MLS Fantasy Boss A Fantasy MLS Community
Last month I posted some stats from the 2013 MLS season. Since then, I’ve done a little more work and I’m ready to talk what we can learn from this data. Specifically, I took a look to see which stats were most related to a player’s final total score. The idea was that by identifying the stats that contribute the most to a player’s score, I could find some promising bargains for the 2014 season.
If you like to see the results, Keep Reading after the break.
A Quick Look at the Math
If you’re not interested in the numbers behind this article, just skip down to the results section, but if you are, here’s how I gathered the data. If anyone finds any errors or has an idea of how to get more accurate results, I’m always open to feedback.
After getting all of the stats, I had the raw numbers for players but not the points. Just because a player has 60 CBIs dos not mean that the player earned 10 points from them. So I needed to look at the breakdown of points earned in each game to determine how many points a player actually earned. (Point breakdowns can be found in the Scoring section of the MLS Fantasy Rules: http://fantasy.mlssoccer.com/rules/)
Once I had the number of points that each player actually earned from each category, and the total number of points that the players earned, the only thing left was to see how they were related. To check for this relationship, I used the Pearson Correlation Coefficient (PCC). This equation allowed me to measure the strength of relationship between two variables. Example: if X= Total Points Earned from Min Played and Y= Total Score, then we could find out if more Min Played had a strong relationship to a higher Total Score. A result of say 0.8 would suggest that the more points a player earns from Min Played the higher the likelihood of a higher score. A result of say 0.32 would mean the opposite. You can get a more in-depth description here.
Obviously, any category that earns players points will have some relationship to a higher Total Score, but this formula puts that relationship on a scale from 1 to -1. The closer a result is to 1, the stronger the relationship.
- Strong Relationship= The higher the stat the higher the total score. Range from 1.0 – 0.75.
- Moderate Relationship= Tendency for higher stats to go with higher totals. Range from 0.74 – 0.5
- Weak Relationships= Small influence on higher stats to go with higher totals. Rang from 0.49 – 0.01.
As for my sample size, it was possible for me to look at every player, but I decided against that. Frankly it was due to time restraints. In the end, I decided to go with the top 25% of players (rounded up) for two reasons. First, this covered a majority of players that I thought would be first choices, and second, the results were not changing significantly with additional players added.
My Results (by position)
For Keepers, I looked at 11 players. The stats I looked at were Min Player, Clean Sheets, Goals Conceded, Saves, Penalty Saves, and Recoveries.
Of all of these stats, only Min Played had a Strong Relationship with a PCC of 0.85. Next was Clean Sheets, which had a Moderate Relationship with a score of 0.6. All of the other stats resulted in Weak Relationships (Goals Conceded=0.32, Saves=0.11, Penalty Saves=0.46, and Recoveries=0.48). I was a little surprised at this result. I expected the +4 Clean Sheet bonus to be the most important stat for Keepers, but Min Played actually made up 1/2 to 1/3 of the top keepers’ total points.
So Min Played and Clean Sheets are most important for keepers, moving on… but wait. The top two highest scoring keepers had three fewer clean sheets that the players with the most (14). So there must be, at least, one more stat that can help right? This is the question that bugged me and when I looked at top scoring keeper Tally Hall, I found the answer I was looking for. Keepers that were able to generate a lot of Recoveries were able to make up for the points they missed out in Clean Sheets. And that’s what helped the top two Keepers earn their spots.
For Defenders, I looked at 40 players. The stats I looked at were Min Player, CBIs, Recoveries, Clean Sheets, Goals, and Assists
Of all of these stats, none returned a Strong Positive Relationship. Clean Sheets was the only stat that had a Moderate Relationship with a score of 0.55. All of the other stats resulted in Weak Relationships (Min Player=0.42, CBIs=0.41, Recoveries=0.41, Goals=0.47, and Assists=0.19).
While I was disappointed that there were no Strong Relationships, I did not find the overall results to surprising. Maximizing clean sheets is still the best way to get the best score from your defenders. Additionally, the value of a defender who is able to generate a lot of defensive actions is desirable. Goals actually had the highest of the Weak Relationships (even more than defensive actions), which only reinforces how valuable attacking defenders can be.
At first glance it looks like you want the perfect defender, but it’s not really that complicated. Center backs are excellent to meet these qualifications. They have ample opportunity to get CBI and Recovery points and many of them come forward for set plays.
For Midfielders, I looked at 51 players. The stats I looked at were Min Played, Crosses, Key Passes, Goals, and Assists.
As with defenders, none of these stats returned a Strong Relationship. Of the three stats that had a Moderate Relationship, Goals was the highest at 0.67, followed by Key Passes at 0.63, and then Assists at 0.55. Min Played (0.33) and Crosses (0.49) both resulted in Weak Relationships.
As with defenders, most of these results fit with my general expectations. What I did find surprising was that Crosses did not have a stronger relationship. As with the defensive actions (CBIs and Recoveries) I expected crosses and key passes to be closer and for crosses to be the higher of the two due to set plays. But it does make sense since players only get points from crosses when they are successful and in the penalty area.
Overall, I believe these results reinforce the conventional wisdom that attacking midfielders and the way to go, unless you find the rare Amobi Okugo who is out of position on defense and is able to make up for it with CBIs.
For Forwards, I looked at 27 players. The stats I looked at were Min Played, Goals, Assists, Crosses, and Key Passes.
Once again, no Strong Relationships, but unlike all of the other positions, every result was a Moderate Relationship. This was the order: Goals (0.74), Assists (0.69), Key Passes (0.63), Min Played (0.61), Crosses (0.55).
It’s no surprise that these results are all so close, all forwards do is attack and the best way to do that is with goals. What I did find interesting was that since the stats have a similar impact on a player’s final total, it’s easier for a well rounded forward to makeup for a lower goal total. The perfect example is Federico Higuain who was able to edge out Camilo Keane because of his ability to also provide crosses and key passes. But these types of players are rare.
What stats do you look for when picking players? If you’d like to continue this discussion about player stats, take some time to visit the Fantasy MLS Subreddit or leave a comment here. And as always, you can always reach me on Twitter.
Why did you choose to look at total score instead of PPG or PP90? That heavily skews the importance of minutes played, which is mostly just a function of a player’s health and lack of international duty for the players we consider in fantasy.
I choose total score because I was trying to find out which stats were most related to higher scores.
I believe you are only partly correct. Min played made up the majority of a players points from the individual stats, and I agree that that does mostly reflect how healthy a player is.
But in 3 of the 4 positions, Min played did not register as a stat that had a Strong relationship to higher total scores. And in the 2 where a Medium relationship was returned, it was not the highest of the Medium relationships.
So I don’t believe Min Played was heavily skewed.