Please allow me to introduce myself, I am a man with no wealth or taste…
Seems like the audience here at Dawg Sports takes to musical references, but for real, my name is Josh Hancher. I am a UGA ‘97 Grad and affectionately known as Dawg_Stats on the Twitter. I got into doing some blogging at the beginning of the 2019 season. Teamed up with Bulldawg Illustrated for the ’19 and ’20 seasons. Last year Graham Coffey, Nathan Lawrence, and I kicked off an experimental video show called “The Battle Hymnal” which took the analytics approach of the Chapel Bell Curve and paired it with some X and O analysis to break down the UGA season. This offseason Graham and I kicked off “Dawg Sports Live” here and have been creating content and analysis all offseason.
I know a lot of the readers have been supportive and watched most of the shows (thanks for that). For those that are just getting back to Georgia Football now that the pads are on, I wanted to introduce myself and offer up some background and a glossary of the terms that I like to track and share with Georgia fans.
If you have seen or read my previous efforts to explain football analytics, feel free to click “back” now. Thanks for the click. For those still with me, I want to share with you the baseline and explain the terms and stats I’ll share here, on Twitter, and on “Dawg Sports Live.” I often preface that using the analytics of football isn’t trying to reinvent the sport or convince anyone that wants to yell “RTDB” from the 600 level isn’t right and just. It’s just another way to look at and analyze the sport. And for many of you that want to dabble in a little betting – all of the sports books are using analytics to set lines. Many of those offering up pick advice will reference or have these stats baked into their picks. That is all to say, advanced analytics is becoming more and more common in football coverage.
There are three main ingredients to my analysis that I will share with you. Really it boils down to efficiency and how to show it. How you score matters. How you prevent scoring matters. If you read any of what Bill Connelly writes – these are all part of his SP+ nomenclature.
• Success Rate
• EPA (expected points added)
Success rate is the base formula for much of football analytics. It is “Yes/No” calculation on the success of a play. It is tracked as a percentage. On offense, a higher success rate is good, and allowing a lower success rate on defense is good. In Bill Connelly’s 2014 “Five Factors” – his numbers lay out that winning the success rate in a game will lead to that team winning 83% of games. This is a binary formula for “keeping the offense on schedule.” Conversely, on defense putting the offense “behind the sticks.”
A play’s success is
· Gaining 50% of yards to gain on 1st down
· Gaining 70% of yards to gain on 2nd down
· 100% of yards to gain on 3rd and 4th down
Simple right? And it makes sense. A 2nd down run for 4 yards on 2nd and 5 is “successful” whereas a 4-yard run on 2nd and 8 is not. And an 11-yard draw play on 3rd and 15 is again, unsuccessful.
And while a high success rate speaks to efficiency on offense, the defense wants to prevent and have allow a lower success rate.
A quick glance one can see the games that stressed us out. Arkansas was a cover, but an ugly one with a 38% success rate on offense but the defense was stifling allowing only 25% to an offense that was 45% successful on the season. Most UGA fans know how Florida game went, but a season low 25% success rate sure hammers it home.
I hope you are still with me, and if you are, this might push your patience with my stats. EPA stands for “Expected Points Added.” It’s nothing more than a number ascribed to a play’s value to the game. Higher is better for offense and opposite is true for defense. Think of it as a number grade for a play. Success Rate is a “Pass/Fail” whereas EPA will give you a larger scale to judge a play. Remember those report cards?
The nerds who paved the way on advanced analytics crunched a buttload of plays from the history of the game and devised a number for each and every play based on
· Field Position
EPA takes a play’s starting EPA (1st and 10 from your own 25 has an expected points added of .922) then takes the next play’s EPA and subtracts it. This gives you that 1st Down’s play EPA.
Let’s take a play from 2020 - as tempting as it is to use Zamir’s 25 yard TD against Florida (6.077 EPA for that) – let’s use a 1st and 10 from UGA’s 26 in the 4th quarter of the Florida game.
That down, distance, and field position has a starting EPA of
Mathis throws an incomplete pass intended for Tre McKitty. Brings up 2nd and 10 at the UGA 26. This 2nd and 10 play has an EPA of
So, the EPA is calculated for that play by subtracted the 1st and 10 .993 from the 2nd and 10 0.259 giving that play an EPA of -0.734
Ok, this calculation is not done on the fly nor does it need to be committed to memory. It’s just a mathematical calculation. So, when you see these numbers, you can understand some of the math behind it. This metric really begins to shape the efficiency analysis. EPA is often shown as an average, but the sum also is apt description of an offense and defense and their efficiency.
This really starts to mathematically show what we saw once JT took over. The Auburn game was about as good as one could hope, but compared to what UGA did on offense versus Mississippi State, Carolina, and Missouri – it somewhat paled.
Before we move on to explosiveness, let’s look at these two metrics and what they were for each team heading into the WLOCP
Offense EPA SR
Florida 0.374 50.2%
Georgia 0.117 41.8%
Florida 0.194 44.7%
Georgia 0.077 36.3%
It’s arguable to say that Florida’s offense was 3 times more efficient than UGA’s judging by EPA. And while UGA’s defense was outstanding it wasn’t 3 times as efficient to compenstate for the inefficiency on offense – not evening considering the cluster injuries on defense . Interesting that UGA was 2.5 point favorite. Most online books showed Florida and points had majority of the bets.
I did a quick video looking at two drives from the Alabama game that shows these stats in real time.
Explosiveness is a word that is often thrown around in media and coach speak. It can be subjective measure. For instance, I don’t think a triple option offense will judge offensive explosiveness in same manner as Todd Monken does. (in the preseason of 2020, he set the explosive play bar as percentage of 20+ yard pass plays and 12+ yard rushing plays).
Well, the analytics community universally describes explosiveness as the EPA of a team’s successful plays. This particular stat certainly needs more context than the previous ones in this post. Especially if I told you that UGA was second in the SEC in offensive explosiveness in 2020.
Think of this as “how good are your good plays?” UGA had their share of big gains, but remember from the “success rate” section, UGA was 7th in SEC. So, while their good plays were good – using the explosive metric – UGA needed more of them. If we sum this stat instead of average it shows a different picture.
UGA was near top in average yards per play on successful plays, but the fact that they had fewer than the elite offenses in evident this Total EPA measure. (For reference, the James Coley experiment left UGA 11th in SEC in this stat in average, and 1.5 yards per play behind LSU’s 13.9 in 2019)
Graham and I did a show about these topics, and they will a part of breakdowns this season on “Dawg Sports Live”
Well, if you have made it this far, thank you. I hope that you understand better some of these numbers bantered around. I’ll be referencing them quite often. Certainly, the traditional stats using yardage, points, and others will be in the breakdowns here and on Dawg Sports Live. I thoroughly enjoy discussing this stuff and welcome questions here and on Twitter. Go Dawgs!