Major League Soccer 2019: Statistically Best Players – data analysis

2020 was destined to be a great year in football with the Champions League matches getting more refined and the lieu for fourth place in the Premier League getting more and more slippery. In the midst of a great season, the whole world was hit by COVID-19 which caused everything from an outbreak to a global pandemic.

With nothing but time on our hands, it is time to return back to good old America and probe deeper into the MLS. While the European leagues got a lot of attention, Major League Soccer had recently concluded another glorious season and started their 2020 season. However, with the closure of the league due to the pandemic, an analysis of four games is redundant.

However, what can be done is to look at the 2019 season. In this mini-series, I will be looking over the 2019 MLS season and picking the statistically best players. A season full of data and statistics will be analyzed for all spots to find the statistically best players in every position. The aim at the end of this mini-series is to find some hidden gems and repolish some bright diamonds.

In this tactical analysis and data analysis, we will look find the statistically best defence in MLS. In this analysis, I will be picking the statistically best goalkeepers and centre-backs through data.

Formation and Style

It is always hard to pick the best player from a league. Why?

Simply because players play in different styles. Some tactics allow a player to flourish while others keep them pigeon-holed in a tradeoff for defensive stability. Either way, it is always hard to pick the best player. To make this process more rational, I have decided to set constraints on the formation and the type of player. For each article in this mini-series, I will lay them out as so you know what type of player are being looked for.

The general formation is a 4-2-3-1. This is simply because the experience that I’ve had with analysing teams in the MLS, there seems to be a trend for going to a 4-2-3-1, either in attack or in defence or in a transition.

For this article, there are certain characteristics that will be emphasised.

Goalkeepers: Apart from saving shots, I am deciding to pick an aggressive goalkeeper. This is because the general goalkeeper image is now perceived to be in the image of Alisson/Ter Stegen  the goalkeeper will have to be aggressive in his actions (punches, claims, etc). While there are claims for having a possession-happy goalkeeper, recent trends in other leagues have shown that it is best to start out with a goalkeeper who will definitely keep out the goals and then some.

Centre-Backs: Building off from the possession-model which is popular in MLS, the centre-backs need to be comfortable with the ball, have some creative vision, and be good defensively. One centre-back will be more attacking-minded while the other centre-back would be more defensive-minded. In general, both centre-backs will be defensively good and comfortable with the ball.

Additionally, the sample size for the data analyses will be players who have played more than 20 games. This strict sample is so that we are only considering those who have put consistent performances. Some restrictions are made here and there and those will be stated in the analysis.

Goalkeeper

Goalkeepers are the single-most critiqued position in the world. There is no chance for mistakes as a mistake is a goal in that area. Contrast this to the mistake of a striker which simply results in the dispossession of the ball metres away from the goal.

Apart from that, it is also the single hardest position to analyse through data. Simple metrics such as xG and xGA do not provide us with usefulness on the impact of the goalkeeper. Even tactical analysis through photos requires careful analysis of frames and frames of photos. What makes this analysis harder is that the single most important metric for goalkeepers  their ability to save goals  is almost impossible if you don’t have access to advanced analytics from the likes of Opta.

However, I have made a simple version of John H.D. Harrisons analysis to find good goalkeepers from the bad goalkeepers. But in order to understand the mathematics, we need definitions.

Save Percentage is defined as out of the total shots on target, how many goals does he save? This is as a shot has a chance of becoming a goal so saving a shot from going in is the same as saying that a goal has been prevented.

Expected Save Percentage is defined as out of the total shots on target, how many expected goals does he save? Each shot has a probability of going in and this is simply a metric telling us what should be his save percentage if he actually conceded the goals he should have.

A simple way of understanding the two is through a screamer. The chances of a screamer are low of going in so your expected save percentage for that screamer is high. However, if you concede, your save percentage was zero because you let that shot become a goal. Doing this for all the shots in total, we get a total save percentage and expected save percentage for the season.

So now let’s put all those words in numbers which may simplify these two concepts.

Save Percentage (SP) = (Shots On Target Total  Goals Conceded)/(Shots On Target Total) x 100

Expected Save Percentage (xSP) = (Shots On Target Total  Expected Goals Conceded)/(Shots On Target Total) x 100

Now onto the complicated part. How do we define good goalkeepers with these statistics?

A difference between the actual save percentage and expected save percentage tells us how you performed. A positive difference means that you are conceding less than expected  this is what good goalkeeping looks like. A negative difference means that you are conceding more than expected  this is what bad goalkeeping looks like. Obviously, these statistics aren’t perfect however they are a good estimate.

There is another problem, however. We need to know if the goalkeepers are doing better with easy shots or hard shots. Ideally, we’d want a goalkeeper conceding less with hard shots. That’s ideal so the closer we get to that, the better. To quantify this, we simply use the Expected Save Percentage (xSP). A higher xSP means that the data shows you were expected to save more shots  you are facing easy shots. A lower xSP means that the data shows you were expected to save fewer shots  you are facing hard shots.

One final mathematics we need to understand is to what Goal Prevented is. It is simply how many goals a goalkeeper prevented. It is simply defined as:

Goals Prevented = Expected Goals Against  Goals Conceded

Positive numbers mean that you saved goals while negative numbers mean that you let in more goals.

With all that hassle, let’s finally analyze and see who are the good goalkeepers.

In this graph, (SP-xSP) are graphed on the y-axis while (xSP) is graphed on the x-axis. In addition to that, Goals Prevented is charted as a colouring aspect. Immediately we see the two good goalkeepers: Matt Turner from New England Revolution and Steve Clark from Portland Timbers.

All the MLS goalkeepers are expected to have a high expected save percentages however only Turner and Clark are actually the best at preventing goals. This is backed up by their Goals Prevented. Clark prevented 8.4 goals during the whole season while Turner prevented 8.21 goals. These are huge margins for Revolution and Timbers respectively.

A deserved shoutout is for Jesse González from FC Dallas who is act