While we have seen advanced stats revolutionize baseball and basketball, soccer has not been as quick to get on the analytical bandwagon. At the end of July, Bruce Arena, the godfather of American soccer even dropped this dropped bomb:
— Adam Serrano (@LAGalaxyInsider) July 24, 2016
I do agree with Bruce that soccer is harder to quantify than baseball or baseball, but his argument is a false cause. The problem is not using stats to analyze player performance, it is that the traditional soccer stats that people use are broken and not useful. Looking traditional stats like number of shots is not an example of using analytics.
The purpose of advanced stats is to create stats that can better quantify player performance than traditional stats. Traditionally in basketball, Field Goal percentage (FG%) was the measure of whether or not you were a good scorer. However, FG% does not account for the fact that a three-point shot is worth more than a two-point and that players can be fouled while shooting. This lead to development of True Shooting Percentage (TS%), which adjusts for both three-point shots and free throws to create a more accurate stat to analyze shooting efficiency. We can do the same thing for traditional soccer stats.
Does Saunders Need to Be Replaced?
Now that Frank Lampard is scoring goals off of various body parts, Josh Saunders has assumed the title of most controversial NYCFC player. Some consider him a fan favorite with great shot stopping ability with a slight propensity for gaffes, while others consider him to be the worse keeper in MLS and incompatible with Patrick Vieira’s play-out-of-the-back system. With the Colorado Rapids replacing Zac MacMath with Tim Howard, and MacMath now publicly demanding a move, many have floated MacMath as a replacement for Saunders.
Traditionally, the statistics that are usually used to compare goalkeepers are Wins, Save Percentage (SV%), and Goals Against Average (GAA). There are issues with all three of these stats. A great keeper can lose games on a bad team, and conversely a bad keeper can win games on a great team. No one is saying that Bill Hamid is a bad goalkeeper because DC United does not score enough goals to win games. SV% can be skewed by the quality of defense in front of the goalkeeper. Having to save 10 shots in the 6 yard box is going to be much different from having to save 10 shots outside the box. GAA does not account for the quality or quantity of shots. With everything equal, more shots (of the same quality) will lead to more goals and are not necessarily indicative of goalkeeping ability.
Goals Saved Above Average (GSAA)
One stat that we can use to paint a better analytical picture for comparing goalkeepers than the three above is Goals Saved Above Average (GSAA) which was created by the folks at Hockey Reference:
The formula figures out how many goals the league average goalkeeper would have saved given a certain number of shots on target and then factors in the number goals the keeper actually saved. This normalizes the number of shots on target for all goalkeepers. If you have a positive GSAA, you are an above average goalkeeper and if have you a negative number you are below average
One problem with this stat is with more games played, an above average goalkeeper will accumulate more GSAA and a below average keeper less GSAA. In order to do an accurate comparison, you would have to make it a per minute. I choose per 90 so you could get an accurate assessment of goalkeeper for a 90 minute game.
Your inner Bruce Arena is probably like: “But it still doesn’t take into consideration the type of shot, distance of shot, free kicks, penalties, team playing with 10 men, etc.” Bruce is absolutely right, but that is a much more complex consideration. If I can find the right data set, I will make these adjustments during the offseason. Still, GSAA/90 gives a better tool to compare goalkeepers across the league than GAA or SV%.
I took at the stats from mlssoccer.com. I had to assume that in their goalkeeper stats that “Shots” really meant “Shots On Target” because their SV% was calculated by Goals Allowed/Shots. With 1427 saves out of 2084 shots, the league average SV% came out to be 68.5%
The one guy that really stood out to me was Jake Gleeson. There are 11 keepers with over 500 minutes played that have a better GAA than Gleeson’s 1.24, but his second place ranking GSAA/90 shows that he is performing remarkably well for the Timbers considering all the shots that his backline are giving up and maybe even a keeper of the year darkhorse.
I think there is still some of the same bias as SV% in GSAA/90. There is still no adjustment for the situations that affect the quality of shot that I mentioned above. All of those things could be impacted by the quality of the backline. In fact, most of the league leaders in Goals Against (Houston, San Jose, LA, DC) are also at the top of GSAA/90. It makes you wonder if the keeper is making the defense or if the defense is making the keeper. On the other end, you have a couple of keepers (Ousted, Rimando, & Blake) that are generally considered to be among the league’s best, but whose teams are in the bottom half of Goals Against. Are these keepers overrated, or are they just good keepers with porous backlines? In order to answer both questions, we are going to have to adjust GSAA/90.
GSAA/90 also gives an interesting comparison for some teams that have split time at keeper since you can assume that the team defense is the same. Tyler Deric (0.659) looks like he should be the preferred option over Joe Willis (0.075). Despite Travis Worra’s (0.067) early season form, Bill Hamid (0.394) is a no brainer for DC. If the rumors of Sean Johnson (0.054) leaving at the end of the season are true, Chicago will probably be fine with Matt Lampson (0.267) taking over full-time.
Saunders vs. MacMath
Saunders does not look good. Like most goalkeeping stats, Saunders ranks near the bottom in GSAA/90 with -0.265. Bobby Shuttleworth (-0.463) is the only full-time keeper that has a worse rating. The question for both keepers is how much of that falls on their shoulders and how much is the responsibility of their respective defensive lines. Once again this is where an adjusted GSAA/90 would be a lot more helpful.
However, this does not necessarily mean that NYCFC should make a move for MacMath. Tim Howard (0.137) has nearly doubled MacMath’s (0.063) GSAA/90. While MacMath’s GAA (0.69) was second only to Tyler Deric (0.60), his GSAA/90 is barely above average and only in 11th place. That puts him in middle third of all keepers with 500+ minutes. That makes me question how much of his GAA was really the product of how good the Rapids defense was and whether MacMath is really just a middle of the road MLS goalkeeper.
While the stats show a big gap at goalkeeper for the Eastern Conference leaders, NYCFC did not show any proclivity for making a move during the summer transfer window. If the Rapids move MacMath, his price will probably be inflated by his 0.69 GAA and will probably not be worth the cost. NYCFC might already have Saunder’s replacement on roster already with unproven 6’7 Nordic monster, former Manchester City academy product, and soon to be green-card-carrying Eirik Johansen. It would make sense to give him a shot before looking at MLS or international player, but Vieira has shown little inclination to start Johansen despite Saunder’s struggles at times this season.
GSAA/90 gives an interesting league wide comparison for goalkeepers. I hope to adjust GSAA/90 to adjust for shot quality in the near future. If you like this article, please give a like/follow/retweet on Twitter and Facebook, and continue the discussion on the NYCFC Forums and Reddit.
Josh Saunders By BrokenSphere (Own work) [GFDL (http://www.gnu.org/copyleft/fdl.html) or CC BY 3.0 (http://creativecommons.org/licenses/by/3.0)], via Wikimedia Commons
Zac MacMath By Ryan Byrne (DSC_0718.JPG) [CC BY-SA 3.0 (http://creativecommons.org/licenses/by-sa/3.0)], via Wikimedia Commons