Football science, applied AI, Snap’s Mordor-ish nightmares and more in the Programmable Edition #86
Every week, we try to select the best articles talking about applied machine learning to build autonomous cars, build personalized news feeds, make search more relevant, or recommend movies you may like. We talk a lot about computer vision, natural language processing, and speech recognition, because breakthroughs in those areas will have huge consequences.
This week, I want to focus on another shift happening in sports, or one I want to see happening, especially in (European) football. Indeed, we currently only use 5% of the data collected to describe and explain the result of the game. More specifically, as Billy Beane puts it, talking about traditional stats, we “only credit outcome, we don’t credit process”. We count the number of shoot, of passes, of goals. But we filter out everything else, like defensive press, numerical dominance, player attraction or movement intelligence to name of few.
This is where data science must step in. The goal is to analyse a greater number of events and come up with new metrics to describe the performance of a football player. Eventually, we should be able to “properly allocate credit or blame to a player” and gain a deeper and wider understanding of the game.
Meanwhile, we can also leverage existing hardware to gather new data. This is what I tried to do by exporting and processing information from a smartwatch that I wore during a game. I also represented my movements on the pitch every second. You can find the article about this project on this link.
Maybe one day, even at the amateur level, we will be able to assert scientifically — that is backed with data — who shone and who struggled.
Sportively,
Closing the online video vs. traditional TV price gap
Content quality and safety are some of the factors that justifies, from an advertiser’s POV, the current price gap between online video advertising (cheap) and traditional TV ad (expensive). Besides adapting to new video consumption usage, YouTube’s new service to come, YouTube TV (35$/mo, starting sometime this spring), aims at reducing this gap by interleaving its content with traditional TV programmes.
Buy now, leverage later
Mozilla, the company behind one of the leading browser Firefox, announced the acquisition of Pocket. The app — bought for an undisclosed sum — allows users to save and retrieve documents and videos from websites. The company said it will leverage Pocket talents to build its ambitious Context Graph project.
Against the power that has risen in the east, there is no victory
While Snapchat is thriving in Western countries, it faces great competition in Asian markets. In China, Korea or Japan, the mobile app Snow beats it by implementing a winning strategy : it proposes additional features fitted to local tastes and builds strong relationships with large information technology incumbents. Now, Snow wants to expand its territory to the west, becoming a potential competitor of Snapchat at home.
Building an AI-first social network
The article describes how AI at Facebook has gone “from being a fairly rare component in products to something now baked in from conception”. Indeed, the company made a deliberate effort to democratize AI among the employees and power every feature by applied machine learning.
Like what you just read? Hit the recommend button so others can find it too and wait for the Programmable Edition of next week.
In the meanwhile you can follow us on Twitter.
