PHILADELPHIA — Soccer is perhaps at its most exciting when its biggest stars — think Neymar, Messi and Luis Suárez — are dribbling through the defense and firing shots into the net from any conceivable angle and position. What most fans likely don’t notice is how those players are set up for opportunities by the rest of their team.
Will Spearman wants to change the perception that only the player with the ball matters. At OptaPro’s first North American event last week, Spearman, a data scientist at Hudl with a background in physics, presented his preliminary research on Off-Ball Scoring Opportunities — what he defined as scoring chances a player, not in possession of the ball, creates by getting into open space in “dangerous” areas.
The model is based almost entirely on a player’s distance from the target goal, which determines their probability of scoring. Of course a number of factors influence that model, such as how well a team finishes on opportunities and a player or team’s aggressiveness and decision-making (i.e. did a player with the ball pass to the player with the better statistical scoring opportunity?).
“So the original idea was this space, this model of controlling space and quantifying the way players control the space. And so then, trying to relate that to scoring is where this came from,” Spearman said in an interview at the event. “I was trying to understand if you know how you control space and you know where you can score goals, and we can start looking and seeing how players’ positioning allows them to create the opportunity to score.”
Spearman acknowledged that his model has a ways to go before it is polished, and he took some feedback from the audience on what other factors the model could account for. The model only considers distance from the goal in terms of “danger” and off-ball scoring opportunities.
“I think there’s a number of issues with the model that could really benefit from some additional research, especially the scoring model,” Spearman said. “So right now it’s just based on distance to goal; I think by looking at not just distance to goal, but number of players that could potentially get into position to block a shot, positioning of the keeper, a lot of things that would go into whether a shot from a given location could actually score, (that) would really improve the model.”
But how could the model enhance broadcasts of soccer and what fans see at the stadium? Spearman said he applies a very visual approach to his research (his slides at the event support that statement). With his research, Off-Ball Scoring Opportunities could easily be represented on a video board and on TV; broadcasters could use a telestrator to highlight where a certain player’s movement helped create a scoring chance, or a player who should have received a pass because he was in a dangerous area.
Visualizing the model for fans could also be a natural extension of Hudl’s platform, which already provides video and performance analysis tools across a range of sports.
“One thing I would like to do when building these sorts of models is making them very visual…And so I think that for me, when modeling, that helps me think it through, but I also think for fans you could have a very intuitive representation of, ‘Oh OK, he’s controlling this space,’ or ‘Oh this is this danger zone,’ and…because it’s very visual, I think it’s something that’s a lot more accessible to fans than some more black-box approaches.”
But Spearman’s isn’t the only data model that could be used to give fans a deeper look at the game. OptaPro‘s Johannes Harkins developed a mechanism that he said he hopes gives context to the events that unfold within a soccer match. The model accounts for a team’s playing style and how that style changes based on the match conditions (e.g. winning or losing, specific players on the pitch, etc.).
By looking at sequences, which Harkins defined as “uninterrupted passages of play,” and possessions, Harkins can analyze a team’s style by how they string together possessions into sequences. Pointing to the New York Red Bulls as an example, he explained that a team with a short sequence time often has greater direct speed (a measure of progress on field divided by sequence time). Other factors, like number of passes and distance to goal, can affect a team’s style as well.
Harkins told SportTechie at the event that OptaPro has taken a model that has existed for some time and polished it, and is ready to put it into feeds that will be available to clients for the upcoming European soccer season.
“I think part of the gap between analytics and tactics, maybe, in the existing landscape has to do with people not necessarily feeling like they can apply their tactical expertise to analytic concepts, and I hope that this bridges that gap as the expertise of people in analytics and the expertise of people in tactics and performance analysis meet in the middle and do a lot of really good work,” he said of the model.
Harkins’ model, he thinks, could be delivered onto stadium screens to show fans at a deeper level how teams make tactical decisions. Though he cautions some of what his model explains might already be intuitive to fans, there are surface-level stats on teams or players that could be gleaned for information on team style.
“And from a technological perspective, I think it’s interesting to consider things like video platforms that are used for scouting or match preparation by teams, or even broadcast, people who try to cue up relevant video to provide an experience for a fan.
“You could use this model to characterize moves during a match and then relate them in terms of similarity in a technological or mathematical way, and then you could say ‘I have this aspect of play that I’m looking to highlight, give me the five possessions by Team A that have that characteristic; even show me similar possessions by Team B to this possession by Team A,’ and those are things you can bake into it with some technical expertise.”
Together, these two models could well work to liven up broadcasts and bring fans deeper into the game.
