2018 Week Five - CFB Analytics vs. AP Poll

Week five is in the books, and what a week it was. The Buckeyes got their biggest win of the season, so far, and a couple of unbeatens lost their first games. There are currently only 14 undefeated teams in FBS, and two of them aren’t even ranked in the AP Top 25. More importantly, week six is upon us. Next week we should see nearly all of the preseason ratings bias disappear from the analytic models. We should also expect some last minute jumps (or falls) from teams when this happens.

Thought I'd throw in Mike Leach's press conference from Oct 1, after WSU beat Utah without rushing on a single play. His comments in the first five minutes on offensive balance are really interesting. And his love for the Wishbone offense surprised me.

Passing yards have a way of spending just as well as rushing yards do...I want all the positions to touch [the ball]. There's nothing balanced about fifty percent rush and fifty percent pass, 'cause that's fifty percent stupid.

There's nothing fifty percent run about the wishbone, they're like 95 percent run, but everybody touched the ball. That's why it's one of the greatest offenses devised...If you think about it, nobody's ever really stopped that offense. They get bored with it, they get tired with it, but nobody's really stopped it.

The whole video, like all his interviews, is worth watching.

Week 5 F/+ Top 40

Week 5 Observations

F/+ Top 10: There’s really not much to say here. TTUN dropped out with a come from behind win over Northwestern and LSU climbed in after another great showing. The only noteworthy observation from this week’s F/+ Top 10 is it’s completely opposite take from the AP Poll on LSU and Washington. It would be easy to point to the common opponent of Auburn and apply the transitive property, but we know that’s just lazy. LSU has the opportunity to make a statement this week in the Swamp. UW won’t get that opportunity until the following week when the Huskies travel to Oregon.

Overrated: LSU moved onto the bubble this week and are now within five places of their F/+. Texas Tech came off this list by falling out of the AP Top 25 altogether. Stanford was overrated before their loss to ND, but dropped in both the polls and F/+, and the Cardinal remain on the list.

AP Top 25 Teams that are currently too high in the polls, based on their current F/+ rating, include Stanford (+6), WVU (+7), UCF (+7), Texas (+8), Texas Tech (+10), and Va Tech (+6).

Our ‘obscenely overrated’ list continues to grow, because AP Voters don’t know what to make of teams 21-43. As a result, we’re seeing some “flavors of the week” (Syracuse & Colorado) and some “well I had them in the Top 25 last week, and I’ll look ridiculous if I drop them completely, so I’m still throwing this team a bone” (Oregon, Cal & BYU) type voting.

Joining #18 Oregon (+24), #21 Colorado (+36), #34 Cal (+33), and #37 BYU (+24) are #28 Syracuse (+19), #34 Maryland (+18), #48 Cincinnati (+19), #54 South Florida (+27), and newcomer #42 Hawaii which just barely ranks in the Top 100 of teams in F/+ rating at #99 (that’s +57 if you don’t want to do the math).

Underrated: AP #10 Washington moved to the bubble this week at F/+ #5. Likewise with Texas A&M which is ranked AP #31 and F/+ #26, and NC State which got some AP love and moved up to #23, with an F/+ of #18.

TCU fell in F/+ and rose in the AP poll, taking them off this list. South Carolina dropped out of the AP altogether so it’s hard to call them underrated anymore; although at F/+ #34, the Gamecocks may find their way back into the AP Poll.

MSU (-7), Miss State (-14), and Oklahoma State (-10), remain on our underrated watchlist. Joining them this week are Iowa (-8), Appalachian State (-15) and Duke (-13). It’s also noteworthy that USC, Fresno State, Texas Tech, South Carolina, North Texas and Houston are all in the F/+ Top 40 in Week Five, but did not receive any AP votes. Based on their statistical performance to date, those teams are all poised to potentially jump into the polls.

Inside the Numbers: This is a great time to discuss the effects of winning and losing in analytics. Week four featured a pair of games between AP Top 10 teams (OSU-PSU & ND-Stanford), which also happened to be closely ranked in analytics. However, the results of those games, and how the teams played, were dramatically different.

Notre Dame produced 550 yards of offense, scored on six drives (5 TD, 1 FG), punted four times and had zero turnovers. The Irish went 9 of 17 on third down, and they averaged 4.9 yards per rush (0.5 above season average) and 8.4 yards per pass attempt (0.3 above season average).
Stanford, meanwhile, was held to 229 yards, only scored on three possessions (2 TD, 1 FG), punted eight times, and threw one interception. The Cardinal went 2 of 13 on third down, and they averaged 2.3 yards per rush (1 below season average) and 6.4 yards per pass attempt (2 below season average).

TL/DR: Notre Dame performed slightly above their season average against a good team; Stanford performed well below theirs.
End result: ND blew out Stanford and moved up slightly in F/+, while Stanford fell five places in F/+, and the gap between their raw numbers widened.

Ohio State produced 389 yards of offense, scored on four drives (all TDs), punted nine times, missed one FG and threw one interception. The Buckeyes went 4 of 17 on third down, and they averaged 3.2 yards per rush (1.8 below season average) and 6.9 yards per pass attempt (2.5 below average).
Penn State, managed to gain 492 yards on offense, scored on five drives (3 TD, 2 FG), punted eight times, missed one FG, lost one fumble, and failed to convert a pair of fourth down conversions including one inside FG range. The Nittany Lions went 3 of 17 on third down, and they averaged 4.7 yards per rush (1.3 below season average) and 8.9 yards per pass attempt (0.2 above season average).

TL/DR: OSU performed well below their season average in offensive yards against a good team, but when they moved the ball it turned into touchdowns; Penn State performed below their average rushing and slightly better than their average passing, but they failed to turn yards into points on multiple occasions.
End result: OSU narrowly beat PSU. Both teams moved down slightly in analytics and the gap between them in raw numbers narrowed significantly.

Way Too Early CFP Prediction

In order: ‘Bama, OSU, Oklahoma, and Clemson. Wildcard: Notre Dame.

Week 6 AP “Top 25” Match-ups

Last week was chalk, and my underdog pick of Stanford turned out to be a real dog. I’ll take a 4-1 showing in week five. WVU covered, and damn if Texas Tech didn’t score only enough points to land right on the Over/Under. ND, UW and Oregon all covered by double digits, which didn’t surprise me.

Only three AP Top 25 matchups, and they’re lopsided this week. I added a few close lines to make it interesting.

#7 Oklahoma vs. #19 Texas – Texas is schizophrenic and OU looks incredible on offense. Vegas has the line at OU -8.5, but S&P+ looks more like 11.5. I think the Sooners run/pass all over the Longhorns and win by two TDs. It could be worse if the Texas that faced Maryland, Tulsa or Kansas State shows up.

#5 LSU @ #22 Florida – This should tell us a lot about the SEC. LSU is the Vegas favorite giving 2.5 points. S&P+ favors the Tigers by less than one point on a neutral field. FEI has a wider margin between the teams. What I think is noteworthy is how much Florida has moved up from their preseason S&P+ projection (7.8 to 14.9), whereas LSU has stayed within two points of their raw S&P+ (14.2 to 15.5). Based on this game being played in the Swamp, one sixth of the analytics still working on preseason data, and LSU possibly looking ahead to UGA, I’m calling for a Gator upset.

#6 Notre Dame @ #24 Virginia Tech – Notre Dame dropped each week in F/+ with Wimbush as their QB. They’ve risen both weeks with Book starting. That’s no coincidence. Book makes the Irish a more potent team. The line is ND -6, but S&P+ has it at 11. Both teams are trending up, but I think ND’s analytical data is skewed lower because of three weeks of Wimbush. I’m taking Notre Dame to cover on the road (Sorry HokiePro, if you’re reading this).

#21 Colorado vs. #40 Arizona State – This is a tough one. Vegas has the Buffs as a 2.5 point favorite at home. S&P+ and FEI don’t help clear it up; not only are they reversed (S&P+ has the Buffs higher than the Sun Devils, but vice versa with FEI), but the spread between the teams in much larger in FEI than in S&P+. F/+ has ASU at 31.7% and CU at -3.9%. I have no idea how to convert those into points. Here’s what I do know: CU barely escaped Lincoln with a one-score win over a bad Nebraska team and they have three home wins against bad teams (CSU, UNH and UCLA); ASU beat UTSA, Sparty and Ore State at home, and they lost two road games by a TD each to San Diego State and Washington. I’m going to ride the Sun Devils to another upset, this time on the road in Boulder.

#13 Kentucky @ #31 Texas A&M – A 7:00 pm kick at Kyle Field, and guess who’s favored? The Aggies, by six. This is interesting, because both FEI and S&P+ favor the visiting Wildcats. My take on this is if you think Kentucky is within two points of Clemson then by all means pick ‘em. I love what Stoops has done in Lexington, and their road win in the Swamp was epic. I just don’t believe he’ll be able to outcoach Fisher or slow down A&M’s offense. I think the Aggies win, but they may not cover.

#8 Auburn @ #37 Miss State – This looks completely lopsided from the rankings, but the Bulldogs are better than their AP ranking according to F/+. Auburn is only a 3.5 favorite in Starksville, and S&P+ has less than three points separating these two on a neutral field. This is where I really like to look at trends. Auburn has barely deviated one point from their week two S&P+ rating. Miss State has fallen every week, meaning they have yet to perform to their preseason potential. I would take Auburn to win this and cover the spread.