Interesting Bill Connelly article on explosive plays in CFB
Bill C., the advanced stats guru behind S&P+ and Five Factors, posted a new article today talking about big plays in college football and how to best achieve them.
If you can't read it, he basically says the following:
When he initally came out with his five factors in January 2014, he said: "Big plays are probably the single most important factor to winning football games."
His reasoning was that while a 20 play 80 yard drive was the ideal outcome, they rarely happen. When offenses, especially in college, are on the field for that long they tend to make drive killing plays (turnovers, penalties, big loses, etc.) more often than not. Big plays reduce the amount of plays an offense is on the field, reduce chances for drive killing plays, and are thus paramount to a team's success.
While the general idea still remains, he walks back a good amount on that in this article. He draws a lot of comparisons to Ken Pomeroy's thoughts on 3 point defense where KenPom says the best 3 point defense is preventing 3 pointers from being taken as 3 point percentages are random as a whole.
In this case, Bill C. shows that according to his data big plays happen at the same rate no matter the down and distance. With this in mind, and looking at his data for Success Rate, PPP and isoPPP, he comes up with the following major conclusions:
"If big plays can happen on any down at any rate, the key to creating explosive plays is an efficient offense. The more plays you're on the field, the larger chance you have of creating big plays."
and
"Efficiency is everything in CFB. Explosiveness is too random to rely on without efficiency. Unless you're Penn State anyway"
While this seems pretty obvious in theory, the data behind it is very cool and it's a big change in philosophy from his old idea of "efficiency doesn't matter, just make big plays."
That last sentence about PSU is also a mini shot at them, supporting any PSU naysayers who think their offense last year was too random/luck oriented to sustain success at the same rate this year.
August 22nd, 2017 at 12:25 PM ^
do you have TL;DR version of your summary?
August 22nd, 2017 at 12:27 PM ^
Efficiency be good. Penn State bad.
August 22nd, 2017 at 12:28 PM ^
thanks
August 22nd, 2017 at 1:46 PM ^
Is this why all the peeps get up and yell and clap on dem big plays?
August 22nd, 2017 at 12:29 PM ^
Still digesting this, but the thesis would seem to play into Harbaugh's philosophy, which doesn't seem to focus as much on getting explosive plays as just being efficient and moving the ball (I think he has said as much in public, but I can't say where). Put enough plays together, though, and Karan Higdon or somebody will take one to the house.
Oddly, our D wasn't great at the big play metric last year, relative to the rest of the numbers the defense put up. I wonder what the correlation is?
August 22nd, 2017 at 12:32 PM ^
Obviously Peppers was overrated.
#Staee'd
August 22nd, 2017 at 12:49 PM ^
#Clevelandbrowns
#FansofNFLteamswhodidntdrafthim
August 22nd, 2017 at 1:00 PM ^
isoPPP (the big play stat), is calculated by only including plays deemed successful (or success rate). Success rate are plays are defined as follows: 50% of yardage gained on first down, 70% gained by second down, and 100% gained by third and fourth down.
Michigan's defense was so good at preventing successful plays (1st in the country at only 29.5% of all plays), that a couple of huge chunk plays early in the year against UCF and Colorado crippled their isoPPP for the year as they weren't adding any sample size to reduce that total as the year went on. They still managed the sneak into the top 25 in that stat after being outside of the top 100 for a lot of the season.
August 22nd, 2017 at 1:18 PM ^
I think our defense did what Bill C is suggesting defenses should do. Be aggressive and focus on getting the offense off the field quickly rather thana bend but don't break and hope they make a mistake passive defense. Occassionally you'll give up a big play but the payoff is the opposing offense having less snaps which results in a better outcome the majority of the time.
August 22nd, 2017 at 2:25 PM ^
The book on Don Brown, for what it is, is that big plays can be had if you're ready for him. He just makes it really really hard to be ready for him, and from what I've read, this has always been a thing.
It's basically a calculated risk and a stark, counterintuitive departure from his peers. The conventional wisdom is that good defenses are aggressive because they're good, not that aggression makes a defense good. Big plays look & feel bad, which is why for anything short of desperation, defenses will yield the space they feel they absolutely need. For mediocre defenses the corners might give a ten-yard cushion and safeties over the top while the linebackers read & react. A good defense will be in your face and crash the safeties because they're confident they can get away with it. That is, defenses will generally play as aggressively as they feel they can get away with, and if it's a good defense, that'll be pretty darn aggressive. If they get burned too many times they'll back off, play bend-don't break.
Brown starts with aggression and never stops, so he always incurs the risk of the offense burning him, to the point that his first-year defenses often aren't much better than the disasters he replaced. Once he's implemented his system though, you can burn Brown once or twice, and teams have, but over the long run his defenses will be stingier than the most efficient bend-don't-break defenses. That he does this fully prepared to eat a long TD or two. . . that's some combination of nerves and headiness. He puts a lot of pressure on himself to make sure it pays off; the first half of last season was surprisingly dramatic. I don't think I've ever seen a DC overcompensate the way he did after UCF.
August 22nd, 2017 at 3:57 PM ^
This also sets-up your offense for success by creating better field position and more opportunities. By attacking the opposing team as aggressively as you can, you are creating the opportunity for more big plays on defense (to kill drives sooner) and more big plays for the opposing offense (man-to-man blitzes mean you are going to surrender some big plays). Either way, you are likely getting the opposing offense off the field more quickly, giving your offense more chances to be successful.
Probably without knowing it, Don Brown is the first advanced analytics football DC.
August 22nd, 2017 at 12:33 PM ^
This is quite an interesting take. Basically stay on the field as long as you can because any given play gives you an X% chance at a big play. But trying to get big plays seems like it would be important to getting big plays also. Penn State is a perfect example of this.
If we never toss a ball deep, our chances of getting a big pass play have to go down, right? Or does the risk of efficiency loss from a failed long pass play cancel it out?
August 22nd, 2017 at 12:37 PM ^
What if we just run for 9 yards on every first down, then throw deep on second? I'm not a smart man, but I think that's a winning formula :)
August 22nd, 2017 at 12:38 PM ^
If we never toss a ball deep, our chances of getting an average pass play go down, too; the defense is not stretched vertically and there is less space for the receivers to operate in. Even teams with heavy emphasis on a short pass philosophy run deep routes, because those deep routes force the defense to keep men deep, away from the rest of the play.
August 22nd, 2017 at 12:46 PM ^
I think this analysis obviously has to be taken in the context of running a normal offense with constraints. Every offense is going to have receivers run fly routes or deep digs as part of their normal offensive attack. I think the analysis is showing that you are better off running more 5 yard slants and hitches and other high efficiency plays over more 4 verts despite the fact that 4 verts has a higher likelihood of producing a 70 yd td.
August 22nd, 2017 at 12:54 PM ^
August 22nd, 2017 at 1:06 PM ^
August 22nd, 2017 at 12:39 PM ^
August 22nd, 2017 at 12:53 PM ^
August 22nd, 2017 at 1:11 PM ^
August 22nd, 2017 at 1:54 PM ^
You're looking at it wrong. This is looking at stats that already happened. A team that didn't make the drive-killing play stays on the field longer and was more likely to have a big play. A team that did make the drive-killing play does not contribute to the explosive big play stat because they were not on the field long enough to do so.
You seem to be suggesting that a team that gets off the field quickly is less likely to make a drive-killing play. But they got off the field quickly because they made a drive-killing play...
August 22nd, 2017 at 1:10 PM ^
August 22nd, 2017 at 2:19 PM ^
Splitting into an odd and even data set would be an interesting way to do it for sure. But I think the scheduling issue is mostly covered by the success rate info he includes.
For example in 2015, the top 10 teams in success rate through the first 7 games had an average success rate of 51.5% and average isoPPP of 1.54. Through the last 7 games, those same teams had a success rate of 48.7% but an average isoPPP of 1.33.
So while he's not adjusting for schedule, you can see that those top 10 teams are still very succesful both in the first 7 games and last 7. There is a slight dropoff, which is probably due to scheduling as you mentioned, but it's a drop of 2.8% which is relatively small. The success rates through the first 7 and last 7 are very similar.
The change in isoPPP however is significantly larger. The drop in isoPPP for those top10 offenses was .21, which is about a 14% drop. So you get a 14% drop in isoPPP but only a 2.8% drop in success rate. So even when isoPPP drops, success rate stays pretty constant, implying isoPPP has little effect on success rate.
Do the same for the bottom 10 teams in 2015. Their isoPPP average inceases from 1.06 to 1.19 or around 12%, but the success rate increases only .6%. So despite being 12% more explosive, success doesn't even rise 1%.
And look at the difference between the top and bottom teams as well. For the first 7 games in 2015 the difference in succes rate is 18.5% and the difference in isoPPP is .48 or 31%. The last 7 games the succes rate difference is 15.1%, but the isoPPP difference is 10%. So a 21% drop in explosiveness leads to a 3% drop in success.
If you look at the defensive numbers you see the same trend. He doesn't adjust for schedule, but evenso the best offenses/defenses are consistently succesful all year long, but even as their explosiveness ratings vary, their success does not.
August 22nd, 2017 at 5:21 PM ^
Unless I missed it, he didn't specifically say how he split up the data. For this type of analysis you should probably split it randomly. If he split it chronologically (first half of season vs. second half) it could cause some issues, as you say.
August 22nd, 2017 at 8:45 PM ^
August 22nd, 2017 at 1:15 PM ^
August 22nd, 2017 at 2:26 PM ^
August 22nd, 2017 at 2:42 PM ^
I don't understand your point.
His point was that he used to believe the ability to create/prevent explosive playsis was the most important metric to winning football games. However now he believs being an efficient offense is more important as the correlation between explosive plays and success isn't strong, and that being efficient lends itself more towards explosive plays anyway.
He says nothing about good players making big plays, or the team with the better player wins. Look at his stats. Texas Tech had the 3rd best S&P offense last year. USF had the 7th best and the 8th highest isoPPP. WKU had the 14th best offense and 2nd best isoPPP. None of those teams are teams with "better players". It's as much about coaching/development as it is anything else.
August 22nd, 2017 at 4:45 PM ^
August 22nd, 2017 at 5:09 PM ^
One possible interpretation of this finding is that big plays are the result of defensive breakdowns (mistakes) that the offense has little control over.
I find it a little hard to believe that the offense can't do much to control the frequency of big plays though. Surely if you throw more bombs compared to runs up the middle, your big play frequency will go up?
August 22nd, 2017 at 5:25 PM ^
Just score. Just prevent other team from scoring. Just win.
August 23rd, 2017 at 7:43 AM ^
Frank Beamer would disagree. No one should score... ever.
August 22nd, 2017 at 9:28 PM ^
Unless I'm reading this wrong (quite possible), it seems like the analysis ignores the effect of tempo, which by producing more plays would create the opportunity for more of them to be explosive. Especially with a dominant defense like Michigan's, it would seem to make sense to squeeze in as many offensive plays as possible.