I love safeties. They're one of the most exciting and interesting plays in football. But keep in mind they're not actually worth 2 points.
Jim Glass at the Community site: What clutch victories mean for a team and its coach.
Tyler Cowen on the economics of Moneyball. Phil Birnbaum's take.
An update to Michael Beouy's betting market power rankings.
Some interesting numbers on Denver's Tebow offense.
Here's an interesting article from The New Yorker on Jon Gruden, and stats are a part of the story:
But Gruden abhors the “stats sluts” who try to replace the judgments of a trained eye with mathematical formulas. He says, “You know what I hate, man? Guys that you know haven’t seen the film: they just quote a bunch of statistical bullshit.” Of course, there’s something absurd about a man who loves data railing against “statistical bullshit.” As Gruden demonstrates every Monday night, it’s not possible to assess football without statistics.
...Gruden energetically upholds the illusion, essential to fandom, that sports is a test of character—that every play, every game, really matters.
QBs who hold on to the ball too long tend to get sacked. That's fine, if not obvious. But isn't there a selection problem here? By definition, if we only look at plays in which the QB was sacked, we're looking at the handful of times they weren't able to throw the ball or scamble to the LOS. Aren't we really seeing which QBs can survive the longest before getting sacked? When a QB does throw the ball, there was a future potential sack that would have occurred at some point had they not thrown, which we'll never know and would never be included in the data.
The Sports Skeptic takes a look at one of Gregg Easterbrook's tenets. Does a reliance on undrafted players help teams win? Here's my take: The undrafted players we do see on Sunday's are the ones who are truly extraordinary. The teams that are least wedded to 'draft round' and status, and most wed to actual measured performance, will be the ones that are generally successful. And they're the ones who are likely to have a substantial number of undrafted players on their rosters. Unfortunately, the numbers don't back up the story.
Dispell your illusions.
Many football fans aren't aware of this, but the #1 college team in the land is playing the #3 team this afternoon--well, at least if you only count actual amateur scholar-athlete teams.
Just read the safeties article and commented there before realizing it was three years old. I don't think "3.6 points" makes sense by the typical logic employed here w/r/t EP...
To sum up that overlong comment, having an opponent pinned at their endzone already carries an EP (according to this site, one of +2.0 EP). So even incorporating future field position, the safety really only gives you two points plus the 1.6 points of good field position but minus the two points you were expecting to score anyways. So the safety itself is actually worth only 1.6 points of net EP. True, it's not two, but if you're going to incorporate the average benefit/cost of the future - as you do when you say that TDs are worth 6.3 and FG 2.3 - you should incorporate the average ben/cost of now. Otherwise if you had a pass take you from you own 20 to the opponent's one, and then a sneak gets you in the end zone, you end up saying the sneak was worth 6.3 EP, when really it was the pass that got you way more of that value.
I saw that Gruden article, too, and was absolutely fascinated, his rant against statistical analysis notwithstanding. And the man has a point insofar as such analysis has to be supplemented by film review. This is especially true in football where so much happens on the field that may or may not show up in the metrics we currently utilize.
Great links, thank you!
Nyet, according to historical data, 4th down from your own 1 is worth -1.6 EP. So a safety is worth at least 2 expected points, which is the same value as a converted 55 yard FG attempt. So only under the most extreme circumstances is a FG worth as much as a safety, no matter if you look at the EPA of the individual play, or the net value of the total drive.
On first down, a safety is worth at least 3.1 EP. That is quite significant.
Great point on the sack data. That's actually fairly standard survival analysis with censored data. You have to be very careful when looking at the data to consider what the effect of censoring is, which I don't think FO are.
Also I don't get quite what question he's trying to answer. There are some good bits of analysis that can come from timing pass plays (e.g. which QBs hold onto the ball the longest, which QBs suffer from a poor O-line) but you do need to look at every pass play to get the full picture.
Jonathan, I wasn't saying it wasn't significant. I said the math of just adding the safety value to the next kickoff value is disingenuous; you need to incorporate the expected value at the time of the sack. And btw, you typically don't get your safeties on 4th down; you get them on the 3rd down that would have preceded the 4th down, so the salient "current" EP measured before a safety is the 3rd down EP state. And as that includes the potential for a safety, it probably doesn't match the 4th down EP. So 4th down EP of -1.6 doesn't matter unless the safety is actually on that (likely punting) down.
Incidentally, where did you get that EP? I was just going off a chart on one of the EP posts here, and you're probably right, that may have just been 1st downs at each field position.
Still, I wasn't trying to give the exact value of a safety; I was just arguing that you shouldn't account for future EP without counting EP when describing the EP gained from a single play. I mentioned this on the original post comment, but it would be like crediting a guy who hits a 2 rbi single in baseball with 2 runs plus the new runs expected because now he's on base. He didn't create all of that added value; the fact that two guys were already on base means that he added value to a situation that already returns a particular amount on average. Similarly, the safety needs to account the EP that was there before it happens. After all, in getting the safety you are sacrificing other opportunity costs (blocking a punt and recovering for a TD, the even better field position you would get from a punt from the opponent's one, etc.) so the safety is not a pure 2 point gain.
I.e., the equation for added EP should always be points scored on the play minus EP prior to the play plus EP (pos. or neg.) following the play.
And of course the safety on 1st down is worth more. A safety that comes from a bad snap on 1st and ten at the 15 is probably worth WAY more. But if you're going to say that a safety averages out to 3.6 points, you would need to show that all of those prior EP's average out to zero. Unless you start counting absurd situations - hey, the other team has 1st and goal at *my* one and I get a safety, that's 2 - (-6.7) + 1.6 for a 10.3 point safety! - I doubt that's going to be true.