Do Advanced Analytics Have a Place in the Beautiful Game?
A relatively simple question. Considering this site has statistics and percentages everywhere, it seems to be fairly obvious which side of the argument we’d be on. But as time has gone on and we’ve read more and more about trying to mimic what is occurring in other major sports, it has become less clear that there is a good solution to this analytical problem. There are certain things unique to other sports that allow them to be conducive to analysis.
Other Sports
Baseball in particular has distinct measurables. In addition, the technology has expanded to such where everything related to every pitch can be captured and analyzed. How fast was the pitch thrown? Where did it start? Where did it end up? Did the batter swing? Did he make contact? What happened after contact? Which direction did it go? The possibilities are endless. Given the vast amount of data, it is far easier to slice and dice the analysis any way you want. It is easy to pinpoint which players have the underlying numbers to prove that they will be good players in the long-term.
Likewise, basketball analytics have grown considerably in recent years. There are formulas to show how efficient players are on offense and defense. There are formulas to adjust teams numbers on a per possession basis (“tempo-free”). Logically this makes sense. If a team wins by 20 points in 20 possessions, that’s far better than if a team wins by 20 points in 100 possessions. There are shot charts and cameras that track a player’s every movement. In other words, it’s now mainstream. American Football is the same way. Numbers that go beyond the numbers to figure out who’s really doing well.
The sport that compares most favorably to football from an analytical perspective is hockey. It’s a game of possession that ultimately results in a team scoring a goal. The scores are relatively low, and each of those goals is truly all that matters towards victory. But there is one major difference when it comes to analysis between hockey and football. During a hockey game, teams are constantly rotating players. They can change a line and bring players in and out as they please. It is fairly easy to see which combinations of players work well together since there are many combinations at work in a short period of time. The same cannot be said for football. 11 players are sent out, and only 3 changes are allowed. It is extremely difficult to determine the “best 11” when a team can’t rotate players in and out.
This exact reason is one of the ways basketball analytics have advanced so greatly. Teams only have 5 players on the floor at a time, and they are free to rotate and change them at will. Perhaps an individual who is skilled does well during a practice setting, but when it comes to the team game, they are not adept. It is easy to flush this out in a short period of time. It just can’t be done in football. Likewise, while we appreciate the “per possession” work people have done trying to analyze football, we don’t think those metrics hold much water.
What Can We Do?
Over recent years, many attempts have been made to transform methods from other sports and apply them to football. It’s not easy. Should things apply to specific players or teams? How do you cater the numbers to optimize a lineup? And most importantly, what numbers are even available to analyze? Perhaps the biggest problem when it comes to development of analytics is what is available. There are mostly heat maps and some individual possession statistics, but that’s really it beyond your generic, fouls, offsides, goals, assists, and cards.
The best thing outside of these items currently comes from individual work. Our friend Bill Vegas at EverybodySoccer.com has developed a metric called Goals Saved Above Replacement (GSAR). This metric is an algorithm for rating how well goalkeepers do relative to a replacement-level keeper. This is very similar to Wins Above Replacement (WAR) in baseball. The problem (awesome thing) about this metric is that Bill personally sat down and watched and graded every keeper’s performance throughout an entire game. Not only is this extremely time consuming, it is tough to stay accurate over countless late hours of pizza and film. But that’s not to say it’s not a valuable metric. There’s probably some real accuracy in what Bill is doing, and this should be the future of the development of analytics. It may have to be developed first before it is able to be calculated directly with computers. And that’s OK. It’s what makes the beautiful game so great.
Recently retired NFL player Rashard Mendenhall had this to say about the current state of the game: “games are analyzed and brought to fans without any use of coaches tape; practice non-participants are reported throughout the week for predicted fantasy value; and success and failure for skill players is measured solely in stats and fantasy points. This is a very different model of football than the one I grew up with. ” These games have become more about money and predictions than about anything else. Perhaps that’s part of the reason that cameras and statistics haven’t become more commonplace in the world of Football.
Our rankings are just that: a ranking of teams. We don’t get into the details of every player like ESPN’s Soccer Power Index does. There are always two sides to an argument, and maybe it’s better that things be left the way they are. For now, we’ll continue to support what others are doing in trying to understand why things work out the way they do.