The spread (listed in parentheses) is shown next to S&P+’s pick for each game. When S&P+ predicts a push (a tie with Vegas, basically), I’m listing the pick on the side that S&P+ would’ve picked, if teams could score in decimals.
This document breaks games (and S&P+’s season performance to date) out into their decimal glory. There wasn’t just a ton of glory for S&P+ in Week 7. More on that at the bottom of the post.
- No. 2 Penn State (-9.5) 29, No. 19 Michigan 18 (Saturday, 7:30 PM ET, ABC)
- No. 13 Notre Dame (-3.5) 32, No. 11 USC 26 (Saturday, 7:30 PM ET, NBC)
While Jim Harbaugh attempts to scrounge up enough free points to make Michigan-Penn State a game, Notre Dame and USC are playing in a Playoff eliminator of sorts. Each has one loss, and while the Irish have looked more dominant at times, USC is certainly going to be comfortable if this game goes the distance — the Trojans have already played in games decided by one, three, three, and 10 points.
- No. 1 Alabama 39, Tennessee (+36) 11 (Saturday, 3:30 PM ET, CBS)
- No. 4 TCU 46, Kansas (+39) 12 (Saturday, 8:00 PM ET, Fox)
- No. 5 Wisconsin (-24) 39, Maryland 14 (Saturday, 12:00 PM ET, Fox)
- No. 8 Miami (-16.5) 38, Syracuse 20 (Saturday, 3:30 PM ET, ESPN)
- No. 9 Oklahoma 39, Kansas State (+14) 27 (Saturday, 4:00 PM ET, Fox)
- No. 10 Oklahoma State (-7) 38, Texas 26 (Saturday, 12:00 PM ET, ABC)
- No. 14 Virginia Tech 35, North Carolina (+21) 17 (Saturday, 3:30 PM ET, ESPN2)
- No. 15 Washington State (-10.5) 34, Colorado 22 (Saturday, 10:45 PM ET, ESPN)
- No. 16 South Florida (-11.5) 34, Tulane 22 (Saturday, 7:00 PM ET, ESPN2)
- No. 18 Michigan State (-7) 27, Indiana 19 (Saturday, 3:30 PM ET, ABC)
- No. 20 Central Florida (-7) 40, Navy 28 (Saturday, 3:30 PM ET, CBSSN)
- No. 21 Auburn 33, Arkansas (+15.5) 20 (Saturday, 7:30 PM ET, SECN)
- No. 23 West Virginia 39, Baylor (+9.5) 30 (Saturday, 8:00 PM ET, FS2)
- No. 24 LSU 31, Ole Miss (+7) 28 (Saturday, 7:15 PM ET, ESPN)
- Houston (-3) 33, No. 25 Memphis 28 (Thursday, 8:00 PM ET, ESPN)
S&P+ isn’t quite sold on TCU just yet, but my friend Brian Fremeau’s FEI loves the Horned Frogs. Saturday night is a TCU showcase on Fox — we’ll see if we get a better read for Gary Patterson’s squad against a lifeless Kansas.
- Arizona (-3) 33, California 26 (Saturday, 8:00 PM ET, Pac-12)
- Duke 32, Pittsburgh (+8.5) 24 (Saturday, 12:20 PM ET, ACCN)
- Florida State 30, Louisville (+6.5) 28 (Saturday, 12:00 PM ET, ESPN)
- Georgia Tech 28, Wake Forest (+6) 24 (Saturday, 7:30 PM ET, ESPNU)
- Iowa (-1.5) 27, Northwestern 24 (Saturday, 12:00 PM ET, ESPN2)
- Minnesota 31, Illinois (+14) 17 (Saturday, 3:30 PM ET, BTN)
- Mississippi State (-11.5) 34, Kentucky 20 (Saturday, 4:00 PM ET, SECN)
- Oregon 36, UCLA (-6.5) 33 (Saturday, 4:00 PM ET, Pac-12)
- Purdue 26, Rutgers (+9.5) 22 (Saturday, 12:00 PM ET, BTN)
- Texas Tech 33, Iowa State (+7) 31 (Saturday, 12:00 PM ET, FS1)
- Utah (-9) 36, Arizona State 24 (Saturday, 3:30 PM ET, FS1)
- Virginia (-7) 34, Boston College 16 (Saturday, 12:30 PM ET, ACCN)
Hey, you know who could move to within a half-game of the Pac-12 South lead with a Saturday night win? This guy:
(Technically, so could Arizona State and Todd Graham, but the Sun Devils are not nearly as likely to beat Utah as Arizona is to beat Cal.)
- Air Force (-6.5) 36, Nevada 27 (Friday, 9:30 PM ET, CBSSN)
- Appalachian State 38, Coastal Carolina (+24) 19 (Saturday, 3:30 PM ET, ESPN3)
- Arkansas State (-12.5) 39, UL-Lafayette 25 (Thursday, 7:30 PM ET, ESPNU)
- Army (-6.5) 32, Temple 23 (Saturday, 12:00 PM ET, CBSSN)
- Ball State (+2.5) 29, Central Michigan 26 (Saturday, 3:00 PM ET, ESPN3)
- Boise State 29, Wyoming (+14.5) 18 (Saturday, 10:15 PM ET, ESPN2)
- BYU 31, East Carolina (+5.5) 26 (Saturday, 7:00 PM ET, CBSSN)
- Colorado State 36, New Mexico (+7.5) 29 (Friday, 10:15 PM ET, ESPN2)
- Eastern Michigan (+3) 27, Western Michigan 27 (Saturday, 2:00 PM ET, ESPN3)
- Florida Atlantic 34, North Texas (+3.5) 31 (Saturday, 5:00 PM ET, ESPN3)
- Marshall 25, Middle Tennessee (+2.5) 23 (Friday, 7:00 PM ET, ESPN2)
- Massachusetts (-7) 32, Georgia Southern 21 (Saturday, 3:30 PM ET, NESN+)
- Miami (Ohio) (-3) 30, Buffalo 25 (Saturday, 2:30 PM ET, ESPN3)
- Missouri 31, Idaho (+14) 28 (Saturday, 12:00 PM ET, SECN)
- Northern Illinois (-14) 33, Bowling Green 19 (Saturday, 2:00 PM ET, ESPN3)
- Ohio (-19.5) 37, Kent State 13 (Saturday, 2:00 PM ET, ESPN3)
- San Diego State 28, Fresno State (+7.5) 25 (Saturday, 10:30 PM ET, CBSSN)
- SMU (-7.5) 37, Cincinnati 24 (Saturday, 4:00 PM ET, ESPNU)
- South Alabama (-4.5) 33, UL-Monroe 28 (Saturday, 5:00 PM ET, ESPN3)
- Southern Miss (+2.5) 33, Louisiana Tech 24 (Saturday, 7:00 PM ET, Stadium)
- Toledo (-15.5) 38, Akron 21 (Saturday, 12:00 PM ET, ESPN3)
- Troy 28, Georgia State (+7.5) 22 (Saturday, 2:00 PM ET, ESPN3)
- Tulsa 36, Connecticut (+6) 33 (Saturday, 12:00 PM ET, ESPNU)
- UAB (-7.5) 33, Charlotte 23 (Saturday, 6:30 PM ET, beIN)
- UNLV 29, Utah State (+4) 29 (Saturday, 6:00 PM ET, MWC Video)
- UTSA (-20.5) 43, Rice 13 (Saturday, 7:00 PM ET, Facebook)
- Western Kentucky (-9.5) 32, Old Dominion 20 (Friday, 6:00 PM ET, CBSSN)
My numbers hate my team.
Week 7 was S&P+’s first truly bad week. It had been below .500 against the spread a couple of times earlier in the year, but that appeared to be a product of Vegas being dialed in. The absolute error for S&P+ picks (basically, how much you miss the point margin on average) was still very much on the healthy side.
Last week: less healthy. S&P+ went 25-33-1 against the spread (43%), and the absolute error jumped from a perfectly acceptable 12.3 to 13.5. (That doesn’t sound like a lot, but it very much is.)
Any time I have a week like that, I investigate, and one thing quickly stood out: the preseason projections that were so effective in September are now holding S&P+ back.
My general plan has always been to slowly phase out the preseason numbers over the course of seven weeks — as in, after one week, everybody’s preseason projections carry X percent weight, after two weeks, Y percent, etc. After seven weeks, the 2017 data stands on its own.
This week, I changed that up slightly: I made it team specific. Instead of phasing out the numbers for everybody week to week, I made it so that, if you’ve only played four games while someone else has played six, your preseason projections count for more than theirs, because we don’t have as much of a sample on you.
This made sense to me. It also backfired. For teams that entered Week 7 having played only four games, S&P+ went 0-3-1 (13 percent) against the spread. For games that featured teams with 10 or fewer combined games played, S&P+ went 9-15-1 (38 percent) with a much higher absolute error than other games.
So basically, the new plan, as good as it sounded, didn’t work very well.
The good news is that there are no Week 8 games pitting teams that have combined for 10 or fewer games played, and there are only six between teams with a combined 11. I smell turnaround.
Each year at Football Study Hall, I have posted weekly S&P+ picks as a way of affirming the ratings’ validity. I use my S&P+ system as a complement to most of my analysis, and the picks are a way of showing it generally knows what it’s talking about.
S&P+ tends to hit between 50 (meh) and 54 percent (great) against the spread from year to year. It isn’t always the single best performer, but it holds its own. And beyond picks, it goes deeper than any other set of college football analytics on the market. You can go into granular detail regarding team strengths and weaknesses in a way that no other set of ratings allows. (See the annual team statistical profiles as proof.)
This year, I will be posting the weekly S&P+ picks at SB Nation instead of FSH.
Because I like to experiment, however, I won’t just be posting the official S&P+ picks. Go to this Google doc, and you will find three sets of picks: S&P+, F/+ (combined ratings from S&P+ and Brian Fremeau’s FEI), and what I’m calling an adjusted S&P+ pick, in which I attempt to account for two additional factors: week of play and type of game.
- Week of play: I’m finding that there are cycles to scoring averages throughout a given season. Week 1 typically falls below the season scoring average, while the final weeks of the season tend to perk up in the scoring department. This adjusted projection will take this into account.
- Game type: It shouldn’t be surprising to learn that the standard deviation of possible results in a game against FCS competition, for instance, is different than that of a conference game. This projection will also adjust for different types of games. This will mean some pretty extreme projections, but we’ll see how it performs.