FST 2017 Retrospective
What we expected and what we achieved
By last December, with the triumph of John John Florence we finished our second year predicting Fantasy Surfing games. A lot has happened and we reached excellent results. Some already expected and others very surprising.
2016, our first year, was all about construction. We begun from zero with only one page. A detailed explanation of that year can be read here.
At 2017 the focus was on growing, enhancing functionalities and achieving goals. We got it right for some goals and failed on others. We expected to do well at the end of the year in the WSL Fantasy Rank, that clue was based on our previous results and simulation of the 2016 season. Beyond Fantasy results we wanted to deliver entertainment to the world of surfing. With the database we built, we could present to our audience more information to follow the WSL by a perspective of numbers.
We failed in delivering a 2.0 version of our model for the 2017 season, however we expect to have it ready for this season. However we also achieved several surprising results. Of course, in the middle of the journey we saw opportunities to build new functionalities. It was also a great year for building new stuff
Our path to deliver much more than Fantasy predictions is being built. We want to bring attention to the World Surfing League by a perspective of numbers.
The story of 2017 is better told with a retrospective of our tweets.
March/April — 2017
At the middle of the journey we saw a possibility of launching a new Model, with a very different strategy from our official one. It would be a possibility to have another version fighting in our internal championship. That was the born of v1.5.1. Our main three versions were up to the battle.
Looking at the World Surf League by a perspective of numbers opened our eyes. If you somehow believe that: John John Florence assuming the leadership at the European leg and straight after winning the championship was a surprise; know that you're wrong. The Hawaiian has been progressing his numbers. Throughout the whole season of 2017 he was better, in numbers, than 2016. It seemed strange, that he wasn't at 1º before, even delivering better numbers. His consistency guaranteed his world title and that was no surprise for us.
Impressionante!!! @wsl @FantasyWSL
May — 2017
At the year of 2017 we wanted to begin predicting results to Fantasy Surfer from Surfer Magazine. The oldest game of Fantasy for surfing. We organised our database, updated some records and left everything ready for the task. There was, however, one information missing and we had to search it by our own. The price of the surfer on every event. We started without that variable and had been updating throughout the year.
The growth of our readers was impressive. We expected to grow, for sure, but no one could predict the way it happened. The relation with our readers is being solidly built.
There was still one problem when we looked at our data. We had more than 4k surfers on our records. How could we know and highlight great statistical performances from surfers in a QS or Juniors? How can we know who are those surfers, beyond the oficial rank, that are doing great? We launched our Stat Rank based only on the performance of the surfers throughout the season. A surfer stays in first if he/she achieves the best results of Average Heat Score and Winning Percentage of the season. John John Florence was an impressive case, in practically all year he had always had the best results of Average Heat Score and Winning Percentage, reaching the maximum sum possible: 100 points.
Our prediction models were doing well. And we knew it would achieve a great position by the end of the season. It is difficult for a prediction model to do extremely bad or good, normally they do average. That is a recipe for working in the long-term. At 2017, atypically, we reached a very bad result at Fiji and an exceptional one at France.
June — 2017
We also launched the pages for seeing event results. Now, for any event from WSL it is possible to know the best performances.
Fiji, for all our models, was a fiasco. Since now, while we are testing new approaches for enhancing versions, we test how it predicts Fiji 2017. That is a stress test for us.
July — 2017
Putting aside our results on Fantasy. Our efforts in other statistical entertainment were making progresses. It was possible to see the evolution of surfers throughout the season. It became evident the way John John surfed a step beyond other surfers.
August/September — 2017
At the event of Tahiti we got an amazing result. It was, somehow, expected, by the predictions we made based on past seasons. Tahiti has always shown similar results, only few surfers do well through different seasons. Reaching the position of 563º was an impressive result at the time. And we didn't know that we would do better. All our bets were placed on Tahiti to reach a high position at a single event. For us, would be impossible to do better on other event.
Having the statistical scores of all surfers and presenting progression graphs brought us a lot of advance. Not only the public gained a great entertainment and data to interpret. We also progress a lot with those analysis, having more insights for future enhancements.
October — 2017
The European leg was the best moment of 2017. The championship shuffled. Before it Jordy Smith was in first, his worst result at the season so far had been a 13th, he still didn't had another bad result at the season to throwaway, a fortunate scenario. However, JJF was leading our Stat Rank and surfing flawlessly throughout the season, by any mistake of his opponents he would reach the leadership. What no one would predict was that Medina would win France and Portugal, a circumstance that made him reach the 2º of our Stat Rank. However, JJF was still at first with several points above.
At the European leg we tried another approach to raise the engagement of our readers, specially after an event has started. We were aggressive and launched a page to follow the event live, with updates of stats at real time, projections and simulations taking into account the scores pre and after the event had begun. Beyond that, it was possible to see the WSL rank updated during the competition, already with throwaways applied.
And it was also at the European Leg that the surprising happened. Unbelievably we reached first place, with one of our models. No other Fantasy player could do better picks than our projections. The model played better than 40k active players at the time. The fact became history on Fantasy for Surfing, until the moment no other model could do a similar result.
December — 2017
Pipe finished the season and our second year for predictions. We are extremely happy by the results we reached. Now we have a solid background to build better models and functionalities for our readers. We want to grow the entertainment around surfing stats, scores and results from competitions. The fight to deliver a better model is higher this year and we will do our best to enhance.
By the end of the season our best model reached the position of 1469º on the overall rank with 5611 points. An accuracy rate of 68,5%. The champion of the season, Jase Jase Florence, ended up the year with 6162 points and an accuracy of 75%. Our goal at the 2017 season was to be around the 1kº. Some mistakes during the season, however, lowered the potential scores of our models. If we had picked the team correctly since the beginning, our main version would have made 139 more points, reaching the 428º with an accuracy of 70%. Enhancing those scores won't be easy, we understand the difficulties and we will do our job to evolve our calculations and the experience of our audience during the events.
We are ready for 2018!
