SITE NEWS: We are moving all of our site and company news into a single blog for Sports-Reference.com. We'll tag all PFR content, so you can quickly and easily find the content you want.

Also, our existing PFR blog rss feed will be redirected to the new site's feed.

Pro-Football-Reference.com ยป Sports Reference

For more from Chase and Jason, check out their work at Football Perspective and The Big Lead.

Data mining the championship games

Posted by Doug on January 17, 2007

Two weeks ago I whipped up a quick system to forecast the wildcard games. The forecasts I got matched up very closely with the Vegas lines. Last week I applied the same system to the divisional games and got what I thought at the time were less believable results. And they did in fact turn out to be less believable. The system differed significantly from the line on three of the four games and was wrong on all three.

Undeterred, I'll post the championship game projections here. Go back to the first link to remind yourself what these tables mean.


TM YR R OPP W L6 SOS PD H B SIM Res
==================================================
chi 3 0 -0.021 81 1 0
ind 2006 w kan 3 0 0.021 51 1 0 928 W 23- 8
ind 2005 d pit 3 0 -0.025 61 1 1 876 L 18-21
hou 1991 w nyj 3 0 0.071 114 1 0 875 W 17-10
atl 2004 d stl 3 0 -0.046 76 1 1 870 W 47-17
stl 2001 c phi 3 1 0.000 95 1 0 865 W 29-24
nyg 1993 w min 2 0 -0.050 96 1 0 856 W 17-10
ind 2003 w den 2 0 0.000 31 1 0 829 W 41-10
nwe 2006 w nyj 2 0 0.037 127 1 0 795 W 37-16
min 1992 w was 2 1 -0.025 80 1 0 795 L 7-24
chi 1990 w nor 3 -2 -0.021 69 1 0 787 W 16- 6
gnb 2002 w atl 3 1 -0.035 -18 1 0 786 L 7-27
sea 2005 d was 3 0 -0.104 115 1 1 782 W 20-10
dal 1994 d gnb 3 1 -0.033 71 1 1 777 W 35- 9
sea 2004 w stl 1 0 -0.042 71 1 0 769 L 20-27
phi 2001 w tam 2 1 -0.046 91 1 0 765 W 31- 9
WEIGHTED AVERAGE: 73.9 pct chance of victory
PROJECTED SCORE: 24.4-15.1

TM YR R OPP W L6 SOS PD H B SIM Res
==================================================
ind 0 -2 0.004 -81 1 0
phi 1995 w det 0 -2 -0.042 -120 1 0 915 W 58-37
oak 2000 c bal 0 -2 -0.004 12 1 0 898 L 3-16
chi 1991 w dal 0 -2 -0.037 -2 1 0 879 L 13-17
mia 1992 d sdg 0 -2 0.008 -35 1 1 849 W 31- 0
sea 2003 w gnb 0 -2 -0.025 -58 0 0 848 L 27-33
sfo 1997 c gnb 0 -1 0.008 -30 1 0 844 L 10-23
cin 1990 w hou 0 -1 -0.046 -90 1 0 841 W 41-14
tam 1997 w det 1 -2 0.023 -37 1 0 837 W 20-10
sfo 2002 w nyg 0 -1 0.023 -25 1 0 825 W 39-38
sea 2006 w dal 0 0 -0.004 -81 1 0 791 W 21-20
mia 1998 w buf 0 -1 0.062 -11 1 0 771 W 24-17
cin 2005 w pit 0 0 -0.017 -60 1 0 758 L 17-31
car 2003 c phi -1 -2 -0.037 -66 0 0 743 W 14- 3
min 2000 c nyg -1 -2 0.042 -56 0 0 737 L 0-41
tam 2002 c phi 0 -1 0.015 -24 0 0 732 W 27-10
WEIGHTED AVERAGE: 59.5 pct chance of victory
PROJECTED SCORE: 23.3-20.7

The algorithm views the Bears-Saints matchup as a pretty boring one: one team is significantly superior to the other. Most of the time, the superior team wins. Not much to see here. Of course, nobody seems to think the Bears are actually superior to anyone, but the reasons for that --- primarily freakishly inconsistent QB play --- are hard to quantify and build into an algorithm like this one.

The AFC game is an interesting case because the Colts only appear to have one advantage: home field. All the other indicators point to the Patriots being the better team. Given the almost nonexistent home field advantage we've seen in the playoffs when two teams with equal records meet, I was a little surprised to see the algorithm pick the Colts.

This entry was posted on Wednesday, January 17th, 2007 at 6:01 am and is filed under General. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.