Breaking down Zone7, Part II: Algorithms & rules vs. auto-learning

Algorithms will change the lives of everyone on the planet — and they already have. But what are they? Founding a company that creates algorithms for a living, it's a question Tal Brown is used to answering.

Algorithms will change the lives of everyone on the planet — and they already have. But what are they? As a company that creates algorithms for a living, Zone7 understands this is a topic that comes up often when explaining what Zone7 does.

Algorithms are simply a set of rules used to solve a problem — especially, by a computer, but not always. When we bake bread or cook pasta, using a recipe, that too is an algorithm. If we did not have a huge variety of powerful algorithms solving problems for us, our lives would be something of a mess.

Using proven rules to tackle problems is a very old idea. The word algorithm derives from a 9th Century Muslim mathematician and astronomer named Abu Abdullah Muhammad Ibn Musa Al-Khwarizmi, widely regarded as the inventor of algebra, as well as the numbers we use to this day.

So a rule-based algorithm is like a recipe. A famous example we all know is the quadratic equation! Sure, this is a bit simplistic, but the purpose here is to illustrate that algorithms are sometimes as simple as a set of pre-defined rules applied to a new environment.

However, technology today relies on more sophisticated algorithms, and we have come a long way since the days of “simple” equations.

One example I often use is Google Maps: before writing this blog, I drove my daughter back from summer camp. Google Maps provided a route that was NOT based on predefined rules but rather on a dynamic analysis of EVERY route EVER driven in our neighborhood. Google also includes data and learning from weather, traffic and many other parameters.

Generally speaking — computer scientists, please bare with me — we call this “Artificial Intelligence”, to mean an algorithm that learns from a data, a lot of data, becoming better the more data it learns from. In Zone7’s world, old-school algorithms are like the quadratic equation — rules for how much running, jumping, sprinting are appropriate. What we have done is to create a Google-like cloud that creates A LOT OF dynamic rules by learning from every soccer/baseball/hockey game and practice ever recorded in our ‘neighborhood’.

In Part 1 of this series, we discussed BIG DATA in sports performance. The bottom line is that today we have access to tons of data about what athletes are doing on/off the field and how their bodies respond to these workloads and efforts. For example, we have second-by-second reports of a soccer midfielder’s movements during a Premier League or La Liga game and training session. Or data about every single baseball pitch thrown in a game since 2014. We also have access to data about how athletes train, sleep and recover.

This “Big Data” allows us to general several kinds of algorithms and insights: A set of clusters showing how things in the data are related — e.g. what patterns lead to specific types of injury — or, a model detecting when players are at high risk of injury based on countless individual properties that change from day to day. Or, a series of optimization prescriptions, completely personalized for each individual athlete, helping drive them forward towards peak performance and away from risk of injury.

The key word here is PERSONALIZED. No two athletes are alike. Each athlete has a unique profile and physique so ‘one size fits all’ cannot work in this field. Zone7’s algorithms are all personalized, meaning the individuality of each athlete, and the uniqueness of their environment/coach/team, must be reflected.

This personalization is absolutely crucial. For decades, innovators have understood how to massively customize every single product, tailored to the individual. Going back to the previous example of Google — driving instructions offered to me are super tailored to my location, driving preferences and tendencies. This is why data about my personal driving history are so valuable!

Many startups trumpet “AI” as a buzzword — but AI technology alone does not guarantee success. Results must be consistent and transparent. Zone7’s artificial intelligence algorithms have undergone extensive testing at the highest level of sport performance. Few of us would take medicine, or be allowed to buy it, without clinical trials. Zone7’s algorithms have undergone the close equivalent of meticulous clinical trials. We can show you the results and invite you to visit our website and case studies, showing Zone7’s injury detection rates in several sports environments.

  • Zone7’s injury detection rates in several sports environments

Results in the AI era are not about ‘fast clicks’ and ‘pretty buttons’. AI is only as good as how much trust users can put in it. In my next post in this series, I will discuss lessons learned at my previous job at Salesforce.com about what it takes for users to trust AI. I was fortunate to be one of the first product managers to work on predictive technology and helped define ways to evaluate results of an AI algorithm for quality, consistency and ‘trust worthiness’.