Saturday, January 07, 2006

Sports: NFL and margin of victory

Is the margin of victory (MOV) dependent on the total? Is the probablity of a teaser winning dependent on the total? These two questions are related. (This post serves as the essay from the list that I started in 2005: #5. Teasers: distributions of scores based on the level of the total)

Intuitive reasoning seems to say yes. The higher a total, the larger the distribution, and the higher the MOV expected. I ran some numbers in games over the past 15 years. The results do not coincide with this intuitive reasoning. In fact, the results show that the total is not a factor at all when it comes to MOV. Maybe the reasoning is that when a total is high, it can also mean the team that is behind has a higher chance of scoring...whereas if the total is low, if a team happens to get a large lead, it is tougher for the team behind to catch up. No matter the reasoning, here are some results:

I ran games from 1989 through early 2005. I split the MOV into 4 categories: Totals below 37; 37.5 to 40.5; 41 to 44; and 44.5 and above. I ran every game that had a pointspread of 4.5 or less (so it ranges from the Home Team being a 4.5 point favorite to a 4.5 point dog). I ran the MOVs of 7.5, 13.5, and 20.5 points to see what percentage of those games were greater. Apologies if the formatting makes the numbers difficult to read.

Lines of 4.5 or less
MOV Samples 7.5 / 13.5 / 20.5
1-37 602 50.7 / 30.7 / 12.6
37.5-40.5 609 50.1 / 31.4 / 14.1
41-44 489 49.1 / 29.7 / 13.7
44.5-100 303 50.5 / 31.7 / 13.5

Those numbers look pretty much the same regardless of the total.

The I looked how teasers did depending on the total. I looked at them in clumps and not each game indivually. For example, I lumped all games where the home team was a 7 to 8.5 point favorite and looked at how they did as 1.5 point faves, which is roughly the level they would be teased down to on average. This may lose a bit of specific accuracy on the indivudal games, but since the lines aren't perfect anyway, it should work nice. When I say "the lines aren't perfect", I mean that even if a game is lined at 8, it is possible to find a place with a line of -7.5 -120 and tease it down to -1.5 without the -120 juice hurting your teaser.

Home team -7 to -8.5. Winning percentage versus a line of -1.5
MOV Samples Winning percentage at -1.5
1-37 125 72.8%
37.5-40.5 144 71.5%
41-44 116 75.9%
44.5-100 79 75.9%

Although the winnign percentage is a bit higher in the higher totals, I don't think it is all that different given the smaller sample sizes....especially when compared to when the away team is the -7 to -8.5 point favorite:

Away team -7 to -8.5. Winning percentage versus a line of -1.5
MOV Samples Winning percentage at -1.5
1-37 54 77.8%
37.5-40.5 37 64.9%
41-44 34 58.8%
44.5-100 41 65.9%

The sample sizes are even smaller, and it has the best winning percentage in the low totals. I combination of these two tables make me think that it is safe to assume that the total level doesn't do anything for teasers on favorites down from -7 to -8.5

Now on to how the teasers from +1 to +2.5 up to +8.5 do.

Away team +1 to +2.5 dogs. Winning percentage versus a line of +7.5
MOV Samples Winning percentage at +7.5
1-37 99 73.7%
37.5-40.5 130 72.3%
41-44 94 63.8%
44.5-100 54 74.1%

There is not much of a patter there. 41-44 totals have a much smaller winning percentage, but then it spikes the highest when the total is the highest at 44.5 and above. So from that I would conclude the total doesn't have much of an effect on the teaser's winning percentage

Home team +1 to +2.5 dogs. Winning percentage versus a line of +7.5
MOV Samples Winning percentage at +7.5
1-37 101 67.3%
37.5-40.5 91 84.6%
41-44 59 84.7%
44.5-100 36 58.3%

Again there isn't much of a patter as far as saying that higher totals may lead to lower teaser win percentages.

The question is: Does the winning percenteage of teasers change depending on the total?

The ansewr is: No. There is no historical evidence to say it does at all.

No comments: