Making a line with the Off/Def Line ratings
The following research is based on the new line-play rating system discussed in the Offensive/Defensive Line Ratings article in the main research section.
Can the offensive line and defensive line be isolated and used successfully to predict future outcomes? There are many people who would have you believe that games in the NFL are mostly won and lost by what happens "in the trenches."
Still, it's a challenge to try to put it all together into a workable methodology for finding overlays. We decided to make a matchup for starters based on the overall line rating (rather than just the rushing rating, passing rating, or some other combination of the two) and by comparing projected offensive line ratings for the two teams, we assume that every 0.2 of rating difference is worth one point of projected margin in the final score.
To put it another way, if Team A is projected to have a 6.4 O-line overall rating, and Team B is projected to have a 5.4 O-Line overall rating, then on a neutral field we would expect Team A to win by 5 points. Of course, with the NFL you need to add in a home field advantage, which for convenience sake in this test we set at three points.
Other details are: we used a rolling season-to-date average for our data set, and all tests include all games from week 5 through the conference championship playoff events.
NFL 2002 Season, Overall Projected Line Ratings
| Overlay |
0 - 1.5 |
2 - 3.5 |
4 - 5.5 |
6 - 7.5 |
8+ Pts |
ALL |
| HomeFavs |
19 - 21 |
12 - 2 |
1 - 4 |
0 - 2 |
0 - 0 |
32 - 29 |
| HomeDogs |
13 - 8 |
6 - 8 |
3 - 6 |
3 - 1 |
3 - 1 |
28 - 24 |
| AwayFavs |
5 - 5 |
1 - 4 |
0 - 0 |
0 - 0 |
0 - 0 |
6 - 9 |
| AwayDogs |
17 - 17 |
12 - 10 |
11 - 3 |
1 - 2 |
0 - 2 |
41 - 34 |
| Favorites |
24 - 26 |
13 - 6 |
1 - 4 |
0 - 2 |
0 - 0 |
38 - 38 |
| Underdogs |
30 - 25 |
18 - 18 |
14 - 9 |
4 - 3 |
3 - 3 |
69 - 58 |
| Home Teams |
32 - 29 |
18 - 10 |
4 - 10 |
3 - 3 |
3 - 1 |
60 - 53 |
| Away Teams |
22 - 22 |
13 - 14 |
11 - 3 |
1 - 2 |
0 - 2 |
47 - 43 |
| ALL PICKS |
54 - 51 |
31 - 24 |
15 - 13 |
4 - 5 |
3 - 3 |
107 - 96 |
| WIN % |
51 % |
56 % |
54 % |
44 % |
50 % |
53 % |
| Spread Range |
10+ points |
5 to 9.5 |
0 to 4.5 |
| Favorites |
2 - 5 |
12 - 9 |
24 - 24 |
| Underdogs |
2 - 3 |
32 - 23 |
35 - 32 |
Comments: not a tremendous amount of predictive value is demomonstrated on just this one season, and the balance is arguably askew since the underdog is selected almost two-thirds of the time. We can of course tweak the values (from our base of 1 pt for every 0.2), and this can improve the results slightly, but that's just backfitting data to a sample. Of note is that the projected lines from the O-Line scheme are very close in general to the actual lines, with only 15 games out of 203 (7%) falling six points or more from the actual spread, and over half the games landing within a point and a half of the true line.
The next idea was to instead use the pass blocking number only, excluding the run blocking, and the results for this run were eye-opening:
NFL 2002 Season, O-Line PASSING Ratings
| Overlay |
0 - 1.5 |
2 - 3.5 |
4 - 5.5 |
6 - 7.5 |
8+ Pts |
ALL |
| HomeFavs |
18 - 15 |
11 - 9 |
8 - 4 |
1 - 1 |
0 - 0 |
38 - 29 |
| HomeDogs |
7 - 3 |
6 - 8 |
3 - 4 |
5 - 0 |
1 - 2 |
22 - 17 |
| AwayFavs |
10 - 9 |
3 - 6 |
0 - 0 |
0 - 0 |
0 - 0 |
13 - 15 |
| AwayDogs |
19 - 11 |
14 - 11 |
5 - 3 |
2 - 1 |
1 - 2 |
41 - 28 |
| Favorites |
28 - 24 |
14 - 15 |
8 - 4 |
1 - 1 |
0 - 0 |
51 - 44 |
| Underdogs |
26 - 14 |
20 - 19 |
8 - 7 |
7 - 1 |
2 - 4 |
63 - 45 |
| Home Teams |
25 - 18 |
17 - 17 |
11 - 8 |
6 - 1 |
1 - 2 |
60 - 46 |
| Away Teams |
29 - 20 |
17 - 17 |
5 - 3 |
2 - 1 |
1 - 2 |
54 - 43 |
| ALL PICKS |
54 - 38 |
34 - 34 |
16 - 11 |
8 - 2 |
2 - 4 |
114 - 89 |
| WIN % |
59 % |
50 % |
59 % |
80 % |
33 % |
56 % |
| Spread Range |
10+ points |
5 to 9.5 |
0 to 4.5 |
| Favorites |
2 - 5 |
24 - 15 |
25 - 24 |
| Underdogs |
2 - 3 |
26 - 11 |
35 - 31 |
Comments: the balance is better (more favorites tabbed), and the results improve dramatically (close to 56% playing every game from week five on). Perhaps most interesting is that in games with favorites of 5 to 9.5 points, the "Passing Ratings" clocked a fabulous 50-26 (65%) mark.
Additional tests were run using the Rushing only ratings (no predictive value seen) and using different timeframes for the stats (last X games, using 1 to 5 -- season to date was superior). Future tests might include home/away or field surface data sets, but since we only have the one year, it's rash to get too intricate.
We have in past seasons created some semblance of line ratings, but we are less confident that they are as good a measure as what we employed for 2002. Nevertheless, the next step will be to look at some past seasons for comparison. Another area for exploration is clearly the over/under avenues of wagering.
It's early, but there's signs that including lineplay projections as part of an overall handicapping analysis could be worthwhile.
| 
