I don’t like moral victories. I’m sure many of you don’t either, but this offseason we're bound to talk about them—all without saying the phrase "moral victory." We'll talk about how good our losses were as if that's a reason to be proud and many of us (probably me included) won't bat an eye. Does that make sense? What do moral victories mean?
Moral victories are for the Gophers dammit https://t.co/o7P2Fpdoy6
— Arif Hasan (@ArifHasanNFL) December 11, 2015
Well, they suck.
I often relate things to my debate career because why not, it's something I know. But this concept is particularly close to what happens in competitive debate circles. You can win or lose debates, that's true, but you also get assigned "speaker points" in the round, which allows you to more properly seed brackets and mark out gradations between debaters.
Speakers will usually get points between 25 and 30, with half-points to allow for greater differentiation. The reason this is relevant is because in debate, you and your debate partner can earn more speaker points than the other team but still lose, often referred to as a high-point loss (or, if you're on the other side of it, a low-point win)—the clearest and most absolute indication of a "moral victory".
It sucks being told in no uncertain terms that you were good enough for a win and lost because of a technicality or an errant piece of data in the evidence.
At the most prestigious national tournament of my senior year of high school, I pulled off a pretty big high-point loss against a high-level team (names besides mine blanked out because they were kids too).
You can imagine my ecstasy at being told I was good enough to win but not capable enough to pull it off.
I bet it sucks a lot more with millions of eyes watching and millions of dollars invested in you, personally, in the outcome.
Moral victories are draining, possibly quite a bit more than out-and-out victories, because knowing that a small thing was the difference between you and your goal is a much bitterer pill to swallow than knowing you simply weren't prepared or ready.
In football, to me anyway, a "moral victory" is only awarded in a loss—if you win 40-0 you probably are better than people expected, but no one needs to hedge anything—and only if you play better than you were "supposed" to. Off the top of my head, the Vikings have had three this year: against the Denver Broncos, against the Arizona Cardinals and against the Seattle Seahawks.
We might be able to find others if we take a look at times where the Vikings covered the spread in a loss. It turns out, those three games are the only games where the Vikings met that condition this year. The first two were encouraging and the third ripped our hearts out—even though, if we're being honest, they all ended the same way: a botched final play to close out the game (sack-fumble, sack-fumble, Gary Anderson).
Maybe with time, we'll see that aching third game characterized in a similar way to those other two games.
Can those be indicators of something to work with? I'll look at that, but...
Well, what about moral losses?
Moral Record
The first thing you probably have to account for is probably "moral losses"—which don't have the opposite definition of a moral victory but come pretty close: when you're expected to win and you lose. I think I would expand that even further and say when a team is expected to win a clear majority of the time, which I'll define as three points or greater on the spread.
For games within three points (pick'ems), I'll say a "moral loss" is when an even game is lost by 15 points or more.
The Vikings, by that definition, only had three moral losses: Seattle, Green Bay and San Francisco.
So I guess, including the playoffs, the Vikings are a moral 3-3. If we exclude the blowouts in pick'em games, then they are 3-0.
Maybe this means the Vikings are on the rise, but are there other teams that suffered through the indignities of moral victories, only to improve a season later? We can take a look at that as well as whether or not the different kinds of "moral losses" matter to that projection, too.
There were about 25 teams with at least three moral victories between 2011 and 2014. They are below:
Tm |
Year |
Record |
Next Year Record |
Moral Victories |
2011 |
0.250 |
0.313 |
5 |
|
2012 |
0.125 |
0.25 |
5 |
|
2014 |
0.188 |
0.438 |
5 |
|
2014 |
0.125 |
0.375 |
5 |
|
2013 |
0.250 |
0.375 |
4 |
|
2014 |
0.438 |
0.188 |
4 |
|
2011 |
0.125 |
0.647 |
4 |
|
2013 |
0.250 |
0.188 |
4 |
|
2013 |
0.500 |
0.722 |
4 |
|
2013 |
0.125 |
0.563 |
4 |
|
2014 |
0.438 |
0.647 |
4 |
|
2014 |
0.250 |
0.625 |
4 |
|
2012 |
0.313 |
0.625 |
3 |
|
2011 |
0.375 |
0.438 |
3 |
|
2012 |
0.313 |
0.25 |
3 |
|
2014 |
0.188 |
0.313 |
3 |
|
2012 |
0.125 |
0.647 |
3 |
|
2011 |
0.188 |
0.588 |
3 |
|
2013 |
0.344 |
0.438 |
3 |
|
2011 |
0.438 |
0.667 |
3 |
|
2012 |
0.469 |
0.438 |
3 |
|
2011 |
0.375 |
0.438 |
3 |
|
2012 |
0.438 |
0.25 |
3 |
|
2012 |
0.250 |
0.438 |
3 |
|
2011 |
0.313 |
0.588 |
3 |
As you can tell, teams with moral victories suck.
In part, that's because you must lose games in order to accumulate moral victories and because you must be expected to lose—a team that is expected to lose a lot of games is obviously a team that is not good.
You can do the math above, but let's demonstrate how teams did when they accumulated a lot of these near-misses:
Moral Victories |
Number of improved records |
% of improved records |
Avg Wins+ |
5 |
4 |
100% |
2.8 |
4 |
6 |
75% |
3.2 |
3 |
10 |
77% |
2.5 |
3-5 |
20 |
80% |
2.7 |
The problem? None of those teams had winning records, and the Vikings didn't just have a winning record, they had a great record. That's a huge problem (for this analysis) because any teams that improved could just as easily have improved as a result of regression to the mean, a great set of draft picks (more likely with a bad record) and/or a coaching change.
We're pretty familiar with this, because of the five available years in the dataset, Minnesota comprises three of those years. Perhaps more interesting might be when we add the two types of moral losses—after all, those can only happen if a team is expected to do well and might account for the regression effect.
The first accounting of moral losses (the accounts that include blowouts in pick'ems) can be found here, but I'll only list the average below (there are 59 teams that are .500+ in their "moral record" by this accounting).
Moral Record |
Number of improved records |
% of improved records |
Avg Wins+ |
1.000 |
7 |
50% |
1.4 |
.750-.999 |
9 |
82% |
3.1 |
.667-.750 |
15 |
71% |
1.0 |
.500-.666 |
11 |
52% |
0.8 |
OK, useless.
What about the other accounting that takes away blowouts in pickems? Because I'm only interested in the Vikings, I'll restrict that to teams that were .667 or better.
Tm |
Year |
Record |
Next Year Record |
Moral Victories |
Moral Losses |
Moral Record |
Improved? |
2014 |
0.188 |
0.438 |
5 |
0 |
1.000 |
1 |
|
2011 |
0.125 |
0.647 |
4 |
0 |
1.000 |
1 |
|
2014 |
0.438 |
0.647 |
4 |
0 |
1.000 |
1 |
|
2014 |
0.188 |
0.313 |
3 |
0 |
1.000 |
1 |
|
2011 |
0.313 |
0.125 |
2 |
0 |
1.000 |
0 |
|
2013 |
0.250 |
0.188 |
1 |
0 |
1.000 |
0 |
|
2013 |
0.438 |
0.375 |
1 |
0 |
1.000 |
0 |
|
2014 |
0.647 |
0.813 |
1 |
0 |
1.000 |
1 |
|
2014 |
0.563 |
0.529 |
1 |
0 |
1.000 |
0 |
|
2012 |
0.125 |
0.25 |
5 |
1 |
0.833 |
1 |
|
2014 |
0.125 |
0.375 |
5 |
1 |
0.833 |
1 |
|
2013 |
0.250 |
0.188 |
4 |
1 |
0.800 |
0 |
|
2012 |
0.313 |
0.625 |
3 |
1 |
0.750 |
1 |
|
2012 |
0.313 |
0.25 |
3 |
1 |
0.750 |
0 |
|
2012 |
0.469 |
0.438 |
3 |
1 |
0.750 |
0 |
|
2011 |
0.375 |
0.438 |
3 |
1 |
0.750 |
1 |
|
2011 |
0.313 |
0.588 |
3 |
1 |
0.750 |
1 |
|
2012 |
0.250 |
0.25 |
2 |
1 |
0.667 |
0 |
|
2013 |
0.500 |
0.722 |
4 |
2 |
0.667 |
1 |
|
2014 |
0.250 |
0.625 |
4 |
2 |
0.667 |
1 |
|
2013 |
0.250 |
0.375 |
4 |
2 |
0.667 |
1 |
|
2013 |
0.625 |
0.647 |
2 |
1 |
0.667 |
1 |
|
2011 |
0.500 |
0.313 |
2 |
1 |
0.667 |
0 |
|
2012 |
0.438 |
0.5 |
2 |
1 |
0.667 |
1 |
|
2013 |
0.250 |
0.125 |
2 |
1 |
0.667 |
0 |
|
2014 |
0.125 |
0.188 |
2 |
1 |
0.667 |
1 |
|
2013 |
0.438 |
0.125 |
2 |
1 |
0.667 |
0 |
|
2013 |
0.188 |
0.25 |
2 |
1 |
0.667 |
1 |
OK, still a bunch of crap teams plus the 2013 and 2014 Cardinals. They seemed to do alright. Anyway, the aggregate:
Moral Record |
Number of improved records |
% of improved records |
Avg Wins+ |
1.000 |
5 |
56% |
1.6 |
.750-.999 |
5 |
63% |
1.7 |
.667-.750 |
10 |
63% |
0.9 |
Mild improvement!
Again, it's with a bunch of teams people should expect to improve anyway because bad teams get better more often than not, but at least it's something. But the good teams (basically just Arizona twice) did improve—though one would argue maybe it's having their quarterback healthy.
Honestly, I know it's all bullshit—the idea of moral victories, and especially the idea of comparing moral victories and moral losses—but we're going to be talking about it for some time. This was a team that was three bounces of the ball away from 13-3 and a playoff win.
As fans, it's pretty convenient to forget when they got blown up by teams they shouldn't have been blown up by or had close games against teams they should have crushed (23-20 to the Bears? Overtime against the Rams?) so it's at least interesting to see if these moral victories mean anything.
In our hearts, we know they don't, but in the interest of at least relatively objective fandom I like to confirm it.
Regression
In that case, what does count for figuring out if a team "played better (or worse) than their record" in a meaningful way—which is to say in a way that translates into next season's performance?
Let's look at what Bill Barnwell used to look at every year for Grantland, which are the indicators that a team may improve or not improve. To the extent that I can, I'll try to go from least granular (i.e. smaller sample) to most granular.
As a reminder, we should sometimes think like a player, not like a GM—sometimes we think regression to the mean happens simply because that is the way of things instead of understanding that a player (or team... or opposing teams) has figured out a process to force correction.
In baseball, if a player's batting average dips, it may be a result of a change in his form or something pitchers have discovered. Traditional baseball statisticians will sometimes dismiss a temporary drop in batting average (or OBP or what have you) because they assume things regress. Because they usually do. But maybe they regress back to the mean because players and coaches are constantly trying to correct changes.
In the same way, regression to the mean may be enforced, not natural, and may be why the Tony Dungy/Peyton Manning Colts almost never regressed to the mean. Having a good coach (like we feel Zimmer is) could also prevent regression to the mean (if it's negative) or enforce it (if it's positive).
With that out of the way, were the Vikings actually "better" or "worse" than their record as far as things they can build on/be wary of for next year?
Close Game Record
Your record in close games typically does not stay the same year after year. Better teams get better either by not getting into close games or by getting into close games with good teams instead of average or bad teams. Sure, closing out games is a skill, but it's one teams typically learn quickly under the same coaching staff or a new one. Regardless, close game records are usually the best way to find teams that will improve or get worse.
Close games can be defined a lot of ways, but Barnwell chose to use "one score" games, so that's what we'll do here. Teams most helped and hurt by their close game record:
Tm |
+/- |
6 |
|
6 |
|
4 |
|
3 |
|
3 |
|
-3 |
|
-4 |
|
-4 |
|
-4 |
|
-4 |
|
-5 |
|
-6 |
So the Seahawks, Ravens, Cowboys, Titans, Browns, Giants and Chargers are all probably better teams than their record indicates because of unusually bad luck in close games, while the Panthers, Broncos, Cardinals and Patriots are a little bit worse than their record indicates for the opposite reason.
The Vikings do not appear to be that much better or worse as a result of close games (they are +2 in close games). Like turnover ratio below, winning close games may be a skill, but the extremes (like San Diego and Carolina) deserve examination.
Turnover Ratio
Turnovers are an indication of skill, and good defenses produce turnovers consistently, while bad offenses do the same. But they are so low frequency (1.94 percent of plays this last season, or just under one every 50 plays) that small chance events produce big changes, especially with their impact on the outcome of games.
That means that teams at either extreme usually cannot sustain it, despite it being a generally good indicator of quality. Teams with double-digit TO Ratio (which is expressed as a difference, not as a ratio, but names stick):
Tm |
TO "Ratio" |
16 |
|
16 |
|
12 |
|
-14 |
|
-15 |
|
-18 |
So the Vikings (who ranked eighth overall in turnover ratio) do not seem to be at risk of negative regression as a result of what is probably turnover luck.
One thing a lot of people did mention early in the season, however, is the amount of fumbles the Vikings recovered. Forcing fumbles is a repeatable skill, but recovery typically isn't (in statistical terms—I have no doubt that if a team stopped practicing fumble recovery, they would lose fumbles at a sustainably high rate). Did the Vikings mask their turnover ratio with good fumble luck?
Let's look at teams most hurt and helped by fumble luck:
Team |
2015 |
62.2% |
|
61.3% |
|
60.5% |
|
40.5% |
|
37.2% |
|
34.5% |
Halfway through the year, the Vikings had the league-best recovery rate (72.2 percent). By the end of the year, it largely had evened out and the Vikings ranked 10th at 53.3 percent. That's about one fumble more than you'd expect. So in this regard, at least, Minnesota was not worse (or better) than their record. Oakland, San Francisco and Houston have a turnover ratio that is perhaps too complimentary of their overall abilities, while the opposite is true of Detroit, New England and Dallas.
"Offensive" Scores Allowed
Safeties and return touchdowns—from the defense and special teams alike—do not tend to repeat year-to-year (for example, the Rams "allowed" 52 points last year from their offense and special teams and ranked first in points given up that way. They ranked 16th in non-defensive scores allowed this year).
Sometimes a defense will be artificially weighed down or propped up by the offensive scores given up on the other side of the ball through no fault of their own. We know the Vikings defense was good, in part because they were 5th in the NFL in points allowed per game. Were they artificially propped up (or, conversely, weighed down)?
Tm |
NonDEFPts |
JAX |
42 |
SDG |
32 |
DAL |
30 |
TEN |
30 |
BAL |
30 |
CAR |
8 |
IND |
6 |
ARI |
6 |
MIN |
2 |
PIT |
0 |
So here, there's finally evidence one way or the other that the team was artificially helped in some manner by the unusually low number of offensive scores allowed. This could be in part because of the low turnover rate of the offense, but it turns out that actually has a surprisingly low correlation—the difference is not as much in the number of turnovers as it is an errant score-saving tackle or a well-placed fumble (the Vikings tended to fumble in the opponents' red zone more than their own, for example).
But, there's a better way to measure this: points per drive and points per drive vs. expectation. PPD is pretty easy to understand. Excluding kneeldowns, we can figure out how many points a team gives up per possession. This will control for pace (the Eagles played much faster than the Panthers, so if their defenses were equally good the fact that the Eagles defense went up against more possessions than the Panthers defense means they would give up more points) while also eliminating other kinds of scoring.
PPDvE controls for field position, because a defense that is up against the opponents' one-yard-line will fare better than a defense starting from their own twenty.
In the regular metric, the Vikings defense ranks a middling 13th. However, controlling for field position, they rank 16th.
That feels wrong, and it probably should—the defense overall looks and feels like it is better than league average by quite a bit—but it may serve as a soft indication that the Vikings overperformed. A good way to double-check that is to look at things that are quite a bit more predictive like strength-of-opponent adjusted points allowed (they rank 5th) and defensive DVOA (14th).
In all likelihood, the Vikings defense overperformed last year but maybe not by as much as the numbers indicate.
For what it's worth, the Vikings offense, ranked 16th in points scored, likely underperformed by a little—ranking 11th in points per drive though rank 24th in field-position adjusted points per drive. That may not offset the difference but it does cushion it a bit.
Point Differential
Perhaps one of the most predictive measures of quality, point differential, one can very reasonably predict next year's record for a team using what is called pythagorean expectations, which basically just looks at points scored, squares it, and then divides it by the sum of the square of points scored and points allowed.
It sounds more complicated than it is; a team that scores 200 points and allows 100 will have the following process applied to them to figure out their Pythagorean wins: 200 squared is 40,000. 100 squared is 10,000. 40,000 divided by the sum of 10,000 and 40,000 is 0.800. In a 16-game season that's 12.8 wins.
it is extremely good at determining "team quality" as it pertains to future wins (both during the season and forecasting the next season), and is even better when testing for the correct exponent (which instead of 2 is 2.37 for the NFL).
By this measure, the Vikings have 9.8 "expected" wins, a bit (1.2) below their actual average—suggesting, along with scoring defense, that the Vikings overperformed by about one win.
We do, of course, have the ability to adjust for the strength of opposition, and by point differential it just so happens that the Vikings had the 13th-most difficult schedule in the league. After accounting for opponents, the Vikings' "true" Pythagorean wins is 10.5, only half a win below their actual win total.
This means they may have gotten lucky a little bit in their actual games, but because they played good teams, that more than offsets the noise provided by luck.
Ranking 9th in the NFL here jives with their 11th overall ranking in DVOA and is not too far from their 7th overall win percentage.
Our Moral Verdict
So do the moral victories mean anything? No, because they are losses.
Also, because the Vikings screwed up some games they were supposed to be in. Had the Vikings tied those games that were predicted to be draws (or gone 1-1-1), they wouldn't just be a better team record-wise (12-3-1), it would be an indication that they don't overperform sometimes and underperform at other times.
This bore out when we looked at other indicators of possible negative (or positive) regression.
The Vikings really are as good as an 11-5 team at the moment, despite the fact that no one really gave them credit for it until the very end.
That's great; 11-5 teams win the Super Bowl sometimes and have made the playoffs over 97 percent of the time (of the 67 teams in the 12-team playoff format to go 11-5, nine of them have won it all). Minnesota is a team that performed better against expectations, on average, than any other team in the NFL (going 14-3 against the spread including playoffs, and covering 82.3% of the time).
So when we get hyped this offseason (and we will) talking about the team—let's not talk about the almost-had-beens and instead talk about the will-bes. Don't talk about the time the Vikings nearly beat three teams, because we'll forget about the times they got crushed by teams they should have hung with.
*****
At that debate tournament, hosted by the Blake Schools at the Hyatt Regency with over 100 teams in attendance, my debate partner and I made it to the round of 32 in large part because we seeded higher than teams with the same preliminary record, courtesy of our speaker points. Our high-point loss turned into a tangible outcome (and trophy).
But that's only because we won the other rounds.
Our high-point loss was an indicator of quality, but it was our record that got us in. The sour taste of an almost-win gave us clues as to how good we were, but it was the winning that happened later on that mattered.
For the Vikings, it's much the same. Winning got them there and there's definitely quality, but there are just as many clues to point to an improvement as much as a decline. Instead of harping on past near-victories, the team and fans should be looking to future statement wins.
So let's think like a player and not a GM—instead of trying to prove that last year's team was better than people thought, let's talk about why this year's team can be better than the rest of the NFL. What process will make it better? Is it new, key players? Is it development? Is it health? Is it coaching?
Did we fix our swing?