Super Duper
Today’s the Super Bowl, and it seems like everybody and his brother is picking the Giants to win. That makes no sense, by the way.
I mean, don’t get me wrong; in one game, luck is way way waywayway overpowering. The surest possible thing is maybe 60% likely to win any given game — just ask the Denver Broncos, who built an entire playoff run (and quite possibly an entire career for Tim Tebow) on luck. So I’m certainly not saying the Giants can’t win. I’m just saying that I can’t find any way of looking at it that doesn’t make the Patriots overwhelming favourites. Let’s go down the list, shall we?
1) Defense: This is the thing, innit? We all know the Patriots can’t win because their defense is shit. Everybody knows the Patriots have shit defense. The badness of their defense is the subject of many an epic example of skaldship. Only one question: is their defense really as bad as people make it out to be? I see here that the Patriots allowed 342 points this season. That’s not exactly awesome, but you know what? League average is 354.9. The Patriots have allowed fewer points than the average team, which is not the hallmark of a terrible defensive line.
One could argue that this number is skewed by the fact that the Patriots never had to defend against their own offense, which scored an outrageous 513 points. I’d suggest that one shouldn’t mention the overpowerance of the Patriots’ offense in an argument about how bad they are, but never mind; the point is valid. So what now? We get nerdy up in this shit, that’s what.
The Patriots scored 32.1 points per game. League average is 22.2. So let’s assume that, instead of facing an average team, the Patriots had had a Mortal Kombat Mirror Match, and played against themselves. We would expect them to allow an additional ten points that game, giving them a season total of 352. This total is still slightly better than average.
It’s not that I’m trying to be difficult, lord knows; I just honestly cannot locate a metric by which the Patriots play meaningfully worse than average defense. Pro Football Ref’s weirdo Defensive SRS stat has them at -0.1 for the year — just a tiny tiny bit below average. And how do they stack up against the Giants? Well, the Giants have allowed 400 points this year and have a DSRS of -1.5. That’s much, much worse defense than the Patriots.
2) Offense:
Okay, defense is a little bit skewed, but offense is downright unfair. Behold:
Patriots: 513 points scored
Giants: 394 points scored
Ouch. Now, the Giants are not a bad team, offensively — they scored an above-average number of points, and PFR has them as a respectable 3.1 OSRS. But the Patriots are obscene. They scored an outrageous number of points, and accumulated a 9.4 OSRS for their trouble. Offensively, there simply isn’t any comparison. But if you need the whole package assembled for you, here we are:
Patriots point differential: 171
Giants point differential: -6
The Patriots are awesome. The Giants… well, it looks like 9-7 involved a bit of good luck. They should be an 8-8 team. And, no, I’m not bothering to renormalise the offensive stats to account for not playing against own defense; trust me, it doesn’t close a 119-point gap.
3) Clurtch: Oh, but the Giants have Momentum and Destiny and Voodoo and all that other shit that doesn’t really exist on their side! I don’t care if they have Bigfoot, Xenu, and Charles Bronson on their side — turns out stuff that isn’t real doesn’t help very much.
But okay, okay. I’ll play along. We’ll pretend this stuff actually matters. How do we define "momentum" here, and why do the Giants have it? By definition, any team in the Super Bowl has won its last few games, so it can’t be that. Maybe it’s the team that’s won its last few games by the biggest margins. Is that it?
Conference championships:
Patriots 23, Ravens 20 (3 points)
Giants 20, 49ers 17 (3 points)
Oh. Well, can we go back a round, then?
Division round:
Patriots 45, Broncos 10 (35 points)
Giants 37, Packers 20 (17 points)
Oh. Uh. I guess maybe that’s not what momentum is about. Maybe… maybe momentum is about who had the longest winning streak at the end of the regular season. That would make sense. Let’s check:
Patriots: won last 8 games
Giants: won last 2 games
Hrm. Nope, no Giants-favourable momentum there either. Actually, it turns out the only way to present the bullshit "momentum" argument in the Giants’ favour is to say that, since the Giants won the last time these two teams played in the Super Bowl, they have the all-time Super Bowl momentum!
Do I need to tell you why that’s insane? I didn’t think so.
So what is it then? I really, honestly can’t find anything to suggest that the Patriots aren’t the overpowering favourite in this matchup. I’m beginning to suspect that people place more emphasis on what they feel in their intestines than they do on actual data!
Might want to reserve your place in the comments now so you can edit in an "I-told-you-so" in a good position when the Giants win.
So… the team that is a 3 point favorite to win is a favorite to win?
Comment by Stephen | 5 February 2012
Haha you lose! ELI MANNING SI WAY BETTAR THAN PEYTON 2 > 1 !!!!
Comment by Dave | 5 February 2012
It’s almost like there are people out there who aren’t just quoting the Vegas line, huh?
Comment by Darien | 5 February 2012
I’m just unsure who all these people you’re responding to were. The Pats were the clear favorites — not overwhelmingly but significantly — so an article confirming the wisdom of professional handicappers as though you were bucking the conventional wisdom just seems weird.
P.s. I was taking the points on the Giants (3 seemed like a perfect line) so I WAS TOTALLY RIGHT!
Comment by Stephen | 5 February 2012
As for a serious response, when the media was reporting that the Pats had the worst defense in the NFL, they were basing that on yards allowed and not points allowed, as somehow that’s the go-to stat in the NFL when judging defenses. Right before the game last night Chris Collinsworth sort of halfheartedly pointed out that the Pats gave up more yards than anybody else but were only middle-of-the-road in points allowed and kind of chalked it up to… I think luck? Or maybe that silly “bend don’t break” thing people are always talking about–which to me just means you’re intending to let the other team move the ball but somehow you’re thinking you’ll keep them from scoring at the last minute, which never really seems to work, and always seems to fail miserably at the worst possible moment.
I always think about it like a pitcher who gives up a ton of walks but manages to keep a low ERA. That shit is going to come back to haunt you, yo.
Comment by Dave | 6 February 2012
My article was neither about nor directed to professional handicappers. There are many people who believed the Giants to be huge favourites — if you don’t know any of them, okay, but I’ve been listening to it on a daily basis for weeks. Those people are the ones I was referring to.
Comment by Darien | 6 February 2012
Yards allowed seems like a very bad way to evaluate team defense; it doesn’t even make an attempt to neutralise for how many times the other team has possession, which will happen often against a team that scores often and scores rapidly, such as the 2011 Patriots. Indeed, a quick glance at this table and this one shows non-trivial correlation between scoring points and allowing yards. A correlation which disappears when we sort the defense table by points allowed rather than yards allowed. Could that be a coincidence? Of course. But it’s still a variable badly in need of being controlled.
The analogy to walks is instructive; a pitcher who surrenders a shitload of walks but also racks up a ton of strikeouts and doesn’t allow many homers is still going to be a pretty good pitcher. Witness 2010 Carlos Marmol, who allowed a preposterous 6 BB/9. That’s really, really bad. For perspective, no player on BBRef’s career leaderboard — which shows the top thousand players — even reaches 4 BB/9, much less 6. If we were to judge Marmol’s performance solely on walks issued, he’d look like pretty much the worst pitcher of all time. But there are mitigating factors: he struck out an equally-preposterous sixteen batters per nine, and allowed only one homer all year.
Of course, I’m not saying that issuing a bunch of walks is a good thing. It’s not. Neither is allowing a lot of yards. But they don’t tell the whole story.
Comment by Darien | 6 February 2012
So you made an overly broad statement (“everybody and his brother is picking the Giants”) based on anecdotal evidence? Did your gut tell you that was the case?
Ok: not being a dick anymore but stats that aren’t adjusted for strength of schedule are very suspect. Total points scored / allowed are really not terribly useful if you don’t adjust them for the strength of opponents.
Also stats that look at the entire season equally are suspect too. Especially where teams may have had important injuries or scheme changes during a season. The Giants in particular had real problems with injuries early in the season.
I actually like Football Outsiders’ totally insane weighted DVOA stat which probably (definitely) does too much but attempts to compensate for these things. Their pre-SB ratings suggest that the Pats are a large favorite (looking at performance from week 8 on) to a small favorite (looking from week 15 on): http://footballoutsiders.com/dvoa-ratings/2012/conference-championship-dvoa-ratings
I think it’s a pretty good argument as to why you wouldn’t assume the Giants are a huge dog.
Comment by Stephen | 6 February 2012
Oh and: beating a very crappy 8-8 Broncos at home — and even blowing them out of the water — is definitely less impressive than beating a very good 15-1 Packers team on the road.
I put as much stock in “momentum” as you do, but the Giants had a more impressive post season just based on who they beat and the fact that they did it on the road.
Comment by Stephen | 6 February 2012
You do seem to have difficulty remembering that sometimes, on my blog, there are jokes.
I concur that more advanced statistics are more useful, but I was trying not to get in well over my head (since football is only a passing interest of mine and not something I understand as well as I do baseball) and trying to keep the tenor of the article relatively light. You’ll notice I did bring SRS into the equation, which has the Patriots as roughly average defensively. However. Football statistical analysis suffers from a crippling sample size problem. It’s really pretty unavoidable; the season’s only sixteen games long. Of course, it’s still better than anything else we have — much better than my stated pick of the Broncos because I believe that God believes in Tebow — but I am highly suspicious of any statistical analysis that purports to be from week 15 forward; that’s only (at most) five games. That’s not a very good data set.
Which isn’t to say that I don’t think beating the Packers on the road is more impressive than beating the Broncos at home. I do; I expected the Pack to win that game by a pretty big margin, and (Tebow jokes aside) I’d have been very surprised had the Patriots lost. But if we’re allowing that "momentum" is a thing, and that said thing consists of something other than a winning streak and/or a streak of winning by a lot, then I’m totally lost as to what it could possibly be. If having "momentum" requires beating better teams on the road, then it’s the kind of thing that, by definition, top seeds (like the Patriots) can’t have.
Comment by Darien | 6 February 2012
It’s easier to find jokes when they’re funny. ZING!
I totally get what you’re saying with sample, but weeks 15-19 constitute a full 1/4 of the total games played. And mojo about momentum aside, football is a game where teams can realistically perform a lot different in the last five games than the first 15.
Sure, some of that is luck but some of it is also because injuries and coaching really matter a ton. Teams like the Giants that start with major injuries like the Giants can take a while to adjust their schemes amd account for that.
It’s not unreasonable to look at a team’s last few games and weight their performance there a lot more heavily than early games. Yeah you’re at the mercy of small samples, but what else do you have? Looking at largely irrelevant data because there’s more of it isn’t helpful.
Comment by Stephen | 6 February 2012
I think it’s fair to point out that coaching matters a TON more in football than it does in baseball. I’ve literally seen players in interviews say something to the effect of “We finally started doing what coach told us to do, and we started winning!” The key phrase is generally “bought into” the scheme or the ideology or the philosophy that the coach is preaching, but what they really mean is they finally started doing what the fuck they were told to do and found out that shit actually worked. So that’s just another way teams can “miraculously” get better during the course of a short football season.
Comment by Dave | 6 February 2012
trololol
The only thing I’m not getting here is what it is we disagree about. Seems to me we’re on the same page pretty much everywhere.
Comment by Darien | 7 February 2012
Aside from my initial sniping that I have no clue who this article was for, I disagreed with this: “So what is it then? I really, honestly can’t find anything to suggest that the Patriots aren’t the overpowering favourite in this matchup. I’m beginning to suspect that people place more emphasis on what they feel in their intestines than they do on actual data!”
I think that the evidence — not my gut, I didn’t really care who wins or loses since I hate both teams — makes the Pats a slight favorite, but that you can make a reasonable argument based on the data that the Giants should’ve been favored.
Comment by Stephen | 8 February 2012
Oh and as an aside, in response to the earlier discussion about PPG allowed vs YPG: the reason you need to really worry about yards and not just points if you’re trying to measure defense is that good offenses (like the one the 2011 Pats had) can hide a lot of defensive flaws.
Consider if you had a perfect offense, one that scored a TD every possession. This means the opposing team gets the ball on their own 20 each time and needs to drive 40+ to have a reasonable shot at scoring a FG. A defense that allowed 40 yards per drive would not be what I’d consider good, but in this hypothetical scenario that defense could allow 40-yards per drive and still give up very few points.
The opposite is true, too: a great defense that’s saddled with a bad offense can end up giving up a lot of points but relatively few yards and it’s not their fault if the offense can’t drive, allowing the opposing team to always start with good field position.
And of course there are a bunch of other ways that offenses help/hurt their defenses: forcing them to play more per game (increasing injury rates and exhaustion), giving the opponent flat-out more chances, etc.
It’s why I don’t think you can do a really good analysis of offense/defense considering them both in isolation. If a team has a great offense, you would expect their defense to look better on any per-game stats (and slightly better on per-drive stats).
Go check out Football Outsiders’ stats by drive which I think is really telling: http://footballoutsiders.com/stats/drivestats
The Pats were 32nd (aka last) in yards allowed per drive but 21st in points allowed per drive. Why is that? Well the defense on average started its drives at the 24.06 yard line — 2nd best in the NFL. The Giants ranked 22 in yards allowed per drive, but allowed more points per drive because they had significantly worse field position (28.56 average). And the Giants had a good offense! It’s just that the Patriots’ offense was so ridiculously good it masked a ton of problems with their relatively weak defense, especially when you look at “Stuff allowed per game” and especially when you look at “Points allowed per game.”
Comment by Stephen | 8 February 2012
As long as I’m thinking about this post: I think another of my issues is with the word “overpowering” as applied to being a favorite. Darien, you had this to say: “The surest possible thing is maybe 60% likely to win any given game.”
Uhm. No. That is really, really low worst-case scenario in the NFL. The best team versus the worst team in the league is probably 4:1 to win (80% favorite). I say this because it just so happens I love to bet underdogs on the moneyline (which means you bet the team to win straight up, not taking points) and routinely get better than 2:1 odds. It’s not uncommon for a few matchups in the year to pay 4:1 on one of these bets. I don’t think the casinos are in the habit of routinely offering gamblers several times the true odds, so I’m thinking that a 60% chance to win at best is really low.
Here’s a chart that provides a rough calculation of point spreads to moneylines in the NFL: http://www.predictem.com/nfl/odds-conversion.php
Basically the negative numbers represent the favorites, so if you bet that way you need to bet that much money to win $100. E.g. a 3-point favorite would be -155, so you would need a bet of $155 would pay out $100 (you’re laying odds taking the favorite). The underdog in that scenario is +135, meaning a $100 bet pays out $135 (or 1.35:1 odds). The fact that the odds don’t add up when you add both sides is the house’s vig.
But if you go all the way down, you’ll see a 10-point spread is a whopping +325 for a bet on the underdog, meaning they’re paying 3.25:1 on that bet — which, if you think the dog has a 40% chance of winning, means they’re paying you more than twice the true odds (a team with a 40% chance to win would be worth 3:2 or 1.5:1 in a fair bet).
And that chart only goes up to 10-point games. There are legitimately a few 15-point games each year! If you really believe that a team’s odds are never worse than 2 in 5 then you should move to Vegas next football season and cleanup. Or hell, I’ll take that action.
I think that’s maybe the misunderstanding to me. A 3-point underdog like the Giants should have roughly a 40% chance to win if the line is good (and it usually is), but in my mind a 3-point dog is far from a huge underdog. In fact, here’s a list of historic Superbowl lines: http://www.vegasinsider.com/nfl/superbowl/history/ (they have the line closing at NYG +2.5 and not the +3 I thought).
That +2.5 is the *smallest* line since 1982. So, no, I don’t think that the Pats should have been overpowering favorites. I think they were probably evaluated pretty fairly — both by Vegas and by the media, which (using purely non-empirical method of my recollecting analyst predictions) seemed to favor the Pats slightly but not hugely.
Comment by Stephen | 8 February 2012