|Viewing Correlations for Tom Mitchell|
Correlation is used to define a relationship between two variables, with a perfect positive correlation indicated by 1.0 and a perfect negative correlation being assigned a value of -1.0. Correlation however does not indicate the extent to which the two variables change together. Because of this, it’s important to understand the co-variance between the two variables also.
The chart above shows 4 theoretical player’s scores over 6 games. Each of Player’s B, C & D have high correlations to Player A indicating they have a perfect (or near perfect) positive correlation. However, it’s clear from the chart that if you were expecting Player A to have a high scoring game, the best player to stack him with would be Player D as he scores significantly better than the others when Player A scores higher. Correlation does not give us any insight into this and it’s by understanding co-variance that we can take advantage of this.
Looking at the numbers we can see that Player D’s correlation is in fact slightly less than the other players, however, his co-variance is much larger. In simple terms, the co-variance indicates how two variables vary together and as such for fantasy sports the larger the value the better.
So, we can see from this simple example that correlation should always be used in conjunction with co-variance as correlation is only half the story.