Tracking Line Movement using Adjustments from week 1

[uva3021]

This line movement contest sparked a fascination into what conclusions one can reach about their understanding of the overall NFL market. I think it reasonable to assume one who excels in predicting line movement over the course of an entire season is more than likely a pretty solid NFL handicapper. One year is certainly constrained to the limits of a 50 game sample size, though the margin of error may be low enough to combat the variance of probability with confidence in prediction skills. Not sure this is a new experiment for me, and perhaps many people here. So an ideal sample size has yet to be realized or manufactured.

I think the best way to track this is to use excel and the graphing functionality it provides. So I extracted the week 1 line movement history from Covers.com, using the five teams I predicted would invoke line movement in their direction (Colts, Miami, San Diego, New England, Chicago), and re-adjusted the line to even price. For instance, a line of -3 -129 would translate to something around ~ -5.2. My formula isn’t perfect, and does not really embrace the manner in which key numbers correlate to the asymmetric distribution of vigorish. One sees -3 -125 more often than a -5 -125, essentially because it is harder for money or the market to move the line off of a key number.

At first I considered something to this effect: Finding the fair value percentage of the line and vig, then add the currently set line number (-5.5, 3) to the product of the number and the fair value percentage less 50%.





However, this does not account for the nature of how each line can convey a different scale of adjustment given the juice.
In a similar fashion, I entertained the formula:





With either formula, I wasn’t please with the results. To be commensurate with the various numbers up and down the scale, some derivative of the formulas above used as a function of a logarithmic formula is probably the answer. Maybe even something that naturally reconfigures to all key numbers. I haven’t quite figured it out yet.


Until a better method can be applied, I basically cheated. Each cent is worth one-tenth of a point. A line such as -3 -121 would now equal -4.6. Its intuitively, mathematically, and logically flawed, yet equally flawed for all games that I am tracking based on the line movement picks.
Additionally with the Covers website, the line is from the point of view of the home time. Assuming one chooses a road favorite, one would fancy a move upwards, as opposed to a downward movement. Conversely, a home favorite would show a move downwards on the graph indicating an increase in the line number.


On the graph, I marked my line with a circle, and compared to the closer by drawing a line from one point to the other. The y-axis is the spread and the numbers on the bottom betray the line movement number sequence. My excel sheet has the consequent dates yet they are not shown here. But my lines posted here were from the Tuesday leading up to the week 1 games, so from that one should be able to judge to a certain degree when each point of movement took place.