Primer on closing line value for NFL betting: Part 3

In the previous installment of this series, I talked about things to avoid when mining for closing line value (CLV). Specifically, the “dead zone” in NFL sides with a spread between -1 and +2. Today, I’m going to show you a couple of different things to look for when searching for CLV, using classification trees as in the last article.

Eliminate favorites

You may recall using this tree to set the stage in the previous piece.

This will help set the table for us today as well. Essentially, we are going to be observing the regions outside of the dead zone we established in Part II. Off to the left, you can see the favorites of at least 1.5 points. I’m going to make this very simple: there is not a ton of CLV to be found here.

The base rate for this group is just 9.4%, and we can’t get to a point of usefulness without shrinking the sample to under 35. It is clear that favorites of at least 1.5 points do not typically see a ton of big CLV movements, and that the situations that do produce CLV are difficult to predict, at least with the inputs I have in this model. 

The Sweet Spots

Fortunately, the other side of the dead zone produces some desirable results. For starters, I observed that for spreads greater than 2.0, there is not a huge difference in results for sides gaining a point of CLV instead of the two points we have been searching for.

A side picking up more CLV covered at a higher percentage, but also greatly reduced our sample. Furthermore, the edge in terms of percent increase to hit over sides without the requisite CLV was fairly similar.

With over twice the sample, we can feel more confident that whatever trends we find will be more meaningful, and such bets are still covering at a profitable rate. Now let’s take a look at some paths to that CLV based on the opening line and key data points about the teams. 

The biggest point of delineation comes from the spread itself. Dogs of at least eight points were 41% likely to receive at least a point of CLV. This is excellent for us because that range is even more profitable for bettors. 

There have been 285 sides since 2007 to open as a dog of at least eight points and absorb at least a point of CLV. Of those, 171 sides went on to cover, good for a 60% win rate. The 424 such dogs who did not see at least a point of CLV covered 47.9% of the time.

That on its own will not make you profitable, but we can move down the tree to find two excellent sweet spots.

If a side has an opening spread of at least eight points AND had at least 387 total yards in their last game, they are 60% likely to see at least a point of CLV based on history. If the side did not make that yardage threshold, but their matchup opponent has a season-long (average) point differential under 1.7, they will still see at least a point of CLV 53% of the time. 

Together, these examples makeup 261 games over the past 13 seasons, and see at least a point of CLV 56.5% of the time. If we want to see how often these sides will cover, it is a mere calculation away.

.565*60 + .435*47.9 = 54.7%

Since you only need to cover 52.4% of the time to be a break-even better, these trends would make you profitable over time by betting opening lines. I’m not sure they are strong enough on their own to make it a solid system, but it should give you strong leans when evaluating the opening line market. Especially if you have other betting systems that are creating profit for you, they could get even stronger when adding these additional filters.