By the (Advanced) Numbers: Keep-a-Wake

Breaking down Virginia Tech's outlook against Wake Forest, the Hokies' extremely efficient offense and the statistical factors for a Tech win.

No question about it, turnover luck played a role in the lopsided 40-14 final score against Boston College. Fumbles and tipped balls are about equally likely to go to the offense as the defense, and the Hokies certainly kept flipping tails on BC errors. But even without that fortune, Tech's defense didn't play as poorly as some believe. The Eagles gained 5.72 YPP, about half a yard higher than their season average, but from a defensive perspective 5.72 YPP is about the median nationally. It's not domination by any means, but also isn't exactly getting walked all over.

It is becoming clear that for many Hokie fans, assessing what "good" looks like will require some adjustment this season. For many years the team relied on a lockdown Bud Foster defense and big special teams plays to win 17-13 slugfests. During those seasons, the defense allowing 30 points would almost certainly be a loss and possibly a blowout. This team is set up to win games that the other team scores 30 in, partly because their rapid scoring gives the opponent more possessions (BC had 12 offensive possessions Saturday; even if the first two drives had been TD's instead of turnovers the 2.33 points per possession would have been average). But more so than that, the offense is simply equipped to keep the team in games much like the defense once did. The Hokies are scoring 3.12 points per possession (non-garbage time) according to Brian Fremeau, and are 18th nationally in points per drive differential with the offense scoring .81 more points per drive than the defense allows. In a typical game, the more drives each team has, the more likely that math is to pay off with a win no matter how high the score gets.

Turning the corner from Boston College, the Hokies now face a team on the opposite side of that statistic β€” Wake Forest gives up .77 more points per drive on defense than they score on offense. The Demon Deacons are coming off a 17-point win over Virginia, and that would be easy to attribute to the absence of the Cavaliers' starting quarterback if not for the fact that the eye opener was the 40 points Wake Forest put on the board. In the loss, the reigning Coastal champs gave up 7.00 YPP to the Deacs, more than the team gave up to No. 1 Clemson earlier in the season.

So should Virginia Tech's suspect defense fear a potential offensive force taking the field in Winston-Salem? Or should the Demon Deacons have the same concern?

Where the Season Stands

Virginia Tech is currently 3-1. The actual and predicted score difference and odds of winning each game are:

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

As things stand right now, Virginia Tech should be favored in each remaining game except for Clemson. The downside is that this doesn't mean fans should expect to win them all. The upside? A 12% chance of beating Clemson is...not bad at all. The team is also at about a 50/50 chance of 8 wins or better, and nearly a 98% chance of "traditional" bowl eligibility, although winning requirements don't mean much this year.

Based on FPI, the odds of each ACC team beating the ACC teams on their schedule is:

Clemson and Notre Dame remain ACCCG favorites, and the second tier of teams (Miami, NC State, VT, and UNC) remains the same even with UNC's improbable loss to Florida State. The highest win total from the third tier is a full two games behind any of those teams, with reigning Coastal champ UVA only favored in a single remaining game (against Boston College).

I'm sure there are plenty of obscure scenarios out there but I want to lay out the current most reasonable path(s) to the ACCCG for the Hokies. With the unusual setup this season, the ACC will be giving the two teams with the highest conference winning percentage spots in the game, but there are restrictions on how many conference games you must play and how tiebreakers will be handled. Based on those rules, for those with an eye on the ACCCG, Tech's best hope to participate involves one or more of the following:

  1. At least one of the favorites (Clemson or Notre Dame) does not meet the minimum number of games requirement.
  2. This may seem obvious, but Virginia Tech wins out including Clemson. The scenarios here are numerous, but in the absence of failing to meet the minimum games requirement Clemson is almost certainly undefeated or one-loss. In a two-loss Clemson scenario with Virginia Tech winning out, the Hokies now have the tiebreaker against Clemson if they are the only two 1-loss teams. If there are multiple, then Virginia Tech needs not only the 3+ team tiebreaker (combined head to head against some combination of Clemson, Notre Dame, NC State, UNC) but also to only have at most one undefeated team. The latter seems clear as Clemson would have defeated Notre Dame and all others have already suffered a loss. But the former? If Virginia Tech wins out, they will be 2-1 against the other potential one loss teams. But UNC and NC State play each other, so only one of those could be among the tied teams; and because Clemson and UNC both play Notre Dame no more than two of those three can make it. To summarize the potential ties here, if UNC is not among the 1-loss teams Virginia Tech would be in the ACCCG and if UNC is in there's still a chance. If you're looking for what to cheer for aside from the Hokies running the table, it is UNC suffering a second loss.

Of course there's only a 2% chance of Virginia Tech winning out and any of this means anything, but good to know they are in a great position should that happen.

Virginia Tech Leads the Nation In...

Among Power 5 teams, the Hokies have the #1 YPC rushing rate at 6.9 yards, a full yard ahead of #2 Notre Dame. Wake Forest has the 7th worst YPC allowed among Power 5 teams at 5.1.

Rankings and Computer Predictions

The computer rankings and predictions for each team:

Wake Forest might be the 18th best team in the country. Or maybe 105th. It's really hard to tell I guess. As regular readers know, the spread tends to sit around the middle of computer predictions but in this game is heavily favoring the Hokies. I suspect that Vegas has factored in the likelihood of many fewer Hokie players out than before, and confidence that Hendon Hooker will start and play all or nearly all snaps. Computers tend to simply look at results and not factor in specific player absences or lineup changes. The odds of a 9.5-point favorite winning is 75.8%.

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

Take a look at this list:

  1. Alabama
  2. Florida
  3. Clemson
  4. Virginia Tech
  5. BYU

Those are the current top 5 in offense efficiency according to FPI. The other four have QB's that are getting plenty of Heisman attention, and Hendon Hooker might also if he had played more than 6 quarters and wasn't being potentially overshadowed by the nation's leading RB. At this point in the season, Khalil Herbert has done everything needed to be a Heisman candidate but will likely never sniff a ceremony invite, even if he continues this pace, because a running back for a team not pushing for a playoff spot just has too many cards stacked against them. Unfortunately Herbert's only shot involves Virginia Tech winning out and the back not just maintaining but accelerating the current pace.

Three thoughts about his explosive start to the season that I keep coming back to:

  1. This is near or at the top of the best offensive lines Tech has had.
  2. Top-end speed is one of the most overrated stats for running backs (ban the 40!), and the elite backs do three things better than the rest: have great vision and patience, explode both laterally and forward, and break tackles. We've seen flashes of some of those from backs in the last few years but it's been a decade since Virginia Tech had a back that could do all three (see Williams, Ryan). Enjoy it.
  3. Kansas has Khalil as a running back and opted for someone else to get the carries. Either Herbert had some breakthrough in the last 12 months or Les Miles and his staff are completely incompetent at assessing talent.

What were we talking about? Oh yeah, this is a Wake Forest game preview. The Deacons have a better offense than defense but both units lag the Hokies.

Who To Watch Out For

In terms of defensive players who could potentially stop the Virginia Tech rushing attack, I thought the linebackers at BC posed the biggest threat and ultimately Herbert and Hooker didn't seem slowed down. No one on the Wake Forest defense appears to be any kind of game changer to me, so instead let's turn our attention to whether there are offensive or special teams players that might keep the Deacons in a shootout:

  1. RB Kenneth Walker III is tied for 4th nationally with 7 TD and 24th nationally with 95.75 YPG on 5.72 YPC. (FWIW: Hendon Hooker has 4 TD's in just 6 quarters and 96.5 YPG on 7.42 YPC.)
  2. QB Sam Hartman is 19th nationally at 8.9 YPA but has struggled to find the end zone in the air with just 3 TD's on the season, resulting in a QB rating of 146.91. (FWIW: Hendon Hooker also has 3 TDS but in just 6 quarters, and an 8.8 YPA resulting in a QB rating of 173.75 - 5th nationally if eligible.)
  3. WR Taylor Morin is 3rd nationally with 14 yards per punt return. (FWIW: Hendon Hooker has no punt returns on the season but would definitely average more than 14 YPR.)

Statistical Key to the Game

The number of Deablos and Wallers cleared to play.

Although I don't anticipate Tech remaining 4th nationally in offensive efficiency for the entire season, the success to date is no fluke and both Hooker and Herbert are legitimate threats. I don't envision Wake Forest winning this thing in a low scoring game. They're going to need to win a shootout and probably win turnovers to keep pace. Turnover luck benefitted the Hokies greatly against Boston College, and although that doesn't tend to swing the other way, it does tend to return to average over time. An even split of turnovers probably means Virginia Tech wins the shootout, but the margin in Wake Forest's favor can't be more than +1 as the Hokie defense probably can't be successful enough to overcome it.

Statistical Prediction

Wake is 4th nationally in giveaways per game and Virginia Tech is 9th, but both teams are above average in defensive takeaways per game; as a result, the teams are 8th and 9th nationally in turnover margin per game and at least one team is thus bound to underperform in this one. They can't both win turnover margin and even a 0 is underperforming.

With the Hokies keeping it on the ground more than any teams other than the service academies and triple option Georgia Southern, ball security is more likely to continue for Virginia Tech. Similarly, there is always hope that the team will finally field a full starting secondary and become a bigger interception threat. Both of these make me think Wake is the more likely team to falter in turnover margin, and this Hokie team is not one that will be easy to keep pace with if you give them the ball.

Virginia Tech 41, Wake Forest 30

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


I really look forward to these each week! Thanks for taking the time to do this.

I'm cautiously optimistic that the defense will begin to take some baby steps each week as the unit gets more practice time together. These next couple of weeks, thought not "lay-up" games by any stretch, should provide an opportunity for us to get better before really hitting the heart of the schedule.

VT 62 (Wake isn't interested in defense)
Wake 35

Is coronavirus over yet?

Thanks and I agree about the defense...the impact isn't just in players on the field on Saturdays but also practice time together.

62 with Herbert going for 200+ and 4 TDs would certainly start to get his name more in the Heisman conversation!

I'm liken' the way y'all are thinking. We need to start pressing for Herbert for the Heisman. His performance on a non-top 5 team makes his accomplishment all the more impressive. (Not slighting our MUCH improved O-line)

Yep, there's really no reason not to start pushing his campaign on free mediums like social media and other low cost measures.

Worst case scenario our run game falters against Miami and Clemson and the Heisman hype dies naturally. Best case scenario we have been pumping it up for weeks before those games so there is already more people thinking of Herbert as a contender as we run the ball effectively on Miami and Clemson.

He's the real thing as a runner, it will be how successful our O-line will be against MIA and CLEM.

I agree to start the "free medium" push!

It's nice when "Virginia Tech Leads the Nation In..." is actually a really relevant and exciting stat. You know, as opposed to "punts per game" or "explosive plays allowed per game."

Warning: this post occasionally contains strong language (which may be unsuitable for children), unusual humor (which may be unsuitable for adults), and advanced mathematics (which may be unsuitable for liberal-arts majors)..

Amen to that!

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

Love this column - thanks for another great write up!

Quick question which may have been answered at a different time, but. Why is the Odds of Winning against Clemson in chart one 12.1% but then the Chances of Beating Clemson in chart three 14.1%?

I could be wrong, but I think the winning percentages on chart 1 are based on historical odds of a team's winning percentage based on the spread. So, that formula just looks at a predicted spread for the VT-Clemson game. Then odds of a 16 point underdog winning are 12.1%. That's my guess.

Then chart 2 is based on FPI.

This is correct but how the hell did you know that?!

There is no logical reason for using two slightly different models though so I'll try to resolve for next week and stick with one to avoid any confusion.

which one is more accurate? do the models you use share their predictive ability?

πŸ¦ƒ πŸ¦ƒ πŸ¦ƒ

Both models are about equally accurate (not enough difference to matter) but for predictive ability, Vegas spread remains long term king even if some computers slightly outperform in the short term. There is a lot of information accounted for in it that is not included in computer models (for better and worse I guess).

I love that LOLUVa's chance of beating us is practically the same as our chance of beating clemson. Despite all the crazy of 2020, at least one thing has returned to the way it should be.

Warning: this post occasionally contains strong language (which may be unsuitable for children), unusual humor (which may be unsuitable for adults), and advanced mathematics (which may be unsuitable for liberal-arts majors)..

the reigning Coastal champs

Proud author of one plaid comment.

Would it help if I wrote it "the reigning coastal champs, who are currently projected to go 3-7 in the ACC this season,..."?

Maybe a little but seeing it just reminds me that all we had to do to be coastal champs was beat them and we couldn't.

Proud author of one plaid comment.

See, I thought that was hilarious to keep mentioning it. It just underscores how far they've fallen this year (or how much of a joke it was they won the division last year).

Warning: this post occasionally contains strong language (which may be unsuitable for children), unusual humor (which may be unsuitable for adults), and advanced mathematics (which may be unsuitable for liberal-arts majors)..

reigning Coastal champ UVA

He said it again!

Proud author of one plaid comment.

Although I don't anticipate Tech remaining 4th nationally in offensive efficiency for the entire season

Me either. I am expecting them to takeover the #1 position in this category.

I'm surprised we are expected to beat Miami. Have to say that if every individual game goes the way it is predicted to go in your top graph, it will be a great regular season. Know I'd be super pleased with 9-2 and the 2 loses being UNC and Clemson. Add in the bowl game and my thinking is:

10 wins (including a bowl win) = I'm ecstatic; top 15 finish and maybe even top 10 based on what other teams do
9-3 = pleased; ranked
8-4 = depends
7-5 or worse = season is at best a disappointment and at worst a failure.

Recovering scientist working in business consulting

tbf, the miami game prediction should be considered a toss-up.

I would guess that anything within 33% to 66% (i.e., middle 1/3rd) should be considered a toss-up. It would be interesting if anyone has tracked the wins and losses and compared with the predicted percentages to see the model performance.

πŸ¦ƒ πŸ¦ƒ πŸ¦ƒ

So, based on the boydsbets results, the spread for the Miami game should be around -3 pts to yield 58%, but I'm wondering how Vegas (or any spread setter) gets to that spread.

I would have thought these percentages would be calculated using Bayesian statistics with validated predictors. Bayes , however, predicts percent likelihood of an outcome, not a spread of points. One can then predict a spread from a predicted percent likelihood of an event by converting the likelihood into a spread of points based on correlation. But it sounds like that Vegas does not do it that way, but instead first predicts a spread that that then can be converted into a predicted likelihood.

Do you have any insights on how Vegas predicts spreads?

πŸ¦ƒ πŸ¦ƒ πŸ¦ƒ

Well, this is probably a topic covering more than one beer but unless you happen to live in north Texas this forum will have to suffice.

The Vegas oddsmakers aren't really using any statistical prediction, Bayesian or otherwise, and actually don't even care what the expected score outcome is. Their REAL job, which thanks to the wisdom of crowds makes for a pretty good predictor of outcome, is to predict the point in which 50% of the money will get bet on one team and 50% of the money will get bet on the other. As long as they do that reasonably well the casino makes money because the payout isn't 100% of the bet for the winners. So if I set a spread and bets for Team A and Team B are both $1000 total, I win as the casino because I took in $2000 and will pay maybe $1900 out to the winners. Doesn't really matter whether I thought they were more likely to win or not.

When too much money is getting bet on one side, the spread shifts to entice more bettors to the other side. So a shifting spread is a sign that either new information has come along to bring more bettors to one side than the other, or just wasn't set well in the first place. That is a risk for a casino, because they occasionally will get burned by a shifting spread where the final score sits between the starting and end point and they have to pay out too much money.

In an event, they are really good at initial spreads, and while there are plenty of stupid bettors out there the combined wisdom tends to have settled at a really good estimate of final score.

The stupid bettors are on both sides so they even themselves out.

Hmm... Yea. I know they shift spreads based on keeping bets 50% on each side, but I guess reaching that initial number is the trick I'm wondering about. I would have thought they use some statistical inference, but I guess not. It would seem they would be often inaccurate in their predictions otherwise, but I guess the data bears out to show they're good at it.

Thanks for the convo. I do geek out about these things a bit, so I truly appreciate the responses.

πŸ¦ƒ πŸ¦ƒ πŸ¦ƒ

No excuses, we should run clean through this Wake Forest team.

lets do it then!
Herbert 19 carries for 203 yards and 3 TDs
Hooker 8/14 for 132 yards and 2 TDs with 21 carries for 165 yards and 2 TDs
QP adds 32 yards on 8 carries and 1 TD through the air on 2/5 passing in the 4th
VT blasts WF 59-31

Onward and upward


You left out Blackshear's 6 carries for 75 yds (he's bound to have a break out game, soon).

that seems like a pretty heavy rushing load on Hooker

Herbert 16 carries, 165 yards, 1 TD + 2 receptions, 61 yards, 1 TD + 4 returns 145 yards (371 apy, 2TD)
Blackshear 13 carries, 110 yards, 1 TD + 1 reception, 15 yards
Hooker 12/19 for 195 yards 3 TD + 12 carries 98 yards, 1 TD
Burmeister 2/3 for 45 yards + 4 carries 70 yards, 1 TD
Turner 4 receptions 100 yards 1 TD + 1 carry 25 yards
Robinson 4 receptions 75 yards 1 TD + 3 returns 30 yards
Mitchell 1 reception 22 yards + 1 carry 4 yards, 1 TD
Johnson 2/2 FG, 8/8 XP

total: 5 rushing TD, 3 passing TD, 2 FG (62 points)

But what about TOP?

Have you entered your PICK 6 on CBSSports yet? Click Here!

a pretty heavy rushing load on Hooker

Warning: this post occasionally contains strong language (which may be unsuitable for children), unusual humor (which may be unsuitable for adults), and advanced mathematics (which may be unsuitable for liberal-arts majors)..

That's 472 yards....seems a bit ambitious.

Yeah, Wake's defense probably does that 1/10 matchups. Figure Holston and King will gets some carries to round out a nice 500 yard rushing day

Les Miles and his staff are completely incompetent at assessing talent.

54-27 Hokies.

WR Taylor Morin is 3rd nationally with 14 yards per punt return. (FWIW: Hendon Hooker has no punt returns on the season but would definitely average more than 14 YPR.)

You're not wrong.

Let's GO!!!!!!!...

At this point in the season, Khalil Herbert has done everything needed to be a Heisman candidate but will likely never sniff a ceremony invite, even if he continues this pace, because a running back for a team not pushing for a playoff spot just has too many cards stacked against them.

I think the road to getting an invite, beyond obviously producing at a high level and staying healthy, is to have a stand out performance against Clemson. If he balls out and has a few 'heisman plays' against that talent then I think it will be hard for anyone to turn a blind eye towards him.

odds of a 10 win season at 2.03!!

Having a 91.7% chance of beating liberty scares me a little. Screams trap game and they haven't exactly been playing poorly...

Relax. Just sit back and watch and enjoy the moment. They can handle it.

You preach the truth. I have faith!!!

Reel men fish on Wednesdays

Last year I got yelled at by some guy in south endzone at the end of the wake forest game. He yelled at me to shut up and I yelled GTFO then he strangely reviled a VT shirt under his wake forest shirt. It was the weirdest shit and still makes me mad. Anyway... go Hokies... let's beat their tails again

Hokie Club member since 2017

I can't even figure out which direction to genuflect.

This is going to be great for the ACC.

I hope the defense starts to shore up some more. I would consider it a disappointment if wake scores 30 on us. Just because. 21 if their last touchdown is a late meaningless one is ok....but I don't know....30 to wake is a lot.

If you don't want to recruit clowns, don't run a clown show.

"I want to punch people from UVA right in the neck." - Colin Cowherd

Wake has played 4 games. Here's their opponent and what Wake scored on each:
Clemson - 13
NC State - 42
Campbell - 66
They are certainly capable of putting up points. I feel pretty confident that we can put up a lot on them though and now that we're getting more of our defense back, that will really help. If they score 30 on us but we score over 40 on them, I'm okay with it. I'd rather our defense hold them to 14, but not going to hold my breath on that.

I took the over so I certainly hope we don't hold them to 14. I don't know that our offense will hang 55 on them (though that would make for a really enjoyable game).

Warning: this post occasionally contains strong language (which may be unsuitable for children), unusual humor (which may be unsuitable for adults), and advanced mathematics (which may be unsuitable for liberal-arts majors)..

You gotta realign your mindset my friend. We aren't the Hokies of old. Giving up 30 to Wake isn't the unthinkable outcome it was a decade ago.