I had the chance to have a chat with Ed Feng of the Power Rank and talk soccer analytics and probabilities in sports with a focus on the beautiful game. It was interesting to play with Ed's model of his Euro 2012 '' prediction tree '' that you can access here as he updates on a regular basis to reflect the results as they happen and see , in a beautiful interactive manner , your team's chances go down the drain, maintain themselves or reach for the sky and glory.
Numbers are nothing without making them look good and attractive. Before the boom of the what we call '' Web 2.0 '' , using the Web as a tool to show that data, organize and make it look good just transforms that data into a living organism by itself. I cannot stress enough to take a look at the Euro 2012 model (click here) and look at the Quarter Final matchups at the Euro and you can see the probabilities of each match-up and chances of the remaining teams at the remaining stages of the competition.
We talk FIFA Rankings vs Power Rank World Rankings, fancy algorithms and Ed Feng, the fan.
More after the jump
1. Sofiane: I had the chance to discover about yourself and your Power Rank through the excellent Amy K Nelson and her SBNation report as she visited you during the NCAA basketball tournament.
a. To introduce you to readers and fans, why dont tell us a bit about yourself and your background?
Ed:
Well, my name is Ed, and I'm a recovering academic. I spent a decade doing research in statistical physics, using mathematics and computer simulations to understand polymers and carbon nanotubes. It was only when I discovered Google's PageRank algorithm that I realized that this background could be applied to other topics like websites. This inspired a new algorithm that applies to ranking sports teams. And since I've always been a huge sports fan, this pivot has been a great move for me.
b. Why don't you explain and vulgarise the basic concepts of the algorithm used to make your Power Rankings? It seems to be applicable to many different sports and competitions, does it give you an edge over the competition?
Ed:
If you only remember one thing about the algorithm, it should be that it accounts for margin of victory and strength of schedule in ranking teams. There's a big difference between winning 1-0 and winning 4-1. The latter more strongly supports a team's strength. As for strength of schedule, a 1-0 loss means different things depending on whether the opponent was Germany or San Marino. Our methods provide a good way for accounting for these differences.
2. In one of your blog posts on The Power Rank, you mention a big difference between your ranking of Croatia and the FIFA Rankings. FIFA Rankings explains its ''fancy'' procedure for its rankings here.
P(oints) = M(atches) x I(importance of Match) x T(Strenght of Opposing Team) x C (Strength of Confederation)
I personally never liked FIFA Rankings as I don't really get an absolute value and ranking of a team and as I consult it , I will more use it as a general idea of the best team and the worst teams. After looking at FIFA Rankings procedures, do you feel that you should fly to Zurich, Switzerland and give them a lesson in statistical mathematics? and maybe even apply your model to their rankings?
Ed:
Sure, I'd be happy sit down with Sepp Blatter and educate him a bit. But the FIFA rankings aren't going away anytime soon. And they will still provide a flawed system for seeding World Cup qualifying. The Power Rank is for fans to have a better understanding of the game than those who govern it. We empower fans with numbers and the stories they reveal about sports. For example, we gave Denmark a 37% chance to make it out of the Group of Death. I think most people would have found that number absurd before the start of Euro 2012. But the Danes put up a great fight, and they might have made it through had the referee called the blatant hold in the box against Nicholas Bendtner in the Germany game.
3. Data Models and Visualisations have provided another way to look at , analyse and present data in different forms and shapes. Your Euro 2012 `` Bracket `` is very interactive as it shows different paths and the associated probability (a bit like a decision tree) . Was it important to present your predictions with nice visuals, all in one screen in an interactive manner?
Ed:
Crunching numbers is worthless unless you can convey this information to a sports fan. My friend Angi Chauhelped me with these interactive visuals, which are a convenient way to access a large set of numbers. I hate looking through a table of numbers. It's incomprehensible to all but the most intense data driven fan. For everyone else, Angi's designs let them access the win probabilities they want.
Angi is really fantastic. I originally thought that the Euro bracket would look similar to the "wheel" we used for March Madness. But, on limited time, she decided it wouldn't look right with only 16 teams. So she developed a new design that cleanly captures the group to knock out stage structure. And then she codes it up. I couldn't ask for a better partner in crime.
4. You watch a lot of sports as a sports fans. Does it change your model and probabilities when you see Spain struggle against Croatia for example?
Ed:
Well, it doesn't change my models and probabilities... yet. We're always looking to improve the model. Watching the games gives new ideas about how to do this. Moreover, watching games is the only way to tell whether your numbers reflect reality. Spain is a great example. I thought they looked terrible against Italy, even disregarding the blunders of Fernando Torres. It wasn't the beautiful play we saw in South Africa. I thought they looked better against Croatia, although they couldn't score that many goals against a rather stout defense. Can we quantify the decline in Spain's play? Can we use data from the club level (Barcelona wasn't the same team this season as last)? We'll see.
5. Any favorite national teams? clubs that you enjoy watching and support?
Ed:
I certainly support the US national team. I like what Jurgen Klinsmann has done with the roster, playing skilled guys like Jose Torres and Hercules Gomez. I almost started a Facebook campaign for Bob Bradley to play Torres more, but now it's unnecessary.
As for clubs, I follow Real Salt Lake because their skilled possession style of play. It's a stark contrast with teams like Portland and San Jose. In Europe, I like Everton, since they always seem to pluck good players out of nowhere and use advanced analytics. Not that I expect them to win the Champion's League anytime soon, but they're a plucky underdog in a league of big spenders. Lyon is another team that fascinates me. I enjoyed the story told about them in the book Soccernomics.