QB sneaks seem more successful on inches-to-go situations than RB dives. We'd like to know if the data back this up. Unfortunately, the play descriptions don't note how long the 'and 1' is, whether it's a long yard or just inches. We'd expect to see more QB sneaks on the shorter distances and more RB dives on the longer distances, which bias the numbers because longer to go distances would naturally be tougher to convert. Still, we may be able to draw some inferences.
The table below lists the success rates for 3rd and 4th down runs with 1 yard to go. It breaks out plays by QBs, RBs, and FBs. QB scrambles on pass plays have been removed. Kneel downs and spikes are also removed. Plays inside the 10 yd line are removed due to field compression effects.
Conversion success rates on 1-yd to go runs (%)
Position | 3rd Down | 4th Down |
FB | 77 | 70 |
QB | 87 | 82 |
RB | 68 | 66 |
Total | 72 | 72 |
The next table lists the frequency of each type of play by down and position.
Frequency of play type on 1-yd to go runs (%)
Position | 3rd Down | 4th Down |
FB | 10 | 8 |
QB | 16 | 31 |
RB | 74 | 60 |
At first glance, it appears QB sneaks are underused and RB dives are overused. It's also notable that the QB sneak is used significantly more often on 4th down than on 3rd down. I suspect this is from the selection bias--4th down attempts are more common on shorter "1 yard" to-go distances because teams would tend to kick more on the longer to-go situations. 3rd down doesn't suffer the same biases, except perhaps that teams might choose to pass on "longer yards" and run on inches to go. Consequently, we can't draw any definitive conclusions about the relative effectiveness of the tactics.
Or can we?
Data from 2-yd to-go distances can help clarify the issue. The one thing we know about 2-yds to go is that it's always longer than 1-yd to go. I realize how simplistic that sounds, but it's an important point. Note below that QB sneaks on 2-yds to go are just as successful as RB dives on 1-yd to go!
Conversion Success Rates on 2-yd to go runs (%)
Position | 3rd Down | 4th Down |
FB | 60 | 50 |
QB | 80 | 62 |
RB | 58 | 56 |
Total | 60 | 57 |
Actually, 2-yd QB sneaks are far more successful than 1-yd RB dives on 3rd down and slightly less successful than 1-yd RB dives on 4th down. Taken overall, this suggests that the QB sneak is an underused tactic as far out as 2 yds to go. It also explains why running on short yardage is more successful than passing--it's the QB sneaks that make the difference.
Note that there is still bias in the data. Teams with strong lines or quick, strong QBs will tend to trust the sneak, and they're also the teams that would be most successful doing so. But the bias runs both ways. Teams with powerful RBs would call their number more often and be more successful doing it. I think we can infer from this data that, whatever coaches believe about the general potential for success of the two tactics, they underestimate the QB sneak and rely too much on the RB dive.
My thanks to commenter Frank for the insight.
Great article again, thanks Brian and Frank (for the insight). One of my favorite things about your articles, Brian, is that you use SIMPLE logical steps to build interesting conclusions (I can definitely take this one to the bar with me). I also like it that you do your best to eliminate arbitrary factors from your stats. Seems like that's why you haven't released your LI stat yet. I'm happy with that, wait until it's up to par with the rest of the site.
Thanks, Ian.
One other point to draw out is that the FB plays are somewhere in the middle between sneaks and RB dives. Success might just be a simple function of the time it take to get the ball over the line of scrimmage.
mmmm, I think this was discussed earlier, that not all QB runs are created equal which may also do some explaining in why runs are more successful than passes. How many of the 2 yard QB runs are dives right up the middle, and how many are draws, run-pass options, or scrambles? Run-pass option plays and scrambles are the sticky wicket, because if they succeed they count as a successful run, but if the run is not there the QB may throw an incomplete pass or take a sack, either one of which would count as an unsuccessful pass. These types of plays would tend to inflate the QB run's success rate, particularly on 4th and 2.
Oh and I totally missed the part where it said QB scrambles on pass plays are removed. Me, fail.Still wonder what percentage and what success a QB draw has in getting 2 yards as opposed to just plunging into the line.
Interesting post, Brian. Good way to analyze the issue given the limited data available.
I thought I'd throw in some thoughts. I'm not sure if your conclusion is necessarily supported by the data. Assuming that QB sneaks are more successful, on average, than RB dives, does not necessarily imply that teams should run fewer QB sneaks and more RB dives.
QB sneaks are a little like onside kicks; they're very successful when they're unexpected but much less successful when expected.
To know whether or not a HB dive or QB sneak is the right play, we need to know the marginal likelihood of success on the next attempt, not the average on all attempts. This all assumes -- which may be faulty -- that if defenses know the exact playcall, a HB dive is a better call than a QB sneak. But if that assumption holds true, that would mean there is an optimal mix between percentage of times to run a QB sneak and HB dive. Once you run the QB sneak too frequently, it becomes counterproductive.
Brian,
I appreciate you're taking the data available to the public to see what if anything the data suggests.
As far as information that could be used to make a decision or to judge a decision, it seems to me we lack the necessary data.
NO lined up in an odd formation with DTs in both A gaps. (ATL's OC went *backwards* after the snap.)
I would want to know WP or success rates for QB sneaks against that specific formation (not to mention other fronts, like even fronts). The difference in success of a QB sneak against NO's sort of Gap 8 defense vs a 5-4, 6-2, etc formation could be huge.
The top-level data incorporates every formation and play call of course, but I have to question whether it applies to the specific defense NO put on the field for yesterday's play. I guess if you assume every defense uses the exact same formation in those exact circumstances...but NO really sold out on the A-gap run first and any kind of power run second.
(Check out the video: NO put ELEVEN men in the box. Let's not forget it wasn't just Mike Smith making a decision, Gregg Williams did, too, and as is his reputation he made his best guess and committed his entire defense, no hedging.)
Interestingly to me at least, it appears to me the same reasons a QB sneak might have failed are the same reasons the play they actually called did fail.
The play appeared to be a long trap to inside of the left TE, not a dive between the tackles.
This call failed in large part because NO tried first to take away plays through the A gaps; in executing that tactic, NO's DL penetrated so deeply the pulling OG could not get to his assignment; NO's safety shot the play gap and made the tackle in the backfield.
Just my two cents. What am I missing?
Good point chase about defensive expectations. A theme that I see developing at this site (run vs. pass, onside kicks, etc.) is that you can't be predictable. While using a QB sneak may decrease the chances of the following one working, you increase the chances of a HB of FB run (or any kind of pass) working the next time. As you eluded to earlier, its about picking the one that the defense isn't expecting. As we saw with the Wildcat craze and onside kicks, they're successful when the Defense isn't expecting them. There may be some intrinsic advantage to one play vs. another (i.e. player skills, risk) but the defensive expectation can nullify it. IMO, the number of situations where an NFL caliber defense knows what's coming and can't stop it are becoming few and far between.
This might be a tangent into another topic, but it put it out here anyway. Previous articles mentioned that NFL coaches minimax based on SR rather than raw EPA. I believe coaches do this to stay out of predictable situations. A successful play does more than just earn EPA for a team. It forces the defense to react and compensate to keep the offense from exploiting that type of play. This compensation (8 in the box, nickle defense) opens up other holes for the offense to exploit. It also keeps all of your options available. There are a lot more plays in the playbook of 1-3 yards than there are for 10+. While most coaches focus on staying out of obvious passing situations, they may also want to be aware of obvious running/QB sneak situations. Keeping the defense guessing seems to be a key to success.
Despite attempts to stay in manageable down and distances, teams will inevitable get behind. This usually manifests itself in 3rd and longs. It is at this point where being able to pass the ball, even when the defense is expecting it becomes incredibly valuable. It allows teams to make up for mistakes or poor plays in a way that the running game cannot. This could help explain why passing is so strongly tied to success. If the defense can't punish you when you're in a predictable situation, they're in big trouble. I would argue that this is even more important than being able to run the ball on an expecting defense to run the clock out, because obvious passing situations occur throughout the game whether ahead or behind.
So to sum this up, to be predictable is a death sentence. To overcome predictable is what champions are made of.
One reason why QB sneaks may be underused (relative to RB/FB runs) is that it puts your most valuable player at greater risk. The increased succcess rate may not be worth the risk.
Good point above about the QB draws. They appear the same as sneaks in the play by play. Scrambles are different.
I wonder if there really is much of a game theory component to this. I think the QB sneak may be the 'dominant strategy.' Tell me one situation when defenses are *not expecting* a sneak on 4th and inches or even 4th and a full yd. And how would the defense's strategy change anyway?
Maybe it's a good place to go play action for a big pass?!
Also, not that the ATL-NO play sparked the interest in the topic, but I'm not necessarily claiming this analysis proves ATL should have called for the sneak in this particular situation, just that the sneak is probably underused in general.
Brian,
I would say that defenses almost always seek to take away the two A gaps first, but that doesn't mean they're going to sell out their entire defense for the A gaps.
In other words, how many choose to defend, what, an 8-foot by 8-foot square to the exclusion of the rest of the field? That's conceding a lot of space.
Most will hedge against something outside or especially in the secondary. They'll have a short yardage defense *and* a goal line defense. And yet Williams it appears to me totally sold out for a run between the TEs.
You wrote of removing plays inside the 10 due to field compression, but Williams' call essentially bet ATL would play as though on a compressed field.
My my, it's like he used a different set of data to make his call!
Mark
I wondered while waking up if there may be selection bias wherein bigger, "sneak"-ier QBS are more likely to go for it. Then I woke up completely and remember you made the point about better RBs having the same effect. But in the CFL, going for it on 3rd* & 1 is very common. The thinking is, the risk is low because the defense lines up a yard behind the line of scrimmage. Very frequently, they'll bring in a backup QB to take the dive.
I wonder if it would be beneficial to have a backup QB specifically for short yardage situations, or even a short yardage "wildcat" formation in which a fullback lines up under center. I'm not sure anyone could stop Brandon Jacobs from diving forward 8 inches to pick up a 1st down, even if the defense lines up the way the Saints did in yesterday's game. A big backup/3rd string QB type would, instead, allow for the possibility of an unexpected passing play.
* Because there are only 3 downs in the CFL.
I'm with you, Brian. I think the QB sneak may be the dominant strategy. But I'm not sure, because it certainly isn't used as frequently as one might suspect. It may have to do with QB injury worries. I believe that is where Schaub hurt his foot yesterday.
This is pretty interesting, but I was hoping to find some data backing up Smith's decision to go for it in that situation, regardless of which play is called. It is due to analyses I've read on this site in the past that I've become a vocal of going for it on fourth and inches; but in this specific case I'm hearing a lot of talk that this was the wrong situation, in overtime from his own side of the field against a team with a good kicker. Is there enough data out there to say definitively that it was the right (or at least a reasonable) call?
Alan you're looking for the post 2 down from this one. Here you go:
http://www.advancednflstats.com/2011/11/falcons-4th-down-in-ot.html
Suggestive: The Patriots seem to run a lot of QB sneaks on 3rd and 4th and very short.
Maybe I never paid close attention to it, but do all gamebooks really make note of quarterback sneaks versus scrambles? I would have imagined in some cases that stuff would be left out, making it difficult to really weed out that data.
Some gamebooks are more detailed than others, such as the ones by Buffalo and Dallas.
Do you have turnover data for those snaps? The thing about a QB Sneak, especially the 1-inch version of it, is you end the play with the ball more exposed. Sure RBs reach out as well but the typical "sneak" play is designed to build a crease for the QB to extend the ball forward, away from his body, where it is more exposed to being knocked away. The dive play generally ends with a RB pad-first into a LB -- perhaps not as much of a fumble risk?
How any team in the NFL can't successfully gain 1 yard is still completely beyond me. Intuitively, for me anyways, QB sneak seems to be the better option and this article just solidifies that belief for me. It's a pretty simple concept...6'+ (>2 yards) tall QB falls forward with his enormously tall & heavy linemen. Even if the defense is anticipating it how could they possibly stop that consistently? It's physically impossible and inevitable the QB will gain at least a yard, simple. Although injuries and fumbles are definitely important factors to consider.
Eliminate the strategy and bank on simple physics.
One has to remember though that if a DT is directly over the center than the QB sneak is usually not a good play call. Its success really depends on the defenses alignment.
In Canadian football, it is common to substitute in the second-string quarterback for the sneak attempt.