By Warren Sharp

Vegas totals are wrong. And sports books know it. Which is why they start off the week with low limits to see which direction the money will go, and adjust accordingly. Often, they adjust quickly and at times, too much. Why? Because they are nervous and scared. They dislike risk, and they know that certain individuals are far more accurate at projecting the totals on games than they are. Those individuals end up working for betting groups or large sports bettors (who operate individually or as a collective) and bet the games they are given with a lot of money. And they win more often than they lose.

But the whole concept of “Vegas” totals is a bit silly. Hopefully, the DFS community uses the buzz word “Vegas” to describe general “betting” lines as opposed to the city itself. Because you can ask anyone from Las Vegas legendary pro bettor Bill “Krackman” Krackomberger to the grinder trying to pay his rent: totals are rarely originated in Las Vegas. Instead, most if not all totals originate at offshore books which do significantly higher volume than the books in Las Vegas. The linemakers in Las Vegas then take these lines (once posted offshore) and either copy them, or tweak their own number to closely reflect the offshore odds. There are a handful of linemakers in Las Vegas who would be capable of originating and booking action on their own lines, but if their number are set too far from the number at major offshore books, the Las Vegas linemakers will get steamrolled with sharp action. To avoid that problem, they tend to align exactly or close enough with the offshore lines. For that reason, when I use the term “Vegas” totals in this article, I (as well as you) should not confuse that to say these lines are born and bred in Las Vegas – but they are the betting line and we’ll leave it at that.

I understand Vegas totals more than most, because while I am a quant and a football analyst, I got my start betting and beating the market on NFL totals. I worked to develop a model used to predict NFL totals. I spent years researching, inventing and testing the model while receiving my Engineering degree and later, my Professional license. I’ve been using the model for 10 years now over at SharpFootballAnalysis.com and have a long term success rate of accuracy of 60%, including recent success rates of 72% last year and 61% in 2014. Many of my recommendations move the Vegas line when I release them. What that means is, if you see a total that opened at 51 on Tuesday, and suddenly you look at the lines and the total is 50, then 49.5, then 49, there is a chance it was a game I recommended to bet under. Pro betting groups work with me and bet my recommendations and the Vegas linemakers who run the books know me. I write for ESPN Chalk where I share a couple recommendations each week, and those hit 66% last year. While perhaps superfluous, the point is: I appreciate, understand and value the attempt at setting the Vegas total more than most. It’s been my primary focus both in and out of the NFL season for well over a decade now.

**Trend in DFS**

More and more, we see Vegas totals being incorporated into DFS. Perhaps this was because, at one point, no one playing fantasy football cared about Vegas totals. Eventually, some in the DFS community started to discuss them. They realized that some of the players that performed well played in high totaled games. And some that performed poorly played in lower totaled games. The “secret” began to get out. More and more people began discussing it. Now, it has permeated the market. DFS players want to start players with high totaled games, or high team totals.

Here is the problem: not only has it jumped the shark, the data shows it is not helpful. The reason the player with a high team total or total may perform well is NOT because the team total or game total is high. It is because he is playing on a good, offensive team which may not have the best defense, and in many cases could be playing against a weak opponent. As a result of that, the team or game total may be high. But the total is just a byproduct of the situation and matchup. You know football, so think fast – which game is likely to be higher scoring (and thus have a higher total) – the Saints at the Falcons or the Titans at the Broncos? You don’t need to know the actual Vegas total to know the answer. To put it in different terms: go to a basketball court with a basketball and a beach ball. You already should know that a shot with the basketball is more likely to go in the hoop because of what it is (its weight, size and the size of the rim) as opposed to the beach ball. You don’t need a physics professor who happens to be stretching out before a pick-up game to come up to you and tell you that you’ll have a better chance if you don’t shoot the beach ball.

So why should it matter what the total is if you already know that a certain offense (Falcons) is going up against a bad defense (Saints) and there are favorable matchups and situations for the game to be high scoring? Does it really matter if the total is 46 or 49 for DFS purposes? Absolutely not. Should you shy away from playing Mohamed Sanu if the total is 47, but play him if the total is 51? Absolutely not. But let’s start discussing math, regression and correlation to explain further.

**How Wrong are Vegas Totals?**

Before we get into this, you should know where to find accurate Vegas totals, and by accurate I mean by real sportsbooks that take sharp action. Also by accurate, I mean the lines are “live”, meaning they change as bets come in and are accurate. HERE is a great link to bookmark from my website which gives you live lines from real sportsbooks (in Vegas and offshore). It also shows you the opening line, so you can see where the line has shifted. But the cool secret about this page is if you click on any specific game’s line, it will pop up and tell you exactly at what time during the week the line moved and by how much. So you can see a documented “history” of the total over time. Now that you can accurately view live Vegas totals, let’s get back to the question at hand.

This question should be viewed with the lens of “…in terms of helping in DFS”, but I’ll discuss that momentarily. First, let’s just look at the basic question. The answer is they are quite wrong.

We can examine first regression, then correlation. When you run a regression of the Vegas total to the actual points scored, you will notice that there clearly is some relationship. Inherently, there should be some type of relationship. But the key is, how weak that relationship truly is: the R^2 is just 8.7% over a 3 year sample, which is extremely small. What does it look like and what does it mean?

Above you can see what it looks like. Instead of a narrow cluster of actual point scored outcomes which bunch closely together near the projected total, you see a true amalgamation of results without a dominant, definitive pattern.

When you add the linear trend line, as shown below, you can see that there does exist a relationship between the points scored and the Vegas total. But it is weak, relatively speaking. Specifically, the R^2 is 8.7%. What the R^2 means is that just 8.7% of the total points actually scored by teams are explained by the Vegas total itself. Only 8.7%.

By most definitions, a low R^2 like this would fail to prove predictive. With a low (good) P value (in this case 1.62E-17), the sample size is adequate. Sometimes low R^2 values are still viewed as acceptable with acceptable P values. For instance, the number of people who eat deli meat once a week and live in cities to the number of people who develop cancer. Because so many other factors influence cancer rates, even an 8% R^2 in that sample would be large. But here is the problem with our sample. Theoretically, Vegas totals are trying to predict the combined points. There are not randomly put up against total points. We’re not correlating number of TV commercials aired with combined points scored, or number of fans in the crowd with combined points scored. The Vegas total’s objective is to try to predict the combined points. That is its goal. And with an 8% R^2, it fails.

Next let’s move to discussing correlation using correlation itself. In this example, we are focusing only on 2015 data. The correlation coefficient between the Vegas total and actual point scored was 0.25. What does this mean? A value of exactly 1.0 means there is a perfect positive relationship. A value of -1.0 means there is a perfect negative relationship. A value of 0 means there is no linear relationship between the two variables. Values between 1 and 0 can be described from strong to weak, and in this case, a value of 0.25 means there is a very weak positive relationship between the Vegas total and actual points scored. Which is what we see after adding the trend line to the graphic.

To add context, let’s compare it to something near and dear to the goal of any of this, which is to try to help you make sense of Vegas totals and (more importantly) win when playing DFS. According to TJ Hernandez of 4for4football.com, the correlation between Draft Kings player price and his actual points scored was 0.26.

This is a big issue. The player price actually has a STRONGER correlation to how many points that player will score than does the Vegas total to how many point will be scored in the game. Why is that a big issue? Ultimately, as a DFS player, you should be concerned first and foremost with how many points the player will score (when factoring in his price). Using Vegas totals to forecast game score still does not get you to your goal. If you factor in Vegas totals and assume the game sees the number of points Vegas is forecasting (a terrible idea given what was presented earlier with the regression analysis) you still need to then take the step of (hopefully) accurately distributing those points to the players themselves. It is a multi-step process. And every bit of low correlation becomes magnified with each iteration.

**High Totaled Games**

Hopefully we’ve explained why DFS players looking to Vegas totals as a secret weapon to win tournaments or cash games is not the best strategy, because the total is actually not very correlated to the actual points that will be scored. Therefore (as an example) a total of 48 is really not measurably more likely to see strong fantasy performances than a total of 44, all other factors aside.

But what about high totaled games? What if we ignore the stepwise differences between totals of 42 vs 43 vs 44, and just try to target high totaled games? Maybe those are likely to produce stronger fantasy performances?

First of all, the reality is that the last 3 years, just 49% of games totaled above 45 have gone over the total while just 48.7% of all games have gone over the total. So picking a high totaled game and assuming it will go over because of its high total is problematic. It is basically a coin flip.

Second, the research above tells us that while there is a relationship between Vegas total and total points scored, it is very weak. Over the course of many years, games with a Vegas total of 49 should see more points scored than games with a Vegas total of 45, for instance. But in the single week you’re selecting your lineup, it’s far from predictive. Even over the course of a couple of seasons, large sample sizes can prove the lack of use in the way DFS players use them.

As an example, over the last two full seasons, games with a Vegas total of 45 (+/- 0.5 points on the total) have actually **outscored** games with a Vegas total of 49 (+0.5 points on the total) across a sample of almost 150 games.

But third, we can turn to some great research done by Evan Silva of Rotoworld.com, and furthered by his colleague Sean Fakete. Evan used a three year sample to try to come up with a profile of the top overall QB on a weekly basis (what traits should DFS players look for) as well as those QBs who scored in the top 6 (Evan’s article is linked in the first sentence). Sean expanded the study for WRs and RBs, and took it past just top 6 but also to top 12. And ideally, when paying DFS, particularly tournaments, this should be what you care about. You need players to not just score points, but to outscore their positional field. So you need top 6 or top 12 RBs or WRs when competing against thousands of other entrants.

What they found was insightful.

- Only 49% of top 6 QBs played in a game with a Vegas total of 47 or more.
- Only 45% of top 6 WRs played in a game with a Vegas total of 47 or more.
- Only 44% of top 6 RBs played in a game with a Vegas total of 47 or more.

The guys you should be targeting to put up top 6 performances are **no more likely** to be found in high totaled games than low totaled games. In fact, by a very slight majority, they are more likely found in lower totaled games. (For reference, the average total last year was ~45 points).

Now what to do? Using Vegas totals to accurately predict points scored isn’t the shortcut. You can’t take the work out of winning by choosing to select player A over player B just because his Vegas total is 3 points higher. And looking only at high totaled games isn’t likely to help you win tournaments or other high stakes games, either. Because more often than not, the top players at each position aren’t even playing in the games with the high totals. So now what?

**Team Totals**

Some research has turned to team totals. But like everything thus far, I’ve come to ruin your day. I’ve read several studies this offseason discussing team totals and what number are ideal for solid results.

The Rotoworld study by Sean with respect to WRs looked at a cutoff of 22.5 points for a team total, and looked at the number of times a team was lined with a team total above 22.5 points, and how often those players hit at a high rate to put them into top 1, top 6, top 12 etc. criteria for weekly performers that week. The study also modeled how often the team was favored by the linemakers. What was not discussed was how extremely similar those two sets of results were.

That is because the reality is this: last year (throwing out games that closed a pickem) there were 245 teams which closed as favorites and (therefore) 245 teams that closed as underdogs. The number of underdogs whose team total exceeded 22.5 points was just 25 (out of 245). So when looking at the teams who have a team total of at least 22.5 points, the extreme majority will be favorites.

I read another study which looked at team totals and suggested that a team with a projected team total of at least 24 is ideal. Again, while not directly measuring this, this is essentially simply targeting favorites, as last year only 8 out of 245 underdogs had a team total set at or above 24 points.

Why is it important that both studies essentially say the same thing – that teams with high team totals may produce better weekly results? Because favored teams typically produce better results. Both studies ran those correlations and found the results between favorites and top performers by position is significantly stronger than any correlation to game total. The team that is better is likely to have their way with the underdog and impose their will more often. Which is why linemakers favor them in the games. And their DFS plays perform better because they are better and win their matchups. The studies looking at high team totals are effectively doing nothing more than isolating favorites.

But there is a second brutal truth about team totals: while DFS pros try to incorporate them into the Tuesday, Wednesday and Thursday research, it is irrelevant. There is no such thing as a Vegas team total on a Wednesday. Sportsbooks will not put up lines for you to wager on team totals until (typically) the day before the game itself. Some don’t even put them up until the morning of the game.

Why? Because the books understand the team total itself is basic math derived from the line and the total. And they don’t like the risk involved in having to juggle the team total around so much during the week as the line could get bet one way by a few points and the total could get bet up or down by more points even than the line. You can calculate your own projected total points on Tuesday or Wednesday, but just know you must adjust as the week progresses, and the number really only becomes accurate by Sunday morning. And by that time, most lineups are already in and finalized, so it is a difficult task (and can be time consuming as well as counterproductive) to even focus on team totals.

I would encourage DFS players to look at the following table, which shows 3 year samples of delta points scored or allowed based on the lines. Delta points scored (dps) is simply the team total set by linesmakers less the actual points scored. As an example: the Bengals have frequently exceeded their team total as a dog, going “over” in 11 of 14 games, 2nd best for an underdog. But as a favorite, particularly on the road, they are below average and frequently don’t exceed their projection.

Meanwhile, the NFC North teams from Detroit and Green Bay have been terrible bets offensively when underdogs, failing to hit their team projection in 16 of 22 games combined. To continue the exercise, a team like the Bills, when at home, has performed tremendously against their expectations. They have exceeded the lined total in 9 of 12 games as a home dog and in 6 of 10 games as a home favorite. Combined, that is 15 of 22 games, or 68%.

Hopefully this analysis changes perspectives on blindly using team totals. Understanding that the linesmakers are typically very wrong on a game to game basis, with some teams scoring considerably more or less than projected, but averaging out in the end. Fortunately, as bettors or DFS players, we don’t have to be right about every player or game we bet. We just need to be right on the few that we target.

**What Next?**

Some DFS players out there won’t care what the real numbers show. They heard that some successful winners use Vegas totals as a key element of their analysis. Perhaps they will still use Vegas totals as a barometer or a comparison tool to help decide on a particular player (“A is in a game with a 49 point total, B is in a game with a 44 point total, so I’ll ride with A”). Again, this is despite the research done by myself as well as Evan Silva and others at Rotoworld which demonstrates that high totaled games are meaningless when trying to target those players who might finish top 6 or top 12 in their position.

As mentioned earlier: the correlation between player price and his points scored is stronger than the correlation between the Vegas total and the actual real points scored in the game. And iteratively, it stands to reason that the platforms have done a better job pricing a player’s actual result than you will when trying to incorporate Vegas lines into your decision making.

But what about the DFS players who incorporate the totals and swear by it? The reality is they are just getting a placebo affect from picking players on teams that ultimately score more points. It doesn’t take a genius to predict that when Drew Brees plays the Jaguars, he’ll likely score more than when he plays the Texans. Yes, you could look at the total on the game or the team total and find out precisely what the linemakers think of the game, but it is pretty common sense to anyone without even looking at that information. Often the players targeted by DFS players factoring in team totals just so happen to be in a number of favorable categories which correlate to success more than the Vegas total itself, such as the fact the DFS player is playing at home, as a favorite (and later in the season) perhaps facing a defense with a key flaw or injury that will be exploited. You can historically look back and try to correlate this player’s success that week and claim it was because of the Vegas total, but the reality is it was the other things that caused this play to be successful, and the Vegas total just came along for the ride.

Additionally, have you ever had moment where you’re debating between two QBs? Drew Brees is playing the Jaguars while Sam Bradford is playing the Panthers. You can look at the total on the games and the team total and see that Brees looks better because his team is likely to score more. But then look at the price, and you will find he is also more expensive. And the reality, to those who have been painstakingly factoring in the Vegas totals, is that the price on the DFS platform (the one sitting right there in the menu of quarterbacks) is actually more likely to tell you exactly how many points Brees will score as does Bradford than the Vegas total would.

Thus, here we find items that are not helpful (not nearly as they are made out to be):

- Game totals – they are actually extremely inaccurate, inefficient and the static pricing on the DFS platforms itself is a better measure of comparative interposition strength than the lined totals.
- Team totals – the studies done on higher team totals essentially scrap all of the underdogs and look at how often good QBs, WRs, RBs etc. play on teams that are favored. And we know that correlation (being the favorite) is stronger, so it’s not the team total itself that is driving the correlation. In addition, any “team total” calculated for Sunday games which you see before Friday at the earliest, likely Saturday, is nothing more than a projection, subject to change.

In sum, the boost that some might find when layering Vegas totals into their evaluation is likely nothing more than residual impact of other factors already incorporated.

**An Underappreciated Research Gem**

One item from the studies I read that is helpful is so elementary it gets overlooked, but is actually far more astute than starting players based on the Vegas total: pick players from games that go over the total.

To quote Evan Silva:

“The fact that 88.2% of overall QB1s and 78.1% of top-six [weekly] fantasy passers played in games that went over the Vegas total is a reminder that identifying high-scoring contests is a critical component in forecasting high-scoring quarterback games.”

Those rates are massive. And Evan is right on the money. Another Rotoworld study on other positions, such as RBs and WRs, found similar results. This should not be surprising. But, frankly, it gets totally overlooked in the modern DFS world, where the fascination is all about “Vegas totals”.

From Evan and Sean’s research, regarding weekly top 6 at each position:

- QBs:
- 78% were from games that went over the total
- Only 49% were from games with a Vegas total of 47 or more

- WRs:
- 64% were from games that went over the total
- Only 45% were from games with a Vegas total of 47 or more

- RBs:
- 64% were from games that went over the total
- Only 44% were from games with a Vegas total of 47 or more

Forget trying to be “cutting edge” by trusting linemakers on their Vegas total to select DFS lineups. This has been proven to be overvalued and frankly, overused. You won’t find the answer to selecting top-6 or top-12 performing players to build tournament-winning DFS lineups from the Vegas total. Just over half the games go under the Vegas total and almost half go over.

**This Is How You Zig While Everyone Else Zags**

As shown above, the focus should instead be on isolating those games you believe will go over, regardless of the Vegas total. Focus on predicting games that will be higher scoring than expected. This is the key. And by selecting players from lower Vegas totaled games, you will not only be targeting lower-utilized players (another edge) but quite possibly paying less for them as well (as their game is not predicted to be high scoring).

This article should help put Vegas totals into perspective for you this season. It is certainly more than fine to talk about Vegas totals in concert with DFS. Discussing projected points is fine as well. It is useful to be aware what the total is. But don’t use them to finalize your lineup decisions. Don’t select players just because the game has a total of 49 points. It won’t help you. Let your DFS opponents make that mistake. Instead, select the player if you believe the game will exceed expectations.

If you digest the above information, what you hopefully learned will help you immensely this season as you play DFS. Knowing what to ignore and what to focus on is critical, especially when others will be focusing in areas which may be less valuable uses of their time. So take Vegas totals with a grain of salt and instead of focusing on high totals, focus on searching for games most likely to shoot out (over whatever total was set).

Step one is realizing that Vegas totals alone won’t help you win DFS tournaments. Everyone knows of them, and frankly, they are not nearly as helpful as they are cracked up to be. At this point you actually would be more likely to win tournaments (with low % lineups) to fade the common “high total” games and select (cheaper) players from lower totaled games which you believe are likely to go over the Vegas total. Frankly, most of you will do much better just given that knowledge and won’t need to enlist further assistance for edges.

Step two, however, is to find a way to accurately get “on” DFS plays from games that have high odds of going over the Vegas total. Conversely (and potentially just as helpful), fading players who are in games that you believe will go under the Vegas total. You need to accurately predict games that will go over or under the Vegas total. The best way would be to build a model. As I mentioned at the onset, my specialty at SharpFootballAnalysis.com is predicting NFL totals more accurately than Vegas. Each week, I recommend games where the Vegas total was set too low (and we bet the over) or where it was set too high (and we bet the under). I don’t share my “real” number for the total on every single game of the week (which would kill the long term edge), I simply circle and share those specific games which are likely to go over or under, along with a detailed write up explaining exactly why I think the total is wrong. My write up drills down to the specific player matchups I think will cause my recommendation to win.

Whether you take step two and build your own model that is more accurate than what is used by linemakers who set the Vegas total (and thus will give you an edge over your competition) or you obtain that information from an information service like mine or a different resource, you will do better regardless if you take away these key elements to get the next edge on the DFS community:

**1) ** We are NOT going to focus on the Vegas total itself but instead we will predict which games will go over the total and get on those games (and stray from those we predict will go under). While quite obvious, it is vastly underutilized and will give us a major edge over DFS competitors who focus on using players from high totaled games. It will also be a far more profitable way to spend our research time. The correlation between landing a top 3, top 6, top 12 etc. player at each position is far stronger on games for games that go over (regardless of the set Vegas total) rather than playing only players in games with high Vegas totals.

**2)** We are going to fade what are likely to be higher utilized players in games with high Vegas totals when we don’t expect the game to be as high scoring as forecasted. These players will likely be pricey, with a high ownership percentage.

**3)** We will identify games with average or lower Vegas totals which we believe are set too low, knowing much of our DFS competition will be focused on the higher totaled games. If we are correct and the game does shoot out, the low-ownership players we took will go a long way toward winning tournaments.

**4)** We will use our best judgement to predict which games are most likely to outperform or underperform expectations from a total points scored standpoint in order to accomplish the three points above. In my specific case, I’ll be using the10 year track record from my NFL totals model. You’re welcome to join me or you can attempt to build your own model or you can simply use your own experienced judgement in your forecast. In any case, we’re focused on which games should have seen their Vegas total set higher or lower than the linemakers set them, and we’ll use that to our advantage while others will blindly observe the number itself and take it as gospel.

By making smart choices like those outlined above, we will have a huge edge on the competition this year. Big credit to Evan Silva and Sean Fakete for their initial research on the subject which ultimately led to the basic yet (in many respects) contrarian strategy for Vegas totals moving forward.