By Jaboner Jackson 8 am | Last week, we looked at standard deviation (SD) as a level of variance. Raiders fans were warned to tread lightly as my lesson did involve math but fortunately many Raiders fans were actually able to follow along. To familiarize yourself with last week’s lesson, check it out here. This week, we are going to look at correlation.
Correlation describes the degree of relationship between two players. A correlation can be any value between –1 and +1. When two players have a correlation of –1, their fantasy football points will move in the opposite direction. When two players have a correlation of +1, their fantasy football points will move in the same direction. Let us work through a real life example to understand this concept more fully. We will once again utilize the Cincinnati Bengals.
In 2010, Heisman Palmer (aka Carson Palmer) had Terrell Owens on his team. They had a positive correlation of approximately +0.5, meaning that whenever Palmer did well, T.O. also did well, and whenever Palmer did poorly, T.O. also did poorly. Accordingly, if you had both of them on your team, you had either feast or famine, which makes intuitive sense. It was obvious without any statistical analysis that their points were related to each other, since Palmer was throwing the ball to T.O.
If your fantasy football team was playing a strong opponent, it made sense to start both Palmer and T.O., since a good game with one of them automatically made the other one have a good game as well. (Hence the positive correlation.) However, if you were playing a weaker opponent, it made little sense to start both Palmer and T.O., since a poor game with one of them automatically made the other one have a poor game as well. In other words, it was worth the risk to start both Palmer and T.O. when you needed to score a lot of points against a strong fantasy opponent but it was not worth the risk to start them both when you only had to score only a decent amount of points against a weaker fantasy opponent and it made more sense to start only one of them.
Sometimes players can be negatively correlated as well. This means that when one player does well, the other player does poorly. This is most obvious in cases of a starting quarterback on an NFL team and his backup. Kevin Kolb and Michael Vick had a strong negative correlation, meaning that if Kolb did well, Vick did poorly. This makes intuitive sense because we do not expect both quarterbacks to be playing at the same time. If Kolb did well, it was because Vick was not playing, and vice versa. In reality, it is hard to find players that are negatively correlated. Football is a team sport and so success and loss is spread out throughout teammates. Nonetheless, the reason for “handcuffing” is because of this negative correlation. A starter and his backup are negatively correlated. (Handcuffing means having your starter’s backup on your fantasy team. For example, you would handcuff Darren McFadden with his backup, Michael Bush.)
Nonetheless, certain players will be less positively correlated than others. Adrian Peterson and Brett Favre had a low positive correlation in 2010, meaning that if Peterson did well, Favre did not, and conversely, if Favre did well, Peterson did not. This again makes intuitive sense, since the more that Favre threw the ball, the more points he had, but the fewer points Peterson had.
Homework: Let the concept of correlation percolate and review standard deviation again. Next week, we will put both of these statistical concepts together to create a winning fantasy football team.