South Carolina Week 2 analytics breakdown

Well, it wasn’t pretty, but the South Carolina Gamecocks pulled out a close win over Old Dominion in Week 1. While 1-0 is much better than 0-1, there are a ton of things to clean up for the Gamecocks and plenty of questions. I’ll address a few of those and look ahead to Kentucky in this week’s analytics breakdown.

Analytics Metric Check-In

South Carolina won 23-19, but it was ugly and the offense really struggled. Here are where South Carolina stands in a few national metrics as well as my thoughts on what happened Saturday at Williams-Brice Stadium.

Bill Connelly’s SP+

South Carolina tumbled in SP+ due to a few things. The Gamecock offense was buoyed last year by explosive plays, but South Carolina struggled in that area Saturday and it’s not like the offense was very efficient either.

All three of South Carolina’s units fell in the analytics metric, and though I’m not concerned about the defense (more on that in a minute), the offense and even special teams units have some kinks that need to be worked out quickly.

Offense: 26.3 (78th, down 19)
Defense: 21.6 (35th, down 5)
Special Teams: 0.0 (82th, down 70)
Overall: 4.6 (48th, down 13)

South Carolina had one of the sharpest falls of any team in the FBS, but I will give a caveat. SP+ historically has overreacted to Week 1 results. Those aren’t my words; they are a direct quote from Bill Connelly’s weekly SP+ column.

That reason is three-fold. First, there’s an art to winning ugly. That means that a team just might not have it one day in a weird result but still pulls out a win (Remember when the 2011 team nearly lost to Navy after beating Georgia?) No team can consistently do that and teams that try are usually just bad teams.

That being said, a ton of favorites struggled, including several teams we know (or at least assume) are very good teams. If South Carolina continues to struggle, it’s obvious the team just isn’t good. If the Gamecocks bounce back this weekend, we don’t have to automatically assume the Kentucky result is the weird one.

[Join GamecockCentral for in-depth Gamecock coverage and The Insiders Forum]

Second, since it’s all about how you win, the early results can be exaggerated, just like any stat. For example, if South Carolina hadn’t had the illegal man penalty, its explosive stats and points per scoring opportunity would be greatly improved. Across a season, one play doesn’t make a huge difference. In one game, a single drop, penalty or bad play can greatly skew the stats.

Third, we just don’t know a lot about any team so far. Every year we have dozens of early results that look significant and aren’t after a few weeks (Remember when people thought Colorado was amazing after beating TCU in Week 1 last year before realizing later that both teams were bad?) The opposite can be true as well.

SP+ is all about how teams perform relative to expectations. SP+ had ODU outside the top 100, but if the model thought more highly of ODU putting the Monarchs in the 70s (like the next model) the performance wouldn’t have dropped the Gamecocks nearly as much.

We see this already with Georgia Tech, which is only slightly up, per the analytics metric. Maybe Florida State is just bad, rather than Georgia Tech being really good.

Maybe a better example is SMU, who won by just five as a 28-point favorite in Week 0. The Mustangs are ahead of where they started, because it’s most likely that Nevada is much better than expected.

Still, it’s a big drop and SP+ doesn’t like the Gamecocks right now. The eye test somewhat agrees, but I’m not sure the sky is falling.

CFN Strength of Schedule Metric

This one’s weird, DM me if you want the full details. Basically, South Carolina’s schedule is harder now than it was preseason, because a few teams on it (Alabama, Ole Miss, Vanderbilt) look better than expected. Like most things in analytics, there’s math involved.

However, South Carolina is projected for 0.3 more wins than it was before the season started. About half of that is easily explained by South Carolina’s win. The Gamecocks won, so there’s no “x% chance they lose to ODU” factoring into the projected record. Before the season, South Carolina had an 84% chance of beating ODU, or 0.84 wins on the projected win total. Now that the Gamecocks have cemented a win, that 0.84 expected wins is 1.0 actual wins, so that accounts for the 0.16 wins. But what about the other 0.14 wins?

In College Football Network’s model — one I had a tiny hand in creating — win probabilities get updated after each week based on performance. Sure, three teams ate into South Carolina’s win probability in those games (Ole Miss, Alabama, Vanderbilt), but two of those represented two of the three least likely games for the Gamecocks to win, according to the analytics, while Vanderbilt represents the third-easiest game remaining on South Carolina’s schedule.

[Win two tickets to the South Carolina-LSU football game]

So, even after that movement, the Gamecocks remain likely to go 1-2 in those games.

A few toss-up games have really changed, though. Namely, South Carolina, despite its poor performance relative to expectation, saw huge jumps in win probability for three games: Texas A&M, LSU and Clemson.

Given that those games are closer to toss-ups, that’s great news for the Gamecocks. Steal a win against one or even two of those teams (two of the three at home) and there’s a great chance South Carolina has a successful season. All the metrics agree that those teams didn’t look great, even considering the competition.

Analytics Review: ODU

I promise never to make the first section that long again, so to make up for it, I’ll spiral some of this review into previewing Kentucky. Here are three things that I saw and whether or not they’re trends that will continue.

Havoc (Sustainable), Sacks/Fumbles (Unsustainable)

Pressures are the most consistent thing defensively. That plays into havoc plays, something South Carolina was dreadful at creating last season. The more havoc a team creates, the more issues it causes opposing offenses. South Carolina can absolutely sustain its havoc rate. I’d expect a bit of a dropoff, but nothing crazy. That means the Gamecock defensive line will be substantially better than last season, full stop.

[Join GamecockCentral for in-depth Gamecock coverage and The Insiders Forum]

Will Dylan Stewart reach 36 sacks, like he’s on track to do? Probably not, but he’s really good and should continue to cause havoc (hits, sacks, tackles for loss, deflections and interceptions in analytics)

The sack numbers probably aren’t sustainable as the Gamecocks were 12th in pressure-to-sack rate amongst teams that created at least seven pressures. Don’t expect the Gamecocks to get to the quarterback every one out of three times they get close and don’t expect a fumble every other sack either. But the pressure will stay (according to analytics) and that leads to good things for a defense.

Explosives on Defense (Potentially problematic, probably fixable)

South Carolina struggled with the big play in magnitude only (not in frequency) on Saturday. The Gamecocks only allowed two explosive plays (15-plus yard runs or 20-plus yard passes in analytics) but those resulted in Old Dominion’s only two touchdowns of the game.

Will those continue? I’m non-committal.

For the first, South Carolina got caught trying to audible against the third-fastest offense in college football. DQ Smith should have made the tackle, but the communication issue won’t be an issue against 90% of college football teams.

[Newsletters: Get breaking news and analysis from GamecockCentral]

The long touchdown run was bad gap integrity, something the Gamecocks never really struggled with outside of that. The bigger issue was that the defense was overwhelmed by the moment and got too aggressive. A play that could have turned the game into a relatively easy win turned into a play that nearly led to a loss.

There will be big plays in every game, but South Carolina has to make the tackle, because at 2.7 yards per play outside of those, it’s likely the Gamecocks wouldn’t have given up touchdowns on those drives.

LaNorris Sellers’ Accuracy (Fixable)

LaNorris Sellers was highly inaccurate Saturday. While the PFF graded him well, he only had a 55% adjusted completion percentage. It’s hard to believe that will continue. We can fault play calling or Sellers’ ability, but he seemed like he was trying to be too perfect.

College football columnist Matt Hinton has a theory that many young quarterbacks are better when a little reckless with the ball. Sellers looked the opposite, so focused on making the right read that he didn’t trust his arm. Sellers held the ball too long (More on that in a second) waiting for receivers to be wide open before releasing, leading to some late throws and some where he seemed to try to aim them rather than throw them. He can fix that.

Kentucky Analytics Preview

SP+ has Kentucky as an 11-point favorite in a 31-20 game. College Football Network’s Strength of Schedule metric has it much closer, with Kentucky as a slight 2.5-point favorite. The Gamecocks have a 44.5% chance to win, per the metric.

Here are two analytics stats I’m tracking for the Kentucky game.

Analytics Metric 1: Sellers’ Time Before Throw

Sellers held the ball longer than any SEC quarterback last week by a wide margin. He also had the second-lowest percentage of throws in under 2.5 seconds (13.3%) of any quarterback in the entire country.

However, Sellers was middle of the pack in terms of screen percentage. That tells us it wasn’t that the Gamecocks didn’t call quick throws, but more that Sellers took longer to throw. His average depth of target was above average, but not overly so.

There will certainly be an emphasis on calling plays that get the ball out of Sellers’ hands quickly, but he also needs to take the quick throws when they’re there. Quick doesn’t always mean short, and that’s something Sellers will learn.

Analytics Metric 2: South Carolina’s Pressure + Vandagriff’s Turnover-Worthy Plays

Brock Vandagriff was the opposite last week. He was largely accurate, but had several turnover-worthy plays. That analytics metric simply counts plays that should have been turnovers. Southern Miss dropped a few potential interceptions and caught another.

South Carolina needs to apply pressure on a team that allowed just four hurries last week. Vandagriff is prone to a bone-headed mistake or two and in what should be a low-scoring game, Dylan Stewart or Kyle Kennard forcing Kentucky’s quarterback to rush a throw could be the difference.

We don’t have a ton of data on these quarterbacks, so this will be interesting to track.

Discuss South Carolina analytics on The Insiders Forum!