This is my last stats preview for the 2021 season once again looking at the Georgia Alabama matchup. Much of looking at football through the prism of stats, advanced or traditional, is couched by “small sample size.” Every sport has variance, but with more data, there is less variance. But, even with a 162 game baseball schedule a team can buck their trend, get hot and bring home a World Series.
I listen to a lot of football podcasts and many of them, not all, talk about analytics. On the most recent episode of one of my favorites, Bet The Process, host Jeff Ma (the main guy in the MIT Blackjack Card Counting Team) talked how using analytics for football play calls has been taking a beating of late. He brought up an important point to consider when looking at analytics - in anything from business and finance to football and sports - that analytics isn't a magic bullet. Analytics is used to gain small egdes. The more edges you gain, the more likely you are to have success. Counting cards in blackjack isn’t going to make you rich winning one hand. A card counting blackjack team gets rich by using the edges to win over a long extended time.
That got me thinking at the Georgia QB situation and maybe the coaching staff’s decision to start Stetson Bennett over JT Daniels. I am certain Monken and Smart are not using statistical analysis to decide who starts, but they probably view Bennett’s upside is lower variance. When you have the best defense in the country, the QB who won’t turn the ball over is better this team. Smart doesn’t plan on winning a shootout, which is a game that probably has a lot of variance. The “guy who gives [Georgia] the best chance to win” is the one who can be predicted to not throw picks and use his legs. Ok enough of that. Let’s revisit some of metrics in the rematch.
Georgia led the county in defensive success rate allowed most of season, only Wisconsin had a comparable defensive efficiency. Let’s look at the game by game output of the Georgia Defense.
Well, if we look at the Alabama offense by game, we can see a lot of extremely efficient offensive games, but also games versus LSU and Auburn with a low success rate and with it a lower score in those games. And let's go to the Texas A&M game, Alabama had a similar offensive efficiency to the Georgia game but a much lower EPA. This indicates that while the Tide moved the ball well, it did not have a lot of explosive plays.
For the season, Alabama average 35 successful plays a game. Against the A&M they ran 39 successful plays on offense, but only 12.8 YPP. Against Georgia, Alabama had 33 successful plays but with 15.3 YPP. That was their third highest explosive YPP average of the season behind Southern Miss and Arkansas. What’s all this got to do with winning a game? Statistically speaking, successful plays keep offense on schedule, ahead of the chains, and put points on the board.
This is my own stat that I have looked at. Specifically to attempt to predict a score. But it certainly tells a picture of these efficiency stats. How do the successful plays add to points on the board? Georgia allows just .4 points per successful play allowed on defense for the season. Alabama is excellent in this stat, too, allowing just .6 points per on defense.
Offensively, Georgia an Alabama both put up 1.2 points per successful play on offense. Which is what Alabama had if you include the pick 6, but 1.0 is what they did against good, not great defenses. Georgia can return to form limiting successful plays and limiting points. Georgia’s offense actually ran more successful plays and had a higher success rate in the SEC Championship, but those 26 successful plays led to only .67 points per. This was the lowest of the season and what Alabama allowed to good, not great offenses. Turnovers in redone and pick 6’s can beat the trend of what a team has done all season. See above about variance.
We are into the smallest of sample sizes now. One game to win it all. I prefaced this post with the face that football stats are very hard to use a predictor. The bettors and predictors plug in as many stats as they can, but always add qualitative analysis, too. Those stats, to me, show that Georgia played it’s worst game of the year in December. I just don’t see that happening twice.
Sportsbooks use many of these same stats to hang a line of a game. The look ahead line immediately after the Bama loss? Georgia -1. After the Semis? Georgia -2.5. I think the books agree that Georgia is a better team than the won who lost to Alabama.
This post is one of a series previewing the Georgia/Alabama CFP National Championship Game brought to you by the folks at DraftKings. Odds/lines are subject to change. Terms & conditions apply.
See draftkings.com/sportsbook for details.