Introduction
Las Vegas lines are known to be historically accurate, and thus a decent predictor of scoring in the NFL. One of the first studies published on the accuracy of Las Vegas lines was a 1986 paper by Stanford statistics professor Hal Stern. In it, he used data from the 1981, 1983, and 1984 seasons to show that the margin of victory over the point spread (defined as the number of points scored by the favorite minus the number of points scored by the underdog minus the point spread) was not statistically significantly different from a normal distribution with mean zero and standard deviation 13.86.
Throughout history, this has proven to remain true as well, even into the past decade. I was able to get data for Las Vegas point spreads (the expected score differential between two teams) and totals (also known as over/under, the expected total score of a game between the two teams combined) for games from 2005 to 2014, and performed some distribution analysis on these lines.
First, if we compare the Vegas spread against the actual spread, we find the difference (meaning how far Vegas was off) is approximately normally distributed with a median difference of zero and standard deviation of 13.7 points; not too far off from what Professor Stern found in the 1980s. In other words, Vegas is on average able to pinpoint the relative strength of one team compared to another quite accurately, with approximately 95 percent of predictions being correct to within two touchdowns.
Distribution of Vegas vs. Actual Spread Since 2005
Similar holds true for Vegas totals. In this case, the difference between the actual total and the Vegas predicted total also has median zero, but this time a slightly lower standard deviation of 13.4 points.
h3. Distribution of Vegas vs. Actual Total Since 2005
What does this all mean for the daily fantasy player? It means that Las Vegas casinos are very good at rolling every bit of information to be, on average, quite accurate in setting lines. This should translate into being able to predict team performance very well. Subsequently, this means player performance should be somewhat predictable as a byproduct of being able to predict team performance.
As it turns out, this is the case. Team performance can be calculated through an expected points (ExPts) value. This value is calculated by taking the Vegas total, dividing by two, and subtracting one half of the individual team’s line. As an example, let’s suppose the New England Patriots are seven-point favorites (which translates to a Vegas line of -7) over the Buffalo Bills, with a Vegas total of 44. This means the Patriots ExPts is 44/2 – (-7/2) = 25.5. The Bills would have an ExPts value of (44/2) – (7/2) = 18.5.
If we plot the difference between the actual number of points scored minus the ExPts value, we expect the distribution to be closely centered around zero. Indeed, the median difference is zero, meaning Vegas does an accurate job projecting individual team scores.
Distribution of Vegas vs. Actual Total Since 2005
This distribution is not quite normally distributed, as there is some skew to the distribution. However, it’s close enough that we can still find the standard deviation to be useful. In this case, the standard deviation is 9.6 points, meaning approximately 95 percent of individual team scores fall within 20 points of the ExPts value.
Since we can predict team performance to some degree of accuracy using Vegas lines, the individual components that make up a team should follow suit. That is what the rest of this course tackles. The lessons will examine the individual skill positions (QB, RB, WR, TE) as well as positional stacks to find new insights for the daily fantasy sports player to use.
By adding this to your arsenal of knowledge, you will be able to keep that leg up on your competition.
Lesson One Quiz:
1. What is a Vegas point spread and a Vegas total?
2. What kind of distribution does the difference between Vegas spreads and actual spreads result in, and what is the standard deviation of that distribution?
3. How do we calculate Expected Points (ExPts)?
4. What is the standard deviation for the difference between ExPts and actual team points?
5. Since Vegas predicts game results and team results accurately, what is the next step we will use Vegas to predict?