TL;DR: CLV can be a useful alternative measurement for performance, but is ultimately a flawed metric
Purpose of CLV

The primary purpose of CLV is an alternative measurement of performance. The theory is that if you’re getting enough CLV to cover the vig, you should be a winner in the long term. Many “pros” claim that it's best to benchmark performance based on CLV rather than actual outcomes. Sportsbooks can also use it as a measurement to assess whether a sports bettor is a “sharp” or a “square”, sometimes limiting or even outright banning bettors who consistently beat CLV. This assertion relies heavily on the efficient market hypothesis.
Efficient Market Hypothesis
Without giving you a financial theory history lesson, very simply the efficient market hypothesis (EMH) states that the price of an asset reflects all known information and that consistent alpha generation is impossible. Sports betting translation: the only way to bet profitably is to generate CLV and it’s impossible to generate +EV if you only bet right before the game starts. If you bet the Closing Line you should expect to lose an amount equal to the vig in the long-term.
Quite simply – this is bullshit.
Various forms of EMH may apply to liquid financial markets, but I’m going to make the argument that while CLV is useful, the Closing Line is far from efficient.
Is the Market Efficient?
Market efficiency is often characterized as having the following attributes:
1. Immediate absorption of new information
2. Important information is freely available to all participants
3. A large number of rational, profit maximizing market participants
Let’s review these assertions one-by-one.
1. Immediate Absorption of New Information
In an efficient market, the only thing that moves the price of an asset is new information. If this were true, we should be able to identify long periods of static lines, as no new information has been revealed.
Let’s check out a recent example of how reactive the markets are to new information:
On January 11, 2020 the OKC Thunder hosted the LA Lakers. Around 1:30pm ET, news broke that LeBron would miss the game. Naturally, that injury announcement had a large impact on the odds for both teams. A time series plot of the Thunder’s breakeven win probability is shown below.
Time Series of an OKC LAL game win probability
The lines almost immediately improved the Thunder’s breakeven win % from ~50% to ~65%. Without giving a chance for the lines to reach a new equilibrium, another bombshell was dropped at 1:54pm ET: Anthony Davis was questionable. The lines continued to move in the Thunder’s direction for the next hour or so before seemingly reaching an equilibrium a little after 3pm ET.
When it was finally announced that AD was downgraded to Out around 45 minutes before tip, the line began to further trend toward OKC.
So how should we judge these movements? Did the market immediately factor in new information?
Although the market reacted fairly well, there was still some opportunity to get a bet in before the market reached a new equilibrium, particularly with regard to the AD news. I would say that the market may not have fully reacted immediately, but this isn’t enough evidence to disprove the EMH.
We are 0 for 1.
2. Important Information is Freely Available to All Participants
Does everyone have access to the same information? Certainly not everyone would agree with me, but I generally believe that most sports information is freely available these days. The barrier to information is lower than it’s ever been. People use information in different ways, to give them certain edges, but I don’t think that information asymmetry is a reason to disprove EMH.
We are now 0 for 2...
3. A Large Number of Rational, Profit Maximizing Market Participants
I think we can all agree that the drunk guy parlaying the Gatorade color and coin flip at the Super Bowl might not be rational or profit maximizing.
The vast majority of sports bettors aren’t profit maximizers, but utility maximizers. Sports betting offers a form of exhilaration and entertainment that can’t be found in other places. A lot of that excitement manifests itself in poor-EV-yet-thrilling wagers (such as parlays, teasers and futures) that sportsbooks happily offer you.
Just how much are non-profit maximizing behaviors costing sports bettors? To answer that, let’s take a peak at the Nevada’s annual sports betting report. In 2019, sportsbooks in Nevada took $5.3 billion in wagers and held $329 million, representing a hold of 6.2%. Previously we discussed how standard -110 odds gave sportsbooks a hold of 4.5%, which we could chisel away at pretty easily with some basic line shopping. Thus, if market participants we’re truly profit maximizers, we’d expect a hold significantly less than 6.2%.
OK – so finally we have some evidence that the EMH might not hold. Let’s see if we can test it with some data.
Testing Weak Form Efficiency
The three forms of market efficiency are Strong Form, Semi-Strong Form, and Weak Form. The Strong Form assumes that all information (private and public) is baked into the market. The Semi-Strong Form assumes that all public information is baked into the market price of an asset. The Weak Form states that historical prices cannot be used to predict future prices.
If we can prove that the weakest form of the EMH can be disproved, we can disregard the EMH.
Straight from Morningstar:
“The weak form of EMH assumes that current stock prices fully reflect all currently available security market information. It contends that past price and volume data have no relationship with the future direction of security prices. It concludes that excess returns cannot be achieved using technical analysis.”
MLB Moneyline Movements
Let’s go ahead and use MLB ML data from the 2015-2018 seasons to see if we can predict the direction of the closing line, and therefore generate theoretical value (CLV) by beating the closing line.
We gathered the Closing Line as well as the line 2-hours to close[1] (T-2) to see if we can recognize any patterns. We can then test the statistical significance of those patterns to give us a sense of whether they have any merit.
The traditional school of thought is that if you’re betting favorites, it’s best to bet them early. If a dog, wait until close to gametime. Does this hold merit?
The first thing we can do is test the average deviation of prices from a 50/50 probability. Closing Lines had an average deviation of 44 cents, while T-2 had an average deviation of 42 cents over 9,813 games in our sample. If we look at the distribution, we see that there are more games with an average deviation greater of 100 or more at close than at T-2.
Average Deviation
Yes, the curves look similar. But if we focus on the difference between the two, we can identify a more significant pattern.
Difference Between Close and T-2
What the above shows is that there are more “close” games at T-2 and more “mismatches” at Close. Huh? How can that be?
Answer: lines must move toward the favorite from T-2 to Close.
Let’s dive a little further and focus on games that have a significant favorite.
We pulled out games that have an underdog of +180 or greater at T-2. In total we had 1,208 games. Of those 1,208 games, 657 (54%) had line movement toward the favorite, 404 (33%) had line movement toward the underdog, and 147(12%) did not have any movement. The average movement of the favorite was -3.4 cents, from -224.0 to -227.2.
Visually, we can look at the distributions of movement below.
Line Movement Distribution
Clearly, the data suggests a movement toward the favorites in the last two hours, suggesting that we can capture positive CLV simply by betting favorites 2 hours prior to first pitch. This “strategy” violates weak form EMH, which states that past prices have no relationship with future price movements.
If this isn’t enough evidence to disregard the EMH, I pose you this: are the MLB markets systemically mispricing favorites two hours prior to first pitch, only to correct this mispricing from T-2 to Close?
I find it hard to believe.
Optimizing for CLV vs Optimizing for Profit
The evidence above provided a theoretically argument why the EHM can be largely disregarded and therefore CLV should not be the target that bettors are optimizing for.
A more practical reason why CLV should not be the target: because CLV is fairly simple to measure, it is the primary way that sportsbooks designate who is sharp and who is square. With so many sportsbooks practicing the strategy of limiting or banning sharp bettors, it’s probably not ideal to optimize for a strategy that 1) rests heavily on the assumption of an efficient market and 2) firmly puts you on the radar of sportsbooks.
[1] This is the line available two-hours prior to first pitch