By the (Advanced) Numbers: Virginia Tech Shall Have Dominion

The Hokies travel to the 757 to face the struggling Monarchs.

Old Dominion made the move to FBS in the 2014 season, and after a couple of years of expected growing pains, the team went a surprising 10-3 in 2016, finishing 2nd in Conference USA. The wins weren't a fluke either — by S&P+ standards, the Monarchs were the 73rd-best team in the country; while not great by national standards, for a C-USA team the rating was pretty good and bookended by better-known programs Duke (72nd) and Oregon (74th).

Then 2017 came around, and the team plummeted to 115th. Rather than building on success the Monarchs took a big step back, and so far in 2018 they look to be staying there. The team is currently rated 111th out of 130 teams and has started the season 0-3.

The full set of statistics will not be available until mid-season, so this week focuses on overall team ratings.

Where the Season Stands

Virginia Tech is currently 3-0 by way of two victories and a forfeit (Author's note: That is a joke. Please reserve complaints about the game cancellation for another column. There have been pleas to not play ECU for years and everyone finally got what they wanted.). The actual and predicted score difference and odds of winning each game are:

The odds of each possible regular season win total are now:

I'll go out on a limb and guess the win totals raised some eyebrows, so let me address that here. First off, Virginia Tech just reduced its total number of games by one, and the one dropped was almost certainly a victory. There are simply less games to win, so whatever expectation you had before the cancellation for wins should be reduced by one unless you were truly concerned about a victory over the Pirates. Additionally, it is easy to think that the Hokies should have upped their win expectation considerably after dismantling Florida State, but there are a couple of headwinds to that argument:

  1. Florida State is not a good team. Not just disappointing relative to expectation, but a genuinely bad team. The Noles' current S&P+ rating is 77th, and that still includes a substantial preseason component. Teams near them include Eastern Michigan and Tulane, and I doubt Hokie fans would be asking for credit for beating either of those 24-3.
  2. The rest of the schedule is not looking nearly so easy now. Two teams in particular — Boston College and Duke — have been downright impressive and will be tough to beat. Enjoy this weekend because it currently looks like a 50/50 ride from here on out.

Rankings and Computer Predictions

The computer rankings and predictions for each team:

There are no computers that consider Old Dominion to be any sort of threat, but at times the Hokies have not put away weaker opponents as easily as expected. Hope ODU is dispatched with the ease in which William & Mary was. The odds of a 28-point favorite winning is 96.7%.

Next is a look at any overall offensive or defensive advantages:

Obviously the Hokies are better on the defensive side of the ball and overall, but the offensive advantage is not as large as one might imagine. I am far from convinced that this Virginia Tech offense is anything better than mediocre, and really hoping that the team puts up a stronger showing against ACC competition. The unit only scored 17 on an FSU defense that gave up 26 to Samford's (not Stanford's) offense and 30 to Syracuse's offense.

Field Position

Brian Fremeau, creator of the FEI ratings, shared the national leaders in Net Field Position, and thanks to some excellent defense and special team's, Virginia Tech is currently the national leader:

Net Field Position is simply the average number of yards a team starts from their own end zone, minus the average start of their opponents. Some context:

  1. FEI only counts games against FBS opponents, so the Hokies' rating is based only on the Florida State game at the moment. A small sample size for sure.
  2. Garbage-time drives — when one team is all but certain to win — do not count towards the ratings.

So how did the Hokies get there? Look at the starting field position for Virginia Tech and Florida State from their matchup:

Doing that to a team is a felony in some states. While Florida State started drives where you expect most teams to — inside their own 30-yard line — Virginia Tech was able to start drives all over the field, including six times on the Seminoles' side of the field!

Alabama is No. 2 nationally, and third is — thanks to a friendly matchup against a team that's just been discussed — Syracuse. Talkin' bout the Noles here.

So is this sustainable? Well the number itself — 17.8 yards — certainly isn't. But neither is Alabama or Syracuse's. The ranking, however, is very much sustainable: the Hokies finished 2017 ranked 2nd nationally in the same metric, and rated 11th in 2016. The James Shibest/Bud Foster combination is lethal.

Old Dominion ranks 30th in this metric.

Special Teams

While detailed statistics are not available for special team's units, S&P+ lists the Hokies as the No. 1 overall unit in the nation - BeamerBall lives on!

Old Dominion's unit is 69th. Nice.

Who To Watch Out For

Don't be fooled...even the lowest ranked FBS teams can still have some playmakers:

  1. SR WR Jonathan Duhart is averaging 21.18 YPC on 17 catches this season
  2. Fellow SR WR Isaiah Harper has averaged 29.2 YPR on kickoff returns
  3. SO LB Lawrence Garner leads the nation with 14 tackles per game

Statistical Key to the Game

After this week, there are no games left on the schedule where it is almost certain that the Hokies will outplay their opponent. The key here is to avoid the upset formula ODU will likely employ, which is limit possessions by bleeding clock and win the turnover battle. The latter is my statistical key to the game — even if possessions are limited, winning should be a breeze as long as the Hokies don't get defeated by turnovers. Give no more than one away, and fans will be heading back to their tailgate in the third quarter.

Statistical Prediction

It's hard to make any specific predictions without enough data to tell what teams are made of this season, but there are a few that I think will happen. I think the Monarchs will burn the Hokies on a couple of long passing plays, but nothing that will put the game in jeopardy. I think we will see at least two quarterbacks play for VT, and Wheatley will be brought in early in this game to better assess his abilities. Mostly Fuente keeps things simple and lets his superior players beat ODU with athleticism, and the game is never really in question.

Virginia Tech 38, Old Dominion 10

As always a thanks to Football Outsiders, cfbstats.com, and Minitab Statistical Software.

Comments

Lets show whos the real Monarch (butterfly lion?) in the 757!!!

I don't know what a Hokie is but God is one of them (excluding JMU/ODU)- Lee Corso

I'm sorry, I'm not doubting you, but is that seriously saying we're losing 2/3 of the times we play Duke, ND, and BC this year?

Something about those numbers seems very off. I get that FSU isn't good, but that just doesn't pass the eye test. Sure, I'll buy those games being 50/50 odds, but 34/66 in favor of Duke? I dunno....

"Some days you’re a horse and some days you’re a horse’s ass. I’ve been a horse’s ass for a little while." - Roy Halladay

Agreed, no way of just a 68% chance to beat UVA and a whopping 58% to beat the mighty Nardouchey. This team is young but very talented and I can't wait to see how they will respond to a quality opponent on the road next week.

I don't know what a Hokie is but God is one of them (excluding JMU/ODU)- Lee Corso

Agreed, we will learn a lot the next 2 weeks against Duke and Notre Dame. I think we'll be good later this year, but VT hasn't had the opportunity to show how good they are yet. I thought our defense passed a big test week 1, but now I don't know.

This looks to me like statistics being wonky because of a small sample size on our end. We only have 1 FBS opponent to gauge ourselves off of and unfortunately that one team is butt. Even without playing us it looks like ECU has found a way to mess with us...(mostly joking)

New this year, S&P+ will include its preseason component all year. It's also hung up on returning production, so it disregards young overperformers.

This is correct...there is still a heavy preseason component that includes how good you were last year, how much production you lost, and what have your recruiting ratings been like recently. Please tell me if any of those things scream to you that VT should have had a high preseason rating, because to me none of them do.

Then, VT has only played two games. One of those was against FCS William & Mary, who themselves have only played one other game (against Bucknell). Against Florida State, the Virginia Tech offense was outperformed by Syracuse and Samford. Samford! I'm optimistic about VT's future, but am baffled that so many people can't understand why an objective rating system would favor BC over VT at the moment. Let's get a couple more games in and we'll have a much better idea of how good each is.

My complaint isn't with the BC prediction (I'd have that game at 50-50 or so). It's with trying to use computer stats at all early in the season for a sport with such a small sample size. It also downgrades VT for playing against an FSU team that was better than the one playing now (FSU's lost additional OL since Week 1, and is now playing guards at both tackle spots).

BTW, the preseason component is permanent now IIRC. That was a new change for this year.

The S&P+ rankings expand the sample size significantly by looking at every play except in garbage time. It looks at five primary factors of how well a team is performing: "explosiveness" (or big-play potential), "efficiency" (or success rate), "finishing drives" by scoring once inside the 40, "field position" (self-explanatory), and "turnovers" (yep).

You're also overlooking the math behind our "turnover luck" against FSU. First, let's look at fumbles. Historically, the team that fumbles the ball retains possession about 50% of the time. It's essentially a coin flip as to which team recovers. However, in the FSU game, the ball hit the ground 6 times, and we gained or retained possession through all 6. Right there, that gives us a +3 bonus to "turnover luck".

As for interceptions, historically a team will intercept one pass for every 4 that a defender gets his hand on. FSU deflected 2 passes, and intercepted none, so that's another +0.5 bonus to our "turnover luck". We intercepted 3 passes and deflected 3 more, for another +1.5 bonus to "turnover luck", which brings our overall total to +5.

Considering that, on average, a turnover creates a field-position advantage worth a total of 5 points, we gained 25 points on FSU merely by "turnover luck", while we only won by 21, 7 of which were off of a blocked punt. Obviously, our defense and special teams were dominant aside from a handful of plays, but outside of the first drive our offense was largely conservative and outright bad. In fact, FSU actually outgained us in the game by 8 yards. Thus, we are currently #41 in S&P+, and FSU has fallen all the way to #77. So yes, statistically speaking, our game against FSU was determined almost entirely by turnover luck, and there's no way that a couple injuries to the offensive line should create a 36 position swing in team ranking.

TL;DR: We absoluely haven't proven anything so far this season, so we shouldn't expect any advanced statistics-based rankings to hold us in any regard at this point. Let's take care of ODU this weekend, and over the following two weeks we will learn a lot more about who this team is.

I was just talking about this. I do "data sciency" like S&P+ things for a living. In the past, S&P+ didn't seem to adjust very much for games that looked like quality W's, when we found out that the loser wasn't very good (ND@UT a few years back is a key example). It's good to see the metrics adjust as more data becomes available.

It will take time to determine if any of Duke's or BC's W's (all of the teams that both have played are somewhat suspect) are being properly attributed within S&P+ since we don't know much about the teams that they played.

I get where the "turnover luck" stat comes from. I was an evangelist for S&P+ before this year.

This year I realized there are just too many things it can't account for, especially early in the season. Trying to expand sample size by changing the sample from games to plays isn't really a 1-to-1 comparison. Also, when you're talking about the difference between 41 to 77, you're in the middle of the bell curve. By definition, there isn't as much difference between teams there as there would be between, say, 81 and 117 (or 1 and 37). So yes, FSU now having to scrape the bottom of the barrel for their OL at multiple positions does make a huge difference (not that they were that good to start with).

I'd agree this team hasn't really proven anything this year, but for now, I'm going to ignore advanced statistics when they have such a low sample size. That, or wonder how bad Memphis would beat VT on a neutral field, I guess.

You and I are arguing different points, my friend. I agree that the advanced stats don't tell the whole story, especially this early in the season. Even now there isn't a big-enough sample size with plays to draw any sort of statstically-significant conclusion. However, by the end of the season it's pretty close, if not actually big enough, whereas 12-15 games will never constitute a large enough sample size for any sort of accurate comparison, at least not while there are 130 teams in the picture. 16 games is just about the minimum required to get a good feel for 32 NFL teams.

I am merely responding to the uproar about how these advanced stats are giving us such terrible odds for some of these games by explaining why it is the way it is. I am also ignoring the stats for now (there's no way we're only the 41st best team in the nation), and am looking forward to seeing these next 3 games to give us a much better feel for how good of a team we have. I do not actually think that BC is a better team than ND, though they are probably closer than the polls would suggest (#23 and #8). I do not actually think that Duke and BC have a 2/3 chance to beat us, though I wouldn't put it much less than 50%. I do not actually think Memphis would beat us on a neutral field more than 3 out of 10 times. But, the absolute truth is that we don't know right now, and we won't for at least a few more weeks.

Yeah I think we're of the same mind here. I have much less of an issue with advanced stats toward the end of the year.

Have an upvote.

Because BC has really proven so much against Wake, UMass, and Holy Cross?

I mean I get it, FSU is a dumpster fire this year, but lets not pretend that BC has been slicing through a murderers row of opponents. 50/50, sure, but this system says they're as likely to beat us as #8 in the country Notre Dame, and I don't buy that at all.

"Some days you’re a horse and some days you’re a horse’s ass. I’ve been a horse’s ass for a little while." - Roy Halladay

In the S&P+ rankings, BC is #17 and Notre Dame is #18. Notre Dame has been extremely fortunate to win each of their 3 games, beating #10 Michigan, #35 Vanderbilt, and #110 Ball State by one score each. On the other hand, Boston College easily handled both of their first two tests (#116 UMass and FCS Holy Cross), and beat #57 Wake Forest in a game that was not as close as its final score would indicate.

So... let me get this straight...

ND beats 2 teams whose combined ranking would still be better than the single best team that BC has played, but because they had a close win hiccup against their third opponent, they're ranked lower than a team who has played a FCS opponent, and teams ranked 116 and 57?

"Some days you’re a horse and some days you’re a horse’s ass. I’ve been a horse’s ass for a little while." - Roy Halladay

In short: yes.

The S&P+ rankings care less about a team winning or losing and more about how well they played. Based on how well each team played, the S&P+ rankings say that Michigan had a 57.5% chance of winning that game (S&P+ calls this "post-game win probabilty"), meaning if the two teams played 10 times, and each team played just as well in each game, Michigan would probably win about 6 of those 10 games.

According to S&P+, Notre Dame controlled the Ball State game with a 98% post-game win probability and the final score was much closer than it actually was, but Vandy had a 56.5% post-game win probability. Thus, Notre Dame should be 2-1, but they got lucky and won both of their games that were essentially toss-ups.

On the other hand, by their metrics, Boston College absolutely dominated all 3 of their games, with post-game win probabilities of 99.9%, 100.0%, and 92.0% over UMass, Holy Cross, and Wake Forest, respectively.

I'm not buying it. I am not a data scientist but I play one on TV. By which I mean I teach a data science course at the undergrad level. I think S&P+ made a mistake weighting the preseason rankings as heavily as they have.

Also, Scot Loeffler vs Bud Foster is not something a statistical model can account for.

If you listen to Bill Connelly on the PAPN podcast he explains this pretty well. The ratings reward or punish you for how well you fared compared to expectations. For instance, Auburn fans were pissed when their team actually dropped in the ratings after they beat Alabama State 63-9 in week 2. Why? Because the S+P Plus ratings predicted a 70- or 80-point win. They underperformed the expectations. BC has been pretty dominating in their early wins, while ND has been pretty underwhelming. The numbers definitely look kinda screwy early in the season when there's not a lot of data to work with yet.

“You got one guy going boom, one guy going whack, and one guy not getting in the endzone.”
― John Madden (describing VT's offense?)

The S&P+ ratings don't care about scoring margins at all, they look at what Bill calls his "five factors of college football" on a play-by-play basis, looking only at non-garbage-time plays. Auburn didn't perform quite as well as expected before garbage time kicked in, so they got penalized for it.

I walked away from the cult of S&P+ this year. Stuff like this is why. The other thing was the FSU result; by S&P+'s reasoning, VT was extremely lucky to beat FSU, and should have lost outright. Time has born out that VT's actually much better than FSU, but S&P+ doesn't like that VT created so many turnovers and gave up so few.

Edit: S&P+ considers 2-1 Memphis to be a huge favorite on a neutral field vs. VT.

Edit #2: After Texas pounded USC, USC jumped 8 spots to pass Texas. Yeah. Also, Troy falls 8 spots for beating Nebraska. Makes sense: it hurt Troy's strength-of-schedule to play a team so bad Troy could beat them, so Troy drops.

With regards to turnovers, the bigger issue was that VT recovered every single fumble that occurred for both teams, and long-term, statistics have shown which team recovers a fumble is pretty close to 50/50. Have FSU fall on three of those instead of VT, and you're looking at a very different game.

Love the starting field position chart. I still think we look faster and stronger this year so far, but your comment and that chart put the FSU game into a little more perspective. That said, the field position was also a result of quality special teams and good coaching decisions. Both of those are clearly important for teams with high aspirations.

...or you could say it was due to lack of offensive production in the 3rd quarter. We basically played that whole quarter on the FSU side of the field, with 3 punts and a turnover on downs (kudos to Bud and Bradburn).

I still like S&P+ for its components primarily, it's overall rankings I believe are much more reliable and accurate with regards to the very top teams by the end of the season. That said, I think the way S&P calculates the middle of the field 30-70ish is pretty logjammed and inaccurate. I rarely care when Bill Connelly talks about "top 40" teams because I think it does a bad job of correctly evaluating 30-70 as I said above.

edit: To clarify, that isn't to shit on S&P+, I don't think I've found any system computer or otherwise that can accurately differentiate between the 30-70 range.

Exactly...to your point there is a 23.7-point difference between #1 and #30, and an 11.3-point difference between #30 and #70. In fact, from 30-100 there is a 21 point difference so the middle 70 teams have less variation than the top 30 or bottom 30.

That said, I think that has less to do with the specific way it ranks teams, and more to do with the bell curve of team ability where the middle of the pack is going to be genuinely filled with teams that aren't a huge amount better or worse than others. The same thing happens on the other end too...the bottom 30 teams are 25.4 points apart. That's why I like to rant about people treating every bottom-of-the-pack team as equally bad but then nitpicking over the ratings at the top. They spread out at the bottom end just the same!

We are definitely on the same page.

They spread out at the bottom end just the same!

So what you're saying is, UVA levels of suck rival Alabama levels of good? Your math checks out.

That seems crazy, that they think we don't stand much of a chance against Duke and BC....Not to mention only 96.7% on beating ODU.

I mean, by statistics, we have played shit tier teams. Just real tire fires. There really isnt much reason to love VT right now unless you bought into the FSU preseason rankings which are meaningless.

So our team isnt exactly represented super well because of that. Gotta see how we perform against Duke before we can really make a judgement on the team.

Agreed. I think this analysis is still solid, but I think it goes to show how impossible it is to draw any real conclusions 2/3 games into the season.

Why is it when Duke and Boston College play really well and destroy baby schools everyone says they're very impressive and going to dominate this year but when we play baby schools and do the same thing everyone says we haven't played anybody yet? Is it because we have a higher standard or something?

BC and Duke have beaten decent P5 teams on the road already this season. We can't say the same.

The Orange and Maroon you see, that's fighting on to victory.

BC has beaten Wake, UMass and Holy Cross. Wake is the only P5 team there, and their biggest win is 23-17 over Tulane.
Duke has beaten Baylor, Army, and Northwestern. Baylor's best win is 37-20 over UT San Antonio, and Northwestern just lost to Akron.

"Some days you’re a horse and some days you’re a horse’s ass. I’ve been a horse’s ass for a little while." - Roy Halladay

This was an attempt at humor.

The Orange and Maroon you see, that's fighting on to victory.

....shit.....

"Some days you’re a horse and some days you’re a horse’s ass. I’ve been a horse’s ass for a little while." - Roy Halladay

I know you bit on a joke, but are you really arguing that FSU is as good as either Wake or Northwestern right now?

FSU beats power house Northern Illinois this weekend and changes everything!

via GIPHY

JP

Why is it people will argue over whether a team should be ranked #10 or #15 in the country, but throw every bottom half team into one bucket as a "baby school"? By the Massey Composite rating of tons of computers, Wake Forest is the 45th best team in the country and BC beat them by 7 on the road; UMass is 121st and BC beat them by 34 at home. Florida State is 73rd, well below the "baby school" of Wake Forest.

Sagarin includes FCS schools and has W&M 8 points worse than UMass. VT beat W&M by 9 more points than BC beat UMass, so basically an equal performance.

Also, you realize that these are computer ratings, right? No person is saying anything...it's just numbers that base don a very small sample size say BC looks like a better team. If VT is actually much better, that will show itself in the games between now and then.

ITT: Everyone gets mad at Joel because his computer is wrong.

Thanks for doing this every week dude.

"What are you going to do, stab me? - Quote from Man Stabbed

Shoutout to the computer model that has us winning by like 67 points.

The Orange and Maroon you see, that's fighting on to victory.

Round up. VT puts in 10 touchdowns. 70-3 good guys.

Hmm, when I round up from 67 I get .... 69

via GIPHY

Would that be the result of a Hooker run?

Never Forget #1 Overall Seed UVA 54, #64 UMBC 74

I want 70 points oohhhh so badly, I remember my freshman year the opener against appy state...they were talking mad trash and we just curb stomped em to the tune of 66-14 I believe. I have literally never seen us drop 70 on somebody and I want the day to come

Taylor, looking desperately throws it deep..HAS A MAN OPEN DANNY COALE WITH A CATCH ALL THE WAY DOWN TO THE FIVE!!!!....hes still open

I remember app st fumbled on their first or second play. David Wilson took the ensuing play to the house. You just knew at that point we had already won despite being like 30 seconds in. Good times

Virginia Tech School of Architecture Class of 2014
Fan of Hokies, Ravens, NY Giants, Orioles

You haven't seen it because it hasn't happened since 1995 against Akron. Not sure when our previous 70+ before that was, but 77 was the most we'd scored in a game since putting up 99 on Emory & Henry in 1919.

As much as we would be derided for running up the score I would love to see the Hokies drop a hundred on a team. Given that we play VMI in a few years I might get my wish.

Never Forget #1 Overall Seed UVA 54, #64 UMBC 74

I would love to see the Hokies drop a hundred on a team

As long as that team is ECU, I'm in. Hate to see that happen with ODU, though.

That was a fun game. We had like 638 yards of offense (which was way less common, of course, back then.) The next year AT Akron, on the other hand, sucked. Shyrone Stith was lightning in a bottle, but we were ranked 15th and barely eeked out a win, like 21-18 or something. Post-game was way worse. That was back when reporters were allowed in the locker room, where there was more Hokie manhood on display than a young Timbo Coles reporter wanted to see.

Help yourself, but stay above the equator!

Seems legit, we have beat up on two very bad team and will likely do so again this weekend. Then probably go .500 or so the rest of the season against real competition. Bowl streak should be alive and well though for our moral victory case...

I can imagine no more rewarding a career. And any man who may be asked in this century what he did to make his life worthwhile, I think can respond with a good deal of pride and satisfaction:
“I served in the United States Navy"

KCCO

That under bet at 8.5 wins I made is looking good right now! Not that I didn't think we'd have a good year, I just didn't think 9 wins was a possibility with this young defense. 8 wins for sure.

Certainly helps your case that the ECU game got cancelled. 8-3 seems like a very realistic result for this team. 9-2, maybe, if we get a few more lucky bounces and get hot towards the end of the season or pull off the upset on Notre Dame.

I doubt the winning % for Duke has accounted for the absence of Jones. We'll see what they can do against *checks schedule* NC Central...

I like the Net Field Position stat!

Hokie in West Africa...sadly, I can't jump up and down hard enough for it to be felt in Lane

Alabama doesn't want none of our field position! Suuuuckkkk itttttt!!

“Stats are for losers.” -Mel Kiper Jr.

Joel I think you should safely be able to interpolate the result of the ECU game* from the results of W&M and ODU. Will fix that unsightly blank spot on your first chart.

Come to Blacksburg and see what the Hokie Pokie is really all about

Interpolate between W&M and ODU...or extrapolate from FSU to W&M?

Old Dominion's unit is 69th. Nice.

LEG!

804

All the numbers.... and I guess we will just have to wait for the real games to begin and prove ourselves. Being an underdog will(should) be motivation.

Every year, I'm reminded about now how much I like this set of articles.......around mid October. Until then, too much diversity of opponents and unequal scheduling to draw many conclusions at all.

FSU was excellent this year, until we humiliated them so badly they gave up.

That's my story, I'm sticking to it.