Sailboat racing is all about choices. Go offshore or hug the coast? Change the jib now or wait till the turning mark? Do we spend the budget on new sails or a coach? Every choice is a risk. Get it right and you gain; get it wrong and you lose. But this is sailing, and there is an existential element to the sport that exists in few others; sometimes the choices are more serious, and coming last in the race is the least of your problems. Just such a decision came up on the third leg of the 2017-18 Volvo Ocean Race from Cape Town, South Africa, to Melbourne, Australia.
It was December 12, 2017, and the whole fleet was sailing east, tracking the movement of a low-pressure system, when another, more ominous feature appeared in the forecast. The charts showed a new storm growing with spectacular speed and violence. By the 13th, it would be a fully formed monster with winds up to 65 knots at its core, approaching the fleet from the northwest. If they held to the optimal course, it would chase them down and run them over. Or they could jibe and take a slower, more northerly route that would avoid the worst of the weather.
There were seven boats in the fleet. Two took the northern route: Turn the Tide on Plastic skippered by Dee Caffari, and Team Sun Hung Kai/Scallywag led by David Witt. In both cases, the boats had relatively inexperienced crews (compared to the rest of the fleet) and not much training together as a group before the start of the race. In both cases, the decision to go north felt like the correct one. It was unlikely they would do better than sixth or seventh even if they had chosen to stay south. They could get that result by going north with much less risk of damage, and they might pick up a place if any of the boats that stayed south suffered damage in the storm.
The opposite was true for the top three boats on the overall scoreboard: Dongfeng Race Team, MAPFRE and Vestas 11th Hour Racing. They had previously taken all three podium places in the first two legs, and all three remained south. Once the storm passed, MAPFRE and Dongfeng Race Team led the fleet by over 50 miles. Vestas 11th Hour Racing was solidly in third place with a lead of more than 100 miles from the next boat. In all three cases, it looked like the right decision to stay south, and going north would have put them out of contention for the overall lead. They also had good reason to believe they could manage the conditions in the south and avoid a bad outcome.
This left two boats: Team Brunel and Team AkzoNobel. Neither crew lacked experience. Skippering Team Brunel was Bouwe Bekking, who was on his eighth attempt at winning the race. And while Team AkzoNobel’s skipper was a first-timer in the role, Simeon Tienpont had won a couple of America’s Cups and been around the world twice before. Tienpont’s crew also included Jules Salter, on his fourth race around the planet, already counting a second place and a win. Chris Nicholson, a six-time world champion, two-time Olympian for Australia, and a second-place finisher as skipper among his six previous races, was also on board.
The crews of Team AkzoNobel and Team Brunel nevertheless faced a more difficult choice than the others. On the one hand, they had every right to regard themselves as peers of those on Dongfeng Race Team and MAPFRE. On the other hand, Team Brunel’s crew had sailed together a lot less than those two boats, while Team AkzoNobel had management issues that had disrupted their preparation, which resulted in several changes to senior crew in the early legs. “You can’t instantly replace those people…” Salter said when I talked to him. “If you look at this race again and again, no one has thrown the boat together six months before the start and gone out and done really well.”
Despite this inauspicious background, Team AkzoNobel and Team Brunel both decided to stay south on the theoretically optimal course. By December 14, AkzoNobel was coming up fast on the invisible limit of the Ice Exclusion Zone (a line drawn by race officials to keep them out of known areas of icebergs). They were going to have to jibe in 35 to 40 knots of windspeed. Nicholson was steering, and with all nine of the crew in position and on deck, he went for it. “We had a bad one,” he said in a video to the race office afterward. “I thought we had a good wave to go down. I should have pulled out of doing it, and we probably would have broached, but it wouldn’t have been the outcome that we had…”
They broke battens and pulled the mast track away from the spar.
It could have been a lot worse, but it could also have been a lot better. “On another day, we might have just got away with it; we might only have broken a little bit of track and one batten, and thought, ‘Oh, that was a bit on the edge and off we go.’ I’ve done that before many times,” Salter said. Instead, a crippled Team AkzoNobel limped into Melbourne in last place, well over three days after the leader, and immediately started the next race—against time to get the boat repaired and ready before the start of the next leg.
Team Brunel fared much better than Team AkzoNobel—they kept it together through the storm and finished in fourth place, less than two hours behind Vestas 11th Hour Racing and just over a day ahead of Team Sun Hung Kai/Scallywag. Both Team Brunel and Team AkzoNobel chose the fastest, southerly route; one came fourth, the other seventh and last, but—and here’s the controversial part—there’s an argument to be made that both boats made the wrong decision and took an unnecessary risk. It worked out for Team Brunel, but that still doesn’t make it the right decision, and here’s why: The crews did not fully set what might be gained by their choice against what could be lost. In both cases, the potential losses of going south outweighed the potential gains.
The upside of staying south was limited.
Could either crew realistically hope to beat the better drilled and prepared Dongfeng Race Team, MAPFRE and Vestas 11th Hour Racing in those conditions? Maybe later in the race, yes. But on this first excursion into the Southern Ocean, it was unlikely they would suddenly make the jump to the podium that had eluded them in the first two legs.
So, if there was little to no chance of a third place or better, why not go north with Turn the Tide and Scallywag? They could expect to beat both of them on that same northern route, which meant they would finish fifth at worst, with every chance of moving higher should any of the leaders have a problem or break down in the storm. And if both Team Brunel and Team AkzoNobel had chosen to go north, they would have been racing for exactly the same fourth place that they were effectively contesting in the south, but in much milder conditions with much less risk of damage or worse.
When I put this to Salter, he wasn’t convinced—or maybe he was. “There is a balance—I think that showed in the next Southern Ocean leg, where we went a very similar way again, and we got to a point where we thought we’re just going to back off a bit here because we know what damage can do. We did, and we dropped back a bit, not a big distance, but 50 or 100 miles behind the leading group. And on the approach to Cape Horn, we gradually caught them all up again. MAPFRE broke the mast track and the mainsail, and Vestas lost the rig soon after that as well. So, you get yourself up to third by doing that. It’s just picking the right time to play that strategy—and offshore sailing is so much about that.”
These are tough calls, and often they are made on gut feel—or after a long discussion that the most articulate and opinionated can win on the strength of their hand-waving skills alone. There is a better way, a way to make the decision more analytical using tools based on formal decision-making systems. The “theory of expected value” assesses the value of all the possible outcomes for each decision and multiplies each of them by the likelihood (the probability) of those outcomes. The results of each of these multiplications [value (V) x probability (P)] are then added together to give the expected value (E) for that decision: E = (V1 x P1) + (V2 x P2) and so on.
Once all the expected values have been calculated, then the decision with the highest expected value is the one to go for. It sounds complicated, but an example will make it clearer. Let’s look at the strategic choices for Team AkzoNobel and Team Brunel as the storm approached. The value for the outcome (V) will be the points available for each position in the leg. These were scored in reverse, and this leg was worth double points, so the winner got 14 points, second place 12 points, and so on down to last place, which was worth two points.
We’re going to make a few assumptions to keep the analysis relatively simple. So, we’ll assume the rest of the fleet made the same choices they did in the race and finished the leg successfully. Two boats went north, both of whom Team Brunel and Team AkzoNobel will beat if they don’t suffer damage, while three boats went south, none of whom they will beat, damage or no damage.
On the southern route, the pairs best possible result was fourth, and that would have been worth eight points. A second potential outcome was seventh and two points, if they suffered sufficient damage to stop them racing competitively but not stop racing altogether. The final possibility was that they suffered more serious damage and were unable to finish the leg, and unable to start the next one. I’m going to give this outcome a score of minus four points, equating to zero points for Leg 3 and a loss of a potential fourth-place finish in the following leg. (The next leg was not double points, so fourth scored four points.)
I’m assessing the probability of them racing hard all the way to the finish (and scoring fourth) as 60 percent, suffering moderate damage and coming last as 30 percent, and suffering more serious damage as 10 percent. So, now we can prepare a table:
Points Scored (V)
|Fourth||60% or 0.6||8|
|Seventh||30% or 0.3||2|
|Not finishing||10% or 0.1||-4|
We could build a much more complex table listing more outcomes and their probabilities, but I think the approach is clear with just these three options. So, the expected value (E) is the sum of the gains from each of these outcomes:
E = (0.6 x 8) + (0.3 x 2) + (0.1 x -4) = 5
On the northern route, the best possible score was fifth, and that would have netted them six points, with the outcome for moderate damage again seventh and two points. The outcome for serious damage was also minus four points.
While it wasn’t guaranteed that they would successfully complete the leg on the northern route, the chances were a lot higher, and I’m going to call it 90 percent, with a 9 percent chance of moderate damage and a 1 percent chance of serious damage. So, the table for the northern route looks as such:
|Outcome||Probability (P)||Points Scored (V)|
|Fifth||90% or 0.9||6|
|Seventh||9% or 0.09||2|
|Not finishing||1% or 0.01||-4|
And the expected value will be:
E = (0.9 x 6) + (0.09 x 2) + (0.01 x -4) = 5.54
This analysis produces an expected value on the northern route of 5.54 points compared to 5 points on the southern route. So, any decision based on expected value would be in favor of the northern route, supporting the earlier conclusion.
Now, maybe others would score the probabilities of damage differently—and it’s worthwhile to play around with the numbers to see how the outcome changes—but there’s no doubt that the analysis helps to clarify the choices. We’ve swapped hand-waving for assigning probabilities, and that has to improve any discussion and any decision made.
An expected-value analysis can help in other areas of our complex sport. One of the big problems facing any race-boat campaign, be it an America’s Cup team or a small keelboat shooting for top 10 at the national championship, is there are more ways to improve performance than time or money to pursue them. It’s always hard to make decisions about where to put limited resources when the outcomes are uncertain, and expected value provides an analytical approach, calculating an expected cost for each unit of performance gain.
The process starts by creating a list of the performance projects that are under consideration. For example, practicing for a weekend, buying new sails, or hiring a coach for the season. A simple version of the calculation would be to assess the potential speed improvement (this would be V) for three cases: the best outcome, worst outcome and most likely outcome for each option being considered. We would then assign a probability (P) to each of these possible outcomes and use these three pairs of numbers [speed improvement (V) and probability (P)] to calculate the expected value just as we did for the route choice ahead of the storm.
In this case, we can go a step further: By dividing the monetary cost of each option by its expected value, we can generate a cost-effectiveness ratio. This allows us to tackle the issue of balancing resources in a parallel way to balancing risks—and expected value can be just as useful in revealing the trade-offs. There are limitations to the approach; for instance, there is no allowance for how these gains will degrade with time. Nevertheless, it’s still a great way of tackling any problem where resources are being assigned in uncertain circumstances—just one of the many tough calls in sailing that the concept of expected value can help to illuminate.