MLB DFS: How Accurate Are Vegas Implied Totals?
When I started playing DFS in 2015, one of the first things I did was read all the books I could find by Jonathan Bales. Like any new hobby I dive into, I try and soak up all the information I can by those who are successful in the industry. I specifically remember reading a chapter dedicated to Vegas lines and implied team totals. If you’re unfamiliar with Vegas implied totals, I love how Notorious defines it here – “An implied team total is the number of points that a team is projected to score based on the Vegas line and the over/under.”
The takeaway for me was to leverage the work being done by Vegas because it saves time and Vegas is sharper than any work I could do. Ever since then, I’ve relied on Vegas implied totals to help gauge ownership and identify which stacks to target.
I’ve seen comments in the forums mentioning that Vegas implied totals are over-rated or unreliable. I decided to look into these claims – are these statements fair to make or does it still hold true that Vegas implied totals are valuable for DFS players?
The nice thing about Vegas implied totals is that they are all available in the RotoGrinders starting lineups page, so it’s possible to back-test. I decided to pull the Vegas implied totals for every game so far in the 2017 MLB season between opening day April 2, 2017 through June 30, 2017. This gave me 1,196 games to work with. Since each game features two teams, I was left with a sample of 2,392 Vegas implied totals to back-test. I then pulled the actual scores for each game, and was able to compare how close Vegas was. In the example above, Vegas has the Dodgers implied for 4.86 runs and the Padres implied for 3.14 runs. For this article, I took the actual final score the team ended up with and subtracted the Vegas implied total. So if the Dodgers actually scored 5 runs, they scored +0.14 more runs than expected. If they had only scored 4 runs, then they scored -0.86 runs than expected. I did this for each game and each team. Let’s dive in:
On Average, How Close Was Vegas?
This finding shocked me a bit, but on average, the actual team score was just +0.33 runs more than the Vegas implied total over the 2,392 sample. In other words, if Vegas had an implied team total of 3.67 for a team, on average, the team would have actually scored 4 runs. I thought that was pretty darn impressive.
Author’s Note – 7/3/2017, Noon ET – I originally calculated the +0.33 average difference using both positives and negatives. So if a team outscored their Vegas total on a day by +5 but underscored their Vegas total on another day by -4, the average is +0.5. I saw some comments below saying they’d rather see the average difference removing the positive/negative factor and instead focusing on absolute values, so in this example the average difference would’ve been 4.5. I can see the benefits of this view, so I recalculated the data and using this absolute value method, teams on average scored +2.4 more runs than Vegas implied for them over this 2,392 sample. Hope this additional note helps for those who prefer to see the data this way.
I know some of you are more visual learners, so I decided to plot out the difference between the actual team score vs the Vegas implied team total for each of the 2,392 scores. This chart may seem like a lot to digest, but here’s how you would read it. Let’s say a team was implied to score 5 runs, but they actually scored 7 runs. The difference is +2, so this would be represented in the chart by a +2 blue dot. But let’s say a team was implied to score 5 runs and they were shut out and scored 0 runs. This is a difference of -5, so this would be represented in the chart by a -5 blue dot. If Vegas was perfect and nailed every score, every time, all you’d see is a cluster of blue dots along the 0 horizontal x-axis. But we all know sports is filled with variance and this will never happen. This point is most visibly shown in that blue dot way up top on April 30, 2017 when the Washington Nationals were implied to score 3.69 runs and exploded for 23 runs against the woeful Mets, for a difference of +19.31. While there are some outliers, as you’d expect, I think it’s very telling that in general the blue dots tend to cluster along the 0 horizontal x-axis. In other words, over a big sample, Vegas has been pretty good and nailing a team’s total.
Does Vegas Get Sharper Over Time?
One question I had was whether the Vegas implied totals get closer to the actual totals over the course of a season as Vegas has more data to work off of and we have a better feel for every team. In other words, does Vegas get sharper as time goes on? At the start of the season, we come in with personal biases, often driven by previous year’s statistics. Did any of us expect the Yankees to be a high-powered offense driven by Aaron Judge back in March? I doubt it. Since we have three months of data to work with, I decided to find the average differences in April, May, and June. If Vegas was getting sharper over time, we’d expect the average difference to get closer to 0 as the months went on.
Here are the average differences by month. Remember, a positive difference means that the actual scores were higher than what Vegas predicted:
April: +0.27
May : +0.37
June: +0.34
What this is saying is Vegas was actually best at predicting game totals in April this season, which is contrary to what I expected. What this also showed me is baseball will always be a game of variance. In some ways, this is good news for tournament players. Even as we get closer to the middle and end of a baseball season, there’s still going to be variance we can attack.
How Often Did Teams Score More Than Vegas Predicted?
Of the 2,392 team scores, 1,169 teams scored more than their Vegas implied total. In other words, 49% of the time a team scored more than their implied total. This basically means it’s been a 50/50 shot on whether a team outproduces their expectations. I thought this finding was interesting and was hoping there would be a better story here, but this speaks to the power of how sharp Vegas has been this season.
Which Teams Are Crushing Vegas?
I was curious if there were certain teams blowing out their Vegas implied totals this year. For each of the 30 MLB teams, I calculated the average difference in their actual team totals compared to the Vegas implied team totals. A positive average score meant that team was generally outscoring what Vegas predicted for them, while a negative average score meant that team was generally scoring fewer runs than Vegas predicted for them. Here’s how it played out:
There were a few takeaways for me. The first is that the Yankees and Nationals have been crushing their Vegas implied total this season, outscoring their predictions by nearly a full run. But take a look at the other end of the spectrum for the negative scores. Some of the high powered offenses from last season (Cubs, Red Sox, Indians, Blue Jays) have not been living up to their expectations. Last year, all four teams were in the top 7 in team wOBA. This year, none of these teams are in the top 7 in wOBA, team ISO, or home runs. In fact the Red Sox have hit so few home runs that even the Angels have hit more than them (no offense, Pablo Sandoval. I mean literally, you’ve provided no offense for your team). I think it’s safe to say at this point in the season that some of these teams we stacked heavily last season are just not as good as we thought. I’m looking at you, Chicago Cubs.
What Should We Take Away?
To wrap things up, I’m sure some of you are asking this question: Is Vegas coming within an average of 0.33 runs of the actual team score over the course of three months good or bad? Here’s the question I would ask you – are you able to do better? For example, if I told you to guess every team’s score for three months, could you have better results over that time frame? If you can, then kudos to you because you are sharper than Vegas. If you can’t, then using Vegas implied totals in your DFS research is the smarter thing to do. Personally, I don’t think I have the time, resources, or brains to do much better.
While Vegas implied totals should have a place in your research, always keep in mind that scatterplot graph up above. It should serve as a reminder that while Vegas is very good, our best friend “variance” still has a role. How many times does this scenario happen to you? Say you’re interested in stacking Team A, who has a Vegas implied team total of 4.3 runs. But you notice Team B has an implied total of 4.7 runs. All your daily research and gut tells you Team A is the way to go, but it’d be foolish to stack a team with a lower implied team total, right? In the end, you stack Team B and of course, Team A goes off. It happens to all of us, and it’s a reminder that Vegas isn’t perfect. In other words, leverage Vegas data, but it is not the end all, be all.
Thanks for reading. I hope this was helpful, and I would love to hear your thoughts in the Comments section below. You can also find me on Twitter here. May variance be on your side.