Man vs. Machine at the America’s Cup
Have you ever sailed a sailboat? It’s perfect to clear the mind — the sort of activity you do when you’re on vacation, or you want to unwind on the weekend. It might cost you $50 at your local sailing center to charter a boat for an hour or two, and off you go, away from the madness of your data-driven existence.
Today, the 2017 America’s Cup concluded in Bermuda with a resounding victory for the kiwi team, Emirates Team New Zealand, over the defender from the past two editions, Oracle Team USA. For those of you who aren’t too familiar with it, the America’s Cup (AC) is run every four years, and it’s the oldest international trophy in all of sports. It dates back to 1851, and the trophy itself is affectionately called The Auld Mug. To most sailors, it’s the pinnacle of sailboat racing.
If you Google America’s Cup, you’ll quickly see what incredible machines today’s America’s Cup boats are. I’m not going to go into the physics of those machines here, because that would take an entire dissertation, and I’m a bit rusty on hydraulic accumulators, foil lift, rake control and wing aerodynamics. If you’re interested, Sailing World’s editor Dave Reed did a really good job of reviewing the basics recently. But I’m going to talk about data — and how data science has become an absolutely essential piece of racing success in the America’s Cup today.
AC boats are flying machines that need to be set up for optimal performance on any given day. If the wind is light (say, 10 knots, like it was for most of the finals in Bermuda) and the water is relatively flat, the boat will need to be set up in a different mode than if it’s blowing a hoolie and the tide is trying to rip you out to sea. Not unlike any other sailboat, for that matter, but these boats take it to a whole new level: an AC boat today is typically fitted with 1,000 IoT sensors, measuring everything from weather conditions (wind strength and direction, for instance) to boat position (GPS, speed, drag and angle to the wind) to forces on structural components of the boat (the wing, hulls, foils, rudders). On the boat from Oracle Team USA, the wing alone has 300 sensors, and those sensors might be capturing data at a rate of hundreds of times a second.
Over the course of the four years leading up to the Cup this summer, every day on the water generated hundreds of gigabytes of sensor data for each team. Most of that data was processed in real-time to help operate the boat (to monitor hydraulic pressure, rudder position, wind shifts, etc.), but analysts also pored over the data at the end of each run to understand what happened, correct course and optimize all sorts of parameters for the next time the crew was out on the water.
And the boat was just one part of the equation. How about the men sailing the boat? And the environment? Data science played a key role there as well.
For most of their four-year campaigns, sailors were fitted with biometric sensors to measure their fitness and psychological state and help optimize their conditioning. That’s the case with all top-level athletes these days, of course, but before the Cup adopted those fast-flying boats a few years ago, that level of preparation was unheard of in sailboat racing. To further complicate matters, each position on the boat is different, and thus requires a different athletic profile. The challenge for the preparation teams was to develop athletes who could perform at the top of their craft, under intense pressure, doing wildly different things but still operating as one unit. There’s no equivalent in the world of sports. Imagine a contest between two competing kitchens, with their chefs and sous-chefs, trying to prepare a meal in a pair of food trucks flying through city traffic. With knives dangling off the wall and a guy shoveling coal into the burning stove. I suspect there’s an equivalent in the world of special force commandos, but I prefer the image of warring food trucks.
For the environment, the teams weren’t just checking the Weather Channel the morning of the races. They developed their own models, taking into account years of local sensor and cross-validation data to understand what a particular mix of conditions (water and air temperature, barometric pressure, tide, ocean swell, water depth around the course, etc.) might do to wind and current gradients over the course of the day. They also lived and trained in Bermuda for months at a time to learn about the local conditions first-hand.
Oh, and that’s without accounting for the soft espionage data that each team was constantly collecting about their competitors (how well their boat was going in certain conditions, how their skipper might react under pressure, etc.).
Data. data everywhere. In the process, each team built a massive reference database that they could rely on to polish their skills. The learning curve never flattened. Peter Burling, the 26-yr old winning skipper, kept repeating that his team “was still on the steep part of the learning curve” all the way to his very last interview. Reporters kept rolling their eyes, of course, after such a dominant performance from his crew, but the guy was totally genuine. There’s a lot left in the tank with these boats.
Oracle is no slouch in the area of big data analytics, but the company has from day one seen the Cup as an opportunity not just to promote its existing capabilities, but to continue learning and push the boundaries. It even partnered with Airbus to fine-tune its wing control system. Another competitor, Team BAR (led by sailing legend Ben Ainslie) teamed up with Land Rover to adapt machine learning capabilities that Land Rover had been developing for autonomous driving. At the speed those boats are traveling, and with races over in twenty minutes, AC skippers need to make quick decisions on the water, and there’s no way they could keep their boats afloat and keep a clear head without data science. That sounds like a sales pitch in the corporate world, doesn’t it? And why not. These races were an incredible showcase for big data analytics, and the tech companies that supported the America’s Cup are now taking every opportunity to tout their battle-tested capabilities.
This is all great for data science, but that doesn’t mean that the sport has been taken over by machines quite yet. Yes, at times, the scene in Bermuda felt a bit like a high-school robotics competition (albeit one where the entry fee is $3 million, the bots/boats cost $10 million and the school invested $100 million to make sure your team got there in time). But on the penultimate day of the 2017 Cup, Jimmy Spithill, the veteran skipper for Oracle Team USA, made uncharacteristic errors — mistiming his starts, dropping off the foils, even going over the boundaries of the course as he was trying to catch his opponent.
At the press conference immediately after racing, Abby Wilson, a reporter from Television New Zealand, asked him what happened:
“Jimmy, start line penalties, miscalculating the line, a boundary penalty today — you’re sailing for a technology company, but are you guys having technology issues?”
After a few good-natured chuckles from everyone in the room, here’s what Spithill said in his reply:
“At the end of the day, you have to make a judgment decision yourself, you know. All teams rely on all sorts of algorithms and systems, but at the end of the day, a human has to decide when to push a button or when to make a move. If you’re asking for why the mistakes happen, or for someone to point the finger at, then you should point it firmly at me.”
As it turns out, there’s still plenty of room for human error, thank goodness, and sailing isn’t man vs. machine quite yet. You can still charter that sailboat at your local beach and sail into the sunset without an ounce of computer technology on-board.
But wear a life vest.
Thanks for reading, it means the world! Check out Data Science in the Age of Trump for a perspective on the use of data science in an entirely different context!
