PDA

View Full Version : Dork Stats/Polls, Football Edition, 2014 Season



loran16
10-20-2014, 11:37 PM
Meant to start this last week, but oh well. Some of you may recall I kept track of this last year - basically the point of this thread is to keep track of the advanced computer numbers and statistics for the Duke team, which should give us a better picture of how good the team is than human polls. Computers and stats do not have an SEC bias, for example. On the other hand, the best computers and stats actually don't have a bias in favor of WINNING - it's about outplaying opponents, not necessarily coming out with the win. These measures also account for competition, so Duke's super weak competition early on won't impress the computers, and may hurt them since vs at least one team (Troy), Duke didn't truly blow them out.

Note I won't be talking about polls and numbers used by the selection committee in this post: those have stupid restrictions like no margin of victory and thus are generally useless. So let's talk about useful ones:

Sagarin:

You all probably know Sagarin, he does every sport. Like in the other sports, we're interested in the predictor poll, which is meant to best predict future games. By the predictor, Sagarin has Duke as the 38th best team in the nation, and #2 in the coastal, behind only VaTech. One neat useful thing about Sagarin is that we can use it to project the betting lines for each remaining game of Duke, which are as follows:
Duke @ Pitt : Pitt favored by 1.
Duke @ Cuse: Duke favored by 1.
Duke vs VT: Duke favored by 1.5
Duke vs UNC: Duke favored by 9
Duke vs Wake: Duke favored by 24.

No that last one is not a typo. Wake is BAD. But as you can see, the next three games are basically toss-ups, whereas the last two are clear Duke favorites. So Sagarin would expect for Duke to most likely win 3.5 games, so they'd finish with either 9 or 10 wins. Depending upon who those wins are against (if the loss is vs Cuse it's pretty okay since that wont hurt tiebreaks), the team could still win the Coastal.

Note that Sagarin simply uses scores to make his rankings, so if teams are getting lucky to score, Sagarin won't know.

Football Outsiders

By contrast, Football Outsiders uses two main statistical measures, FEI and S&P+, to try and determine which teams are truly good instead of being lucky with points.

FEI has for two years now been far more favorable to Duke than S&P, for a simple reason: S&P is seemingly based upon yards gained per various types of plays, while FEI is focused more on drive efficiency - so a bend but not break D will look more favorable in FEI if it's working. It's debatable which is a better predictor than the other, so Football Outsiders also likes to combine these measures into a measure called F/+. Unfortunately, F/+ is not yet updated for this week, so I'll post it here when it's updated, probably tomorrow.

Anyhow, FEI currently has the Isles as the 24th best team in college football. FEI thinks Duke is the 57th best offense, but the SEVENTH best defense in the COUNTRY. FEI also ranks Duke as the 6th best in special teams efficiency (due to the #13 field goal performance, #15 kick return performance, and #6 performance in punting and punt coverage).

S&P is a lot less optimistic, putting Duke at 56th overall and behind UVA, VT, NC State (it REALLY likes State for some reason), and GTech. By S&P, Duke is 60th on O and 53rd on D. S&P also has run/pass breakdowns - Duke's O is 50th in rushing, 97th in Passing. Duke's D is 78th in stopping the run but 37th at stopping the pass.

--------------------------
One other football outsiders resource you might want to look at are their line play rankings, which rank the Offensive (http://www.footballoutsiders.com/stats/ncaaol) and Defensive lines (http://www.footballoutsiders.com/stats/ncaadl. The O Line is ranked 49th in the Country, but its so low only because Duke hasn't been able to perform well punching it through on 3rd and short. Duke is SECOND IN THE COUNTRY at preventing sacks. That's pretty awesome.

Naturally our D Line is well below average - 93rd in the Country, below average in practically every category.

Anyhow, except for F/+, which I'll update tomorrow, that's it for this week.

loran16
10-22-2014, 07:13 PM
Update:

http://www.footballoutsiders.com/stats/fplus - F/+ is up, and Duke ranks 24th in that as well, with the #59 O and the #14 (!!) D.

Reilly
10-22-2014, 09:56 PM
based on SRS ...

@Pitt ... Duke fav by 3 on neutral field so toss-up at Pitt
@Cuse ... Duke fav by 8.5 on neutral field so Duke fav by 5.5 at Cuse
VT ... Duke dog by 1.35 on neutral so Duke fav by 2.5-ish at home
UNC ... Duke fav by 7.5 on neutral field so Duke fav by 10.5 at home
WF ... Duke fav by 18+ on neutral field so Duke fav by 21+ at home

2014 School Ratings | College Football at Sports-Reference.com

Nugget
10-22-2014, 10:35 PM
Update:

http://www.footballoutsiders.com/stats/fplus - F/+ is up, and Duke ranks 24th in that as well, with the #59 O and the #14 (!!) D.

The Football Outsiders opponent-adjusted stats really are amazing -- can't ever have imagined seeing the day when a Duke Football defense ranks 14th in the country! Just 4-5 years ago we were probably within the 10 worst defenses in all of college footbal, with almost literally no chance of stopping anyone unless their runners tripped over their own feet or QBs missed wide-open receivers.

What a remarkable turnaround.

And the ACC's Ds as a whole look like they should get a little more respect, with Va Tech, Clemson and Louisville joining more-widely recognized Alabama, Ole Miss and Stanford as the top 6 defenses.

The #59 ranking for our O confirms the "eye test," that we definitely have room for improvement there -- could use more consistency from Anthony Boone and performance from our O-line on first and 10/3rd and short rushing situations that comes closer to matching what they did last year (finishing the year at #33 in Offensive Efficiency per Football Outsiders).

loran16
10-22-2014, 10:35 PM
Bill Connelly writes at SB Nation and Football Outsiders, and is behind S&P+. He wrote an ACC Projection today:

http://www.sbnation.com/college-football/2014/10/22/7037997/acc-football-standings-predictions-projection-2014

Via F/+, he gives Duke the following odds at winning:
@Pitt: 43.8%
@Cuse: 92.1%
VaTech: 60.6%
UNC: 93.5%
Wake: 98.4%


That Cuse # is far higher than any other system. He also projects us as most likely to go 6-2 and for the only other likely 6-2 team to be Pitt - so beating Pitt next week will be critical. You should all really read the analysis, it's REALLY GOOD.

loran16
11-02-2014, 10:53 AM
Updated Sagarin #s after last week's game:

In the predictor, we are 37th, basically the same as before. As of now, we are favored by Sagarin clearly in all remaining games, with the projected lines as follows:
Duke -4 CUSE
DUKE -6 VaTech
DUKE -11 North Carolina
DUKE -25 Wake Forest.

The computers are still not impressed with UNC. And being home swings VaTech very much in our favor.

Reilly
11-02-2014, 04:18 PM
Duke is the 34th best team per S-R:

http://www.sports-reference.com/cfb/years/2014-ratings.html

Acymetric
11-02-2014, 04:43 PM
I don't really disagree with our rankings in Sagarin, but has he ever addressed the difficulty in good statistics with the lack of cross conference play? I know in basketball it takes a while for teams to be "fully connected" in KenPom and could imagine in football that with the high level of variance in single games and lack of games against other conferences (easy for the few non conference games to show results that don't truly reflect the strength of either team) it would be difficult to have meaningful "geek" rankings in college football.

roywhite
11-02-2014, 05:35 PM
After the games of Saturday, Nov. 1, Duke checks in at #22 in the AP and #20 in the USA Today.

http://espn.go.com/college-football/rankings/_/week/11

devildeac
11-02-2014, 05:41 PM
After the games of Saturday, Nov. 1, Duke checks in at #22 in the AP and #20 in the USA Today.

http://espn.go.com/college-football/rankings/_/week/11

Wait! Texas A&M in "others receiving votes?" Must be the Duke (football) curse now:rolleyes:.

loran16
11-02-2014, 09:25 PM
I don't really disagree with our rankings in Sagarin, but has he ever addressed the difficulty in good statistics with the lack of cross conference play? I know in basketball it takes a while for teams to be "fully connected" in KenPom and could imagine in football that with the high level of variance in single games and lack of games against other conferences (easy for the few non conference games to show results that don't truly reflect the strength of either team) it would be difficult to have meaningful "geek" rankings in college football.

Well, this is an issue really with all football leagues - pros to - not just college. Your sample size is very small so it's always going to be very volatile. That said, gaps between teams tend to be larger in college, so it's less of an issue than you'd think.

The FO stats use data from past seasons for like 6 games, for this reason, but obviously it still is an issue.

loran16
11-03-2014, 11:30 PM
FO's stats are all up right now.

First the S&P+ rankings, which rely upon yards and efficiency gaining/stopping yardage. Duke is now #50 in the country, with the #40 Offense and #56 Defense. S&P+ allows us to break down how Duke is doing in run and pass O and D, and we are as follows:
Offense: #34 Rushing, #68 Passing.
Defense: #54 Passing D, #86 Rushing D.

Second, FEI. Duke now ranks 22nd in FEI, 40th in offense, and 12th in defense. Duke is also #2 OVERALL in the nation in special teams, per FEI. For those curious how special teams is rated so high, Duke is:
10th in Field Goal Kicking efficiency
2nd in kickoff return efficiency - (Thanks Devon!)
7th in Punt efficiency (how well opponents do on our punts - thanks Will Monday!).

Pretty damn good.

Finally, the combined ranking, the F/+ rankings:
Duke checks in at 26th overall now, 41st in Offense and 32nd on D (#2 again in Special Teams, but F+ special teams is basically just FEI).

Overall, we've basically remained pat in the rankings through our bye and games vs Pitt. Not much of a surprise. Of note, our remaining opponents are ranked through these systems:

Cuse: 71st F/+, 77th S&P+, 61st FEI
VT: 30th F/+, 26th S&P+, 31st FEI
UNC: 72nd F/+, 75th S&P+, 58th FEI
Wake: 84th F/+*, 113rd S&P+, 86th FEI

(Note the gap between 84th and 72nd is the same as the gap between 72nd and 49th).

loran16
11-06-2014, 08:09 PM
Updates:

ACC Projections from Bill Connolly are up here: http://www.footballstudyhall.com/2014/11/5/7161467/acc-football-projections-predictions-duke-florida-state

The money picture:
http://cdn2.vox-cdn.com/uploads/chorus_asset/file/2428440/ACC-Week11.0.png

F/+ actually thinks Miami should be favorites over FSU, which is the problem. Still Duke is up to 56% to go 7-1.

In addition, Saturday's predictions can be found here:http://www.footballstudyhall.com/2014/11/6/7163513/college-football-fplus-projections-week-11

F/+ has Duke beating Cuse by either 15 or 26.3, depending upon what method you use. This should NOT be a close game per the numbers.

uh_no
11-07-2014, 12:48 AM
Updates:

ACC Projections from Bill Connolly are up here: http://www.footballstudyhall.com/2014/11/5/7161467/acc-football-projections-predictions-duke-florida-state

The money picture:
http://cdn2.vox-cdn.com/uploads/chorus_asset/file/2428440/ACC-Week11.0.png

F/+ actually thinks Miami should be favorites over FSU, which is the problem. Still Duke is up to 56% to go 7-1.

In addition, Saturday's predictions can be found here:http://www.footballstudyhall.com/2014/11/6/7163513/college-football-fplus-projections-week-11

F/+ has Duke beating Cuse by either 15 or 26.3, depending upon what method you use. This should NOT be a close game per the numbers.

our chances of winning the division should be:

p(7-1) + p(6-2)*(~p(miami 6-2)) == .56+ .36 * .58 = ~.77 = 77% chance of winning division

OldPhiKap
11-07-2014, 01:02 AM
our chances of winning the division should be:

p(7-1) + p(6-2)*(~p(miami 6-2)) == .56+ .36 * .58 = ~.77 = 77% chance of winning division

[PolySciMajor] It's only 50%. Either we will, or we won't. [\PolySciMajor]

bob blue devil
11-07-2014, 06:18 AM
F/+ has Duke beating Cuse by either 15 or 26.3, depending upon what method you use. This should NOT be a close game per the numbers.


i love these analyses. only problem is that my most trusted number, the vegas line, has us as only a 3.5 point favorite - oh dear. i wonder why the huge divergence.




our chances of winning the division should be:
p(7-1) + p(6-2)*(~p(miami 6-2)) == .56+ .36 * .58 = ~.77 = 77% chance of winning division


<seeking an honorary position with the grammar police>

i believe your notation should reflect a NOT in front of the miami going 6-2.
p(7-1) + p(6-2) * (1 - (~p(miami 6-2))) == .56 + .37 * (1 - .45) = ~.76
and of course there is the weird 5-3 outcome where we actually win the division, which i'm guessing you assumed away as ~0. or do the tiebreakers already eliminate us there (as they do with a 6-2 tie that includes miami)?

Deslok
11-07-2014, 06:52 AM
i love these analyses. only problem is that my most trusted number, the vegas line, has us as only a 3.5 point favorite - oh dear. i wonder why the huge divergence.




<seeking an honorary position with the grammar police>

i believe your notation should reflect a NOT in front of the miami going 6-2.
p(7-1) + p(6-2) * (1 - (~p(miami 6-2))) == .56 + .37 * (1 - .45) = ~.76
and of course there is the weird 5-3 outcome where we actually win the division, which i'm guessing you assumed away as ~0. or do the tiebreakers already eliminate us there (as they do with a 6-2 tie that includes miami)?

~p is not p(in mathematical logic terms). So either as it originally was, or 1- p(miami 6-2).... yours is 1 - p(miami not 6-2).

uh_no
11-07-2014, 09:31 AM
<seeking an honorary position with the grammar police>

i believe your notation should reflect a NOT in front of the miami going 6-2.
p(7-1) + p(6-2) * (1 - (~p(miami 6-2))) == .56 + .37 * (1 - .45) = ~.76
and of course there is the weird 5-3 outcome where we actually win the division, which i'm guessing you assumed away as ~0. or do the tiebreakers already eliminate us there (as they do with a 6-2 tie that includes miami)?

Sorry! i minced symbol meanings

~p == "not p" or (1-p)

~.76 = "about .76"

Wander
11-07-2014, 09:46 AM
and of course there is the weird 5-3 outcome where we actually win the division, which i'm guessing you assumed away as ~0. or do the tiebreakers already eliminate us there (as they do with a 6-2 tie that includes miami)?

I think there's only one very specific scenario where we win at 5-3: we beat VT and UNC but lose to Wake and Syracuse, Miami loses to UVA and/or Pitt, and Georgia Tech goes 1-1 against NC State and Clemson. So, yeah, pretty close to 0%.

AncientPsychicT
11-07-2014, 10:28 AM
I think there's only one very specific scenario where we win at 5-3: we beat VT and UNC but lose to Wake and Syracuse, Miami loses to UVA and/or Pitt, and Georgia Tech goes 1-1 against NC State and Clemson. So, yeah, pretty close to 0%.

It doesn't have to be that specific. All that needs to happen for us to win at 5-3 would be for only us and GT to finish 5-3 (i.e. everyone else is 4-4 or worse). Since we beat GT head-to-head, we would win the division on the first tiebreaker. It wouldn't matter to whom those specific losses that got us and GT to 5-3 would be (although us losing to UNC would increase the chances of UNC also finishing 5-3).

Anyway, the amount of losses necessary to teams that have shown they are decent if not better makes this whole point moot regardless. This ain't gonna happen.

Wander
11-07-2014, 10:38 AM
It doesn't have to be that specific. All that needs to happen for us to win at 5-3 would be for only us and GT to finish 5-3 (i.e. everyone else is 4-4 or worse). Since we beat GT head-to-head, we would win the division on the first tiebreaker. It wouldn't matter to whom those specific losses that got us and GT to 5-3 would be (although us losing to UNC would increase the chances of UNC also finishing 5-3).

Anyway, the amount of losses necessary to teams that have shown they are decent if not better makes this whole point moot regardless. This ain't gonna happen.

Oops. Yeah, this is right. What I described is (I think) the only way we win a 3-way tie, but of course if Miami loses 2 of their last 3, then yeah we'll win a head-to-head with GT (though note if UNC is one of our losses that get us to 5-3, things get more complicated as they may be in the mix).

loran16
11-11-2014, 11:35 PM
Late update! Duke rises to 34th Sagarin in the predictor.

Football Outsiders' #s are more interesting.

Per FEI (and by default, F/+, since F+ uses FEI for special teams), Duke has the NUMBER one special teams in the nation! http://www.footballoutsiders.com/stats/feist Three top 10 units per FEI - field goal kicking, punting, and kickoff returns, with punt returns being top 30 and only kickoffs being around average. Neither our O or D is elite by itself, but elite special teams gives both units a head start on opponents that has been huge all season.

Duke is now 20th in FEI overall, 37th on Offense and 14th on Defense to go with the #1 special teams.

In S&P+, Duke is now 45th overall, with the 43rd best O and the 49th best D.

As for FO's combined metric, F/+, Duke is back in the top 25 at #23, with the 44th best O, the 23rd best D, and the #1 Special teams in the land. Yeah that's pretty nice.

Tomorrow there should be updated ACC projections, which should be very interesting.

HK Dukie
11-12-2014, 02:25 AM
Duke is now 20th in FEI overall, 37th on Offense and 14th on Defense to go with the #1 special teams.


The ACC has 6 teams in the top 20 of this FEI system. Notre Dame at 24th also for good measure.

Those 6 teams (FSU, Clemson, Louisville, Duke, GaTech, Miami) are solid football teams. Why is everyone so down on the ACC this year? Even the worst ranked team in the Coastal beat Ohio State. BC and Virginia went 1-1 against those southern California schools and they are not even on this list.

I'm looking forward to seeing what the ACC can do in bowl season. It will have an impact on where everyone starts ranked next year. Additionally, hopefully we individually finish strong enough to start the season ranked next year so we don't have to worry having a better record and still getting passed over by another ACC team for a bowl game.

OldPhiKap
11-12-2014, 09:09 AM
Late update! Duke rises to 34th Sagarin in the predictor.

Football Outsiders' #s are more interesting.

Per FEI (and by default, F/+, since F+ uses FEI for special teams), Duke has the NUMBER one special teams in the nation! http://www.footballoutsiders.com/stats/feist Three top 10 units per FEI - field goal kicking, punting, and kickoff returns, with punt returns being top 30 and only kickoffs being around average. Neither our O or D is elite by itself, but elite special teams gives both units a head start on opponents that has been huge all season.

Duke is now 20th in FEI overall, 37th on Offense and 14th on Defense to go with the #1 special teams.

In S&P+, Duke is now 45th overall, with the 43rd best O and the 49th best D.

As for FO's combined metric, F/+, Duke is back in the top 25 at #23, with the 44th best O, the 23rd best D, and the #1 Special teams in the land. Yeah that's pretty nice.

Tomorrow there should be updated ACC projections, which should be very interesting.

Bingo. Nicely put, and exactly right.

And as for last week:

1. Will Monday was the Ray Guy Award punter of the week.
2. On the stopped fake kick from Syracuse, Cut said in his PC yesterday that they watched film of all fakes Syracuse did last year and noticed that formation. They showed it to the team the night before. When Syracuse lined up, every one on Duke's side knew it was a fake and even the sideline players were yelling it in to the guys on the field.
3. Beamer knows special teams; we will be well-matched in preparation this week (not sure how they are executing though).

fisheyes
11-16-2014, 02:36 PM
Duke holds on to the 25 spot in both the AP and Coaches polls!
Ahead of some pretty impressive power-house names (Clemson, ND, LSU).

Go Duke!
Go To Hell Carolina!!!