massive superminds possible.

Trading Intelligence

“All trading is gambling”… I first heard that phrase from my friend Daniel Jeffries and I’m sure you’ve heard it too.

Whether you agree or not, the fact is that — like gambling — trading is not just about chance. It is also about probability, game theory, statistics, fundamentals and many more fields of knowledge that can be applied to make better trading decisions.

Today’s trading is still done both by people and machines.

My guess is that, in the future, it will be mostly machines doing the job, but as research on collective intelligence suggests, humans will still be behind the curtains working together with machines, each focusing on what they do best: machines will use specialized trading intelligence to perform the actual trades and people will use their general intelligence to improve the algorithms used by the machines.

Is it possible that a singularity-like event ever happens in the specific field of trading intelligence?

May superalgos capable of outperforming every other trading entity ever emerge?

Let’s assume for a moment the answer to both questions is yes.


What is the path leading to the emergence of superalgos?

This is the question that keeps the team at awake at night. We’ve spent short of two years crafting the answer under the radar, and are now ready to open up the doors for more people to join.

The rest of this article lays out our vision for the path that will allow superalgos to emerge and tells the story of how we started building on our vision for the future of trading, step by step.

The Path to Superalgos

The path is in fact a long list of problems to solve. If we can solve them all, superalgos will emerge. However, the list is far from being a collection of technical problems alone.

Because general artificial intelligence is still a distant goal of science and technology, superalgos will emerge from a group of machines and people thinking together.

We would argue that the easiest part is to get machines to think together. It is a lot more challenging to incentivize and coordinate a large number of people to think together in a way that they contribute pieces of intelligence to a supermind of humans and machines.

Exponential Increase in Intelligence

The current state-of-the-art in the markets have both human traders and algorithmic traders, also known as algos, bots or trading bots.

Under a certain light, algorithms can be classified in amateur and professionals. Amateur are the ones created as a hobby by aficionados while professionals are the ones developed by companies, small and big.

Under a qualitative lens, we could also argue that professional bots are those whose quality would allow them to be marketed.

It is safe to assume that the average amateur bot is not very intelligent and that its intelligence increases slowly with time. The increase in intelligence of amateur bots is limited by the capacity of the person or small group of collaborators participating in the development of the bot.

On the other hand, the average professional bot is already quite intelligent, and its intelligence grows at a faster pace. However, the increase in intelligence is limited too, only this time the limit is set by the talent hired by the company, which may be significantly larger.

Still, most companies work under vertical segregation schemes by which subsets of the overall firm’s talent work in their own silos, sometimes due to profit center divisions.

Image ©

It is also worth noting that, except for amateurs who might use open source code, professionals don’t share their algorithms. On the contrary, they do as much as they can to keep them secret because they assume markets are a zero sum game.

No company wants the rest to know their winning formula, otherwise algorithms might lose their edge.

For superalgos to emerge surpassing both amateur and professional algorithms, our system’s intelligence needs to increase an order of magnitude faster than everything else in the industry.

Only in this way it is possible to start with little intelligence and end up outperforming both amateur and professional bots in a reasonable timeframe.

Companies’ intelligence is constrained by the number and quality of human resources they can hire. What if we could incentivize an ever-growing crowd of traders, developers, analysts, aficionados, etc., to become a part of a supermind and contribute their intelligence to the collective?

Once we understood this, the ever-growing crowd became a necessary ingredient in the system.

We started imagining a project involving such a crowd.

How would this community be organized?

What mechanisms could be used to transfer people’s intelligence to the project’s supermind?

We found the answers inspired by evolution in nature.

Evolutionary Model

Image © Yurchanka-Siarhei , Shutterstock.

We looked in nature and started playing a little mapping game…

Amateur bots are as intelligent as worms, while professional bots are as intelligent as birds. In such scenario, the project’s target is to become as intelligent as humans, starting from the intelligence of a unicellular organism.

Tough job, right?

In biological life, a cell has DNA which contains genes. In our model, bots would have source code, configuration and parameters. Many biologists compare DNA to source code in computer software, so the mapping seemed quite straightforward…

Cells have a biological life ruled by the laws of chemistry and physics. The cycles of life, reproduction and death define evolution, so this lead us to believe that algorithms should also have some kind of life… maybe a financial life?

What could possibly mean to be financially alive?

Companies have a kind of financial life: they have an income and expenses; when they run out of money they go bankrupt and usually die.

Companies are created by humans, bootstrapped with capital and in turn pay dividends to their creators. They also are superminds — groups of individuals working together — and, as such, have intelligence. They are regulated by human-made laws which give them the status of legal persons.

We thought that if our algorithms were to have a financial life we should expect them to have an income and expenses to pay. If they would run out of money they should die, for the system to have a first layer of natural selection.

Income would need to be related to their trading performance, so that the best performers have better odds to survive and reproduce.

Someone would have to create them — the human crowd — but at some point in time they should become autonomous so that they can live their financial life independently.

Creatures in nature are created by their parents and become autonomous sooner or later, which pushes them to fight for their own survival. This fight for survival is one of the key drivers of evolution.

Things were starting to take a recognizable shape, to the point we started calling algos financial beings.

We just needed to figure out the rules that would govern financial life.

It was clear to us that it wouldn’t be arbitrary human-made rules, as the rules governing different forms of life are usually defined by the environment they live in.

In the case of financial beings and our evolutionary model, the environment would be a software platform connected to the markets.

This lead us to understand that free market laws were to govern financial life.

Markets would be the unpredictable ecosystems in which financial beings would do their best to thrive, interacting with the environment pretty much like animals and plants do with their ecosystems.

Early financial beings’ actions would have little effect on the markets, pretty much like plants or animals’ actions do within a specific mountain, river or prairie ecosystem. However, as intelligence and dominance increases, bot’s effect in the markets would become progressively more important, like the growing human footprint on planet Earth.

It took millions of years for humans to emerge out of biological life’s natural evolution model.

It was obvious that mapping our model with natural evolution would not suffice. We needed an accelerated model that would allow for rapid evolution.

Fork the Winners!

At that stage we had figured out there would be financial beings living and dying in an environment governed by free market laws. The biggest piece missing was the role of the community.

Yes, people would be the ones creating financial beings… but why?

Why would they do it?

What would be the crowd’s incentives to invest time and energy in creating financial beings?

What would this crowd look like?

In nature, every living creature competes for mating and food above all other things.

In our model, competition for food would be analogous to competition for resources in the markets. The best financial beings would have access to abundant resources, which would help them grow and become dominant, as people in the crowd would be willing to invest more time improving winning specimens.

However, the crucial feature of evolution by which only the dominant specimens get to mate was still not present in our model.

We needed a way to establish which were the best algos.

We needed a mechanism for successful algos to reproduce and pass on their genes to their descendants.

We needed the crowd to breed the best algos!

And most importantly, we needed the crowd to want to do it!

That is how we came up with the idea of introducing algorithmic trading competitions in the model.

Competitions have multiple built-in incentives that resonate with different kinds of personalities. Because they cater to a multitude of profiles, competitions attract relatively more people than other activities, while promoting diversity at the same time.

We compete for the ludic and entertainment value of competitions…

For the challenge…

For fame and glory, honor and pride…

To measure our skills…

For the sense of belonging to the community…

Or simply, to win!

Our drive to compete is programmed in our DNA. Especially when it’s fun!

Competitions would be the default mechanism both to establish which are the best financial beings and also to engage the community.

In our minds, we had identified the winners and the incentives for the community to want to breed bots and compete.

The next element in the model fell on our heads like a ripe fruit from the tree of imagination:

Open source the winners’ code and… fork the winners! Improve them and compete again! Boom!

That is how the best bots pass on their genes to new generations! We got it!

That is how the wisdom of the crowd is put to work in the most efficient possible way!

That is how knowledge and new discoveries are spread instantaneously so that no one in the community is left behind!

That is how the playing field is leveled after each competition, keeping the whole community engaged and with fair chances of doing well in the next competition!

But most importantly, that is how we beat nature and speed up evolution!

We don’t need to wait for new traits to be tested through a biological lifetime spanning years and hope that the best features in each individual will catch on eventually.

We just hack it!

We increase competition frequencies to the most efficient rate and create all sorts of competitions, with all sorts of formats and rulesets so as to promote diversity and accelerate evolution throughout the trading space!

After discovering these solutions, we knew we had a real shot at rapid growth in intelligence, the main requirement for superalgos to emerge.

The ALGO Token

Eat or die!

Earlier in the article we mentioned food.

Anthropologists now speculate that it may have been access to a diet rich in fatty acids that propelled human intelligence to new heights when a small group of hunter-gatherers moved from central Africa to the shores of current-days South Africa to survive a dramatic drought that lasted forty thousand years during one of the ice ages.

Competing for food plays a key role in evolution.

In nature, individuals capable of granting themselves a constant supply of food become strong and dominant, are more likely to mate and thus are more likely to pass on their genes to new generations, transmitting important lessons learnt during their lifetime, meticulously stored in their DNA.

Our experience in the crypto industry quickly lead us to understand that autonomous bots that would have an income and expenses would require a medium of exchange.

The ALGO token would be the medium of exchange within the ecosystem and would also play the part of a scarce resource necessary for financial life sustainability.

Just like living creatures need food to live day by day, financial beings would need to earn ALGO to survive.

ALGO would be a capped crypto token (1 billion total supply), and its price would be discovered at crypto markets.

This feature would constitute yet another regulator of life in the financial biosphere.

If the population of financial beings increased too much, food would become scarce and expensive, condemning inefficient bots to die and allowing the best performers to survive and thrive.

On the other hand, if the population dwindled, food would become abundant, creating an incentive for the population to grow back again.

Algorithmic Trading World Championship

FIFA 2018 eWorld Cup Grand Final in London | Image © Getty Images

Increasing the frequency of competitions would increase the rate of birth of new generations of algorithms. This surely would have practical limits, since running a competition every minute wouldn’t allow time for the crowd to improve winners’ offsprings.

Another path to increasing the rate of birth of financial beings would be creating different types of competitions. After all, each individual trading bot would specialize in a single strategy, maybe even using one particular indicator.

There could be competitions solely for each type of strategy.

Current science in Collective Intelligence points that — many times — non-experts are able to make better predictions than experts.

So one of the ideas we put on the table was opening up the community to all sorts of people. Smart people with little knowledge about trading could contribute a lot to the collective intelligence of the supermind too.

Furthermore, asset holders looking to benefit from the trading intelligence of the community would want to have a role as well.

Maybe there would even be an audience following competitions!

Thinking along those lines a curious idea emerged…


In our model, competitions would be organized in a poker-inspired multi-layer league system. In our case, the tournament would be designed to drive evolution towards trading big capitals.

The crowd would self-organize in teams to create and train trading bots. This allows teams to incorporate all sorts of talent to the mix.

Teams would choose to participate in either minor or major leagues. This would help segment people with different interests and customize their experience accordingly. For instance, people that wish to compete for the fun of it would probably play in the minor leagues most of the time, with gamified themes and flows, while people with a professional orientation would mostly play in the higher leagues, with a user experience oriented to professional traders.

The bottom leagues would run short competitions, with minimal capital requirements and relatively small prizes. In this way, entry barriers would be minimal and competitions would be exciting to follow.

Higher leagues would have longer competition periods, higher capital requirements and offer significant prize money. Giving algos longer periods of time to prove themselves is a requirement to achieve reliability in performance.

Teams winning a league would get enough prize money to cover the capital requirement of the next league, affording mobility to teams who perform well.

Most traders would object the notion of open source trading bots under the assumption that a strategy that becomes popular loses its edge really quickly.


In fact, all strategies are born with a death certificate, as they all have a limited life expectancy. In our view, this may even be an advantage as it is a factor encouraging relentless evolution.

Moreover, with such level of diversification by which thousands of teams would be producing thousands of different ever-evolving strategies that compete under different schemes and rule-sets, we believe no single strategy will become dominant among the rest.

What we envision will happen is that higher order trading organisms will use single-strategy bots as individual tools in huge toolboxes, dynamically deciding what strategies to deploy given all sorts of variables ranging from market conditions, trends, fundamentals and every imaginable source of data available to the ecosystem.

Trading Intelligence Marketplace

Let’s imagine that the expenses each bot would need to pay for are related to execution time, storage and memory used every time they ran.

Now, what about their income? Where would income come from?

The trivial answer would be from competition prizes; then bots would literally compete for food.

A more sophisticated approach would involve additional income resulting from offering their services in a marketplace. Put in other words, bots would be available for others — both humans and other bots — to rent their services for a price paid in ALGO.

Consider the following scenario:

  • On one side of the counter, there would be trading bots with all sorts of specialties available for rent.
  • On the other side of the counter, there would be consumers willing to pay for bot’s services.

A market transaction is the most efficient type of interaction between any number of parties to discover the value of a product or service at any point in time. That is, market participants would decide which bots to feed and which to let die.

Does this sound like yet another natural selection mechanism? It certainly is!

This marketplace would act as a higher order intelligence capable of deciding which bots are worth keeping alive.

This is how the ever-growing community emerges:

The more intelligence the supermind acquires, the better performance its bots would display in the markets, and the more asset holders the ecosystem would attract. Therefore, the more people would add intelligence in the system by selecting the best bots!

In addition, the more asset holders willing to rent trading bots, the more value is injected in the economy as people bring money from outside the ecosystem to buy ALGO tokens to pay for bot’s services. Therefore, the more incentives there would be available to bot makers to keep making better bots and the more bot makers would be attracted to the supermind!

The marketplace is the last key element in our self-balancing evolutionary model.

The model may be a gross simplification and may require quite a bit of tuning, thus, like everything else within the project, the model itself is subject to the forces of evolution.

Time Frame

Started by two people in 2001 and competing with the likes of Microsoft Encarta or Encyclopædia Britannica, Wikipedia — maintained exclusively by volunteer editors — has dominated the encyclopedia market for years, spanning 302 languages and authoring almost 6 million articles in its english version only.

How is that possible?

Because the community features around 125 thousand active editors as off these days out of over 34 million editor accounts opened so far and several more million sporadic contributors without an account.

Wikipedia managed to attract a legion of editors, without any economic incentives.

Topcoder, a crowdsourcing company with an open global community of designers and developers participating in design and coding competitions, grew to 1 million community members over the span of 14 years.

At the Superalgos project we aim to attract 1 million developers, traders and data analysts during its first 10 years. The overall community including asset holders renting asset management services, and the audience following competitions should be at least an order of magnitude larger by that time.

We won’t venture irresponsible predictions as of when superalgos may emerge.

The truth is we do not know.

What we do know is that large companies producing trading algorithms usually count with specialized human resources in the order of hundreds.


As hinted earlier, in nature parents give life to an offspring.

A father and a mother.

We believe it is instrumental to make sure we are correctly mapping the masculine and the feminine principles into our system, in order to create an environment that can support the continuous rapid evolution.

Masculine and feminine principles in equilibrium are competition and dynamism balanced by nurturing and support.

Continuous growth should happen playfully, in a supportive atmosphere.

Picture the scenario of a newcomer entering the community for the first time…

What this person will find is an open community offering free access to its collective intellectual property, by generously offering her to fork any of the bots. The community supports her by educating her — for free — and by welcoming her as one of its own.

The community then encourages the newcomer to participate and give back, adding her own knowledge to the collective wisdom, by improving the bot, putting it to compete, and setting it free for it to live a financial life.

Now the newcomer’s creation will attempt to provide for her by winning in any of the myriad of competitions. If her bot does well, it may get forked and its seed may be passed on, along with her wisdom.

The community will help her iterate her creation into a marketable bot and will provide a marketplace for it to sell its services. The bot will become a worthy inhabitant of the financial biosphere, and in appreciation, will pay back her creator and everyone that helped it evolve into a successful individual.

It is this mix of dynamic, focused effort, together with a nourishing and supportive environment that we see as an abundant breeding ground for superalgos.


By now, you have probably asked yourself a rather important question: why?

Why are we doing this?

Or, put in sceptical terms…

Why is this project important?

Why is it worthy?

We believe the project has the potential to disrupt financial markets, in particular the segment of investable funds.

Our project will democratize access to the best trading technology and, at the same time, dramatically lower entry barriers for investors and bot makers so that everyone can participate and extract value from the world’s financial system.

Because we are starting with crypto markets, our project will become a main driver for crypto adoption in the mid term: as the quality of asset management services offered increases, more and more people will be seduced to put part of their investable assets in crypto in order to be able to participate and benefit from the ecosystem in the early stages.

Later on, as we move into traditional financial markets and by the time our model is capable of producing superalgos that outperform every other trading entity in the market, investment funds, investment banks and other investment managers may eventually abandon their silos and start managing their funds leveraging our technology.

In addition, our ecosystem will offer a wealth of opportunity for all sorts of people from all sorts of backgrounds to make a living and enjoy fulfilling jobs working as a part of a massive, borderless, open global supermind.

Finally, we should all consider the not so long-term dangers inherent to a technology singularity…

Our project is based on a system in which humans come first and are always in control.

A system by which machines work for the benefit of humans.

All humans.

Current State

The first line of code dates back to August 2017.

Our MVP is now up and running in alpha stage.

Algorithmic trading competitions start in February 2019!

Join the supermind!

Game on!

Part Two in this two-piece introduction dives under the hood of the Superalgos project and explains how we are Building a Global Trading Supermind.

A bit about me: I am an entrepreneur who started his career long time ago designing and building banking systems. After developing many interesting ideas through the years, I started Superalgos in 2017. Finally, the project of a lifetime.

Follow Superalgos on Twitter or Facebook; or visit us on Telegram or at our web site.

More by Luis Fernando Molina

Topics of interest

More Related Stories