Luckiest and Unluckiest Teams - Week 14

For some reason this is my favorite stat.

I estimate team "luck" by using my efficiency regression model to calculate each team's expected wins--how many wins a team can normally expect, on average, given their actual performance in offensive and defensive running, passing, turnovers, and penalties. The difference between the expected wins and actual wins is what I loosely call team luck.


What do I mean by luck?

In my own life, I'm a big believer in hard work, preparation, focus, execution and everything else that isn't luck. Coaches and players can't let themselves think any other way for a single second. But once we account for all those things, what do we call what's left over? Statisticians call it "residual," and a substantial portion of any residual is due to random effect, including sample error and what I call "bunching." In a bounded and meticulously measured system like sports, a vast amount of the residual from any decent model will be due to randomness. A season of 16 games simply isn't long enough for the breaks to even out.

There are plenty of things my model does not consider, special teams being the most prominent. But special teams plays are the most random events in the sport, save for the coin flip. Luck is a punt that lands on the 5 and skids into the end zone for touchback instead of bouncing into the air and being downed at the 1. A kick or punt return for a touchdown certainly requires skill, but when the kick return (or missed field goal or anything else) occurs means everything.

A missed FG when a team is already ahead by 20 points doesn't mean much, but when a team is behind by 1 in the 4th quarter, it means the game. Teams and players can't control when those events occur, or else they'd save up their successes for when they matter most and their failures for when they don't. So in a very substantial way, they are luck, at least when it comes to deciding game outcomes. (For a more thorough discussion, see this post.)

Schedule strength is part of luck too. Teams fortunate enough to have a soft schedule (so far) are likely to have more wins than a team unlucky enough to face a tougher schedule. Even though schedule strength is something real that can be measured, it still lies outside of a team's control.

2009's Lucky and Unlucky Teams So Far

The luckiest teams so far this year include the Vikings, Colts, Bengals, and Saints. Usually when teams have extremely good records, they are both good and lucky. The reason is simple: It's extremely rare for a team to have a good record and at the same time be unlucky.

One tangible example of what I'm talking about was on display at the end of regulation in the Saints-Redskins game. Kickers miss short field goals periodically, but they usually go unnoticed unless it costs the game. (Now former) Redskins kicker Sean Suisham missed a chip-shot FG that would have put his team up by 10 points and iced the victory. Instead, Suisham missed and New Orleans was able to drive 80 yards for a TD to send the game to OT. The Saints were both lucky and good.

On the other side of the coin are the Redskins (unsurprisingly), Steelers, Buccaneers and Rams, who have received the short end of the stick more than their fair share this year.

The Charmed One (Part 2)

Two years ago the Brett Favre-led Packers topped this list as the luckiest team. At the time, my Packer-fan brother-in-law astutely observed that my model was not measuring luck. Rather, it was measuring "Favre-ness." Then last year, the Favre-led Jets topped this list going into the final games of the season, while the Favre-less Packers became the unluckiest team.

Guess which team tops the list this season. The Favre-led Minnesota Vikings are the luckiest so far in 2009, with 2.7 more wins than we'd expect given their general on-field performance. This is not a knock against Favre. In fact, my Win Probability Added (WPA) analysis put him at the top of the MVP list as the player who has contributed most to his team's success. Maybe my brother-in-law is right.

Then again, maybe it has more to do with missed last-second field goals, a weak schedule, and some good old-fashioned luck. In terms of efficiency (yards per play), the Vikings don't look that great on paper. On offense they are 9th in pass efficiency, 18th in run efficiency, 14th in fumble rate, and 1st in interception rate. On defense they are 13th against the pass, 10th against the run, and 26th in interception rate. Overall, they are 13th in the league in penalty yards per play. Yet the gods of football are smiling on the Metrodome, blessing the Vikes with the third best record in the NFL at 10-2.

Here is the full list. Click on the headers to sort.


RANK
TEAM
GWP
Expected W
Actual W
Luck
1
MIN
0.61
7.3
10
+2.7
2
IND
0.78
9.4
12
+2.6
3
CIN
0.54
6.5
9
+2.5
4
NO
0.82
9.8
12
+2.2
5
ATL
0.35
4.2
6
+1.8
6
ARI
0.52
6.2
8
+1.8
7
MIA
0.36
4.3
6
+1.7
8
OAK
0.21
2.5
4
+1.5
9
JAC
0.50
6.0
7
+1.0
10
SEA
0.35
4.2
5
+0.8
11
CAR
0.36
4.3
5
+0.7
12
CHI
0.37
4.4
5
+0.6
13
KC
0.21
2.5
3
+0.5
14
DET
0.17
2.0
2
 0.0
15
GB
0.68
8.2
8
-0.2
15
SF
0.43
5.2
5
-0.2
17
SD
0.77
9.2
9
-0.2
18
DAL
0.69
8.3
8
-0.3
19
DEN
0.71
8.5
8
-0.5
20
CLE
0.13
1.6
1
-0.6
21
BUF
0.39
4.7
4
-0.7
22
PHI
0.73
8.8
8
-0.8
23
NYJ
0.57
6.8
6
-0.8
24
BAL
0.58
7.0
6
-1.0
25
TEN
0.51
6.1
5
-1.1
26
NYG
0.70
8.4
7
-1.4
27
HOU
0.54
6.5
5
-1.5
28
NE
0.71
8.5
7
-1.5
29
STL
0.24
2.9
1
-1.9
30
TB
0.27
3.2
1
-2.2
31
PIT
0.69
8.3
6
-2.3
32
WAS
0.53
6.4
3
-3.4

  • Spread The Love
  • Digg This Post
  • Tweet This Post
  • Stumble This Post
  • Submit This Post To Delicious
  • Submit This Post To Reddit
  • Submit This Post To Mixx

23 Responses to “Luckiest and Unluckiest Teams - Week 14”

  1. The other Brian says:

    "Maybe my brother-in-law is right."

    I'm glad you have developed a statistical model to prove this.

  2. Another David says:

    dear sweet lord, the redskins are awful. but at least now we have an excuse - it's just bad luck!

  3. Jason says:

    Steelers fan and I were talking about his team yesterday. Of their 6 losses, two have come in the last few seconds of regulation and three in OT. That team probably "should" have 8-9 wins.

  4. Will says:

    Since the expected wins regression is based on average offensive and defensive performance, whereas actual wins are largely a matter of timing as you point out, would it be more accurate to call the luckiest teams "most clutch?" Or at least, luckiest and most clutch? In the Saints/Skins example, the missed FG was pure luck, but the TD pass to tie it was the perfect outcome at the perfect time.

    As the Favre case illustrates, WPA accounts for timing as well as overall achievement. Taking a knee on the 1 yard line can give a player a very high WPA relative to scoring the TD, but also result in a large residual against the model, which is based largely on scoring (correct me if I'm wrong here). Maybe WPA can help tease apart luck from timing?

  5. Alex says:

    Your brother in law has an interesting point. Is there an easy-ish way to see if anyone has been on a lot of lucky (or unlucky) teams over his career? Maybe that would point to something missing in the model.

  6. Chase says:

    Cool post, Brian. Agree with these findings. Worth remembering that while the Saints and Colts are two of the luckiest four teams, they're still the top two teams in expected wins. So they should be the best two teams, and they are, but they just should have more/some losses.

  7. Anonymous says:

    What would be the difference if luck were measured by percent change in wins vs wins expected.

  8. Zach says:

    Is the GWP for each team adjusted to their SOS? A team like the Vikings may not be as lucky if their .61 GWP would be ~.70 based on their easy schedule.

  9. Ian says:

    Detroit rank 14th. I don't think they've been so high up a list ever. Still, at 2-10 I think as a fan you'd rather be able to say "we're unlucky". As it is, Detroit fans just have to face reality - they are plain baaaaad :)

  10. Jeff Clarke says:

    By several measures, the Washington Nationals were an extremely unlucky team this past year. I think DC is screwed. I'm glad I live in Seattle now...

  11. Nate says:

    I wonder if there's any correlation between a team's luck during the first half of the season and their luck during the second half, or between one year and the next.

    If there's no correlation, it would suggest that the stat really does mean something like luck. If there's a strong correlation, it would suggest that it's merely picking up something that's missing from the model.

  12. Marver says:

    "Teams and players can't control when those events occur, or else they'd save up their successes for when they matter most and their failures for when they don't."

    Precisely. Which is why calling the leader in WPA the MVP is a farce. Just because his timing happened to be best for that team doesn't mean he would have been as valuable for any other team in the league.

    Take for example a kicker that makes his first twenty kicks of the season and misses his next 5. It's likely his WPA grades lower than a kicker who missed five meaningless field goals, instead making the 20 most important. That's a function of luck; nothing more, nothing less. I'm sure there are actually some teams who, in that season, would have won more games if their team hit field goals 1-20 than a random spread of 20 over 25. It doesn't mean the first kicker is any less valuable than the other...the only information you know about those kickers that has any meaning is they're both 80% kickers in that season. That makes them equally valuable.

    Giving Favre the MVP award based on WPA and then turning around and basing a 'luck' article on team efficiency makes no sense. Either say Favre's the Most Lucky Player, or give the award to the player most influencing the most correlated statistic to consistent winning, ie. the quarterback of the team with the highest passing efficiency of the league.

  13. James says:

    I'm curious about the answer to Zach's question too. Does expected wins factor in a team's GWP AND their opponent's GWP?

  14. Dan Schlauch says:

    I was curious about that too, Nate. I checked the pearson correlation of of GWP and luck and found none (-0.001).
    This at least suggests that any real elements missing from the model (special teams, clutchness, etc.) are not correlated with the quality of the teams, as I would have expected if something important was missing from the model.

  15. Anonymous says:

    If the residual were truly random "luck" then the distribution of the error terms should be random. There shouldn't see any pattern in which teams are "luckiest" and which are "unluckiest". Given the rather clear pattern that we see among which teams are luckiest and which are unluckiest, I think you are either missing some relevant factors in your regression equation and/or have misfit the data (e.g. using a linear regression where a curvilinear one is needed). It is patentlly clear that the numbers you report are NOT simply luck.

  16. Nate says:

    Anonymous,

    What makes you think the distribution is not random? What clear pattern do you see?

  17. Dan Schlauch says:

    I see what anonymous is saying, even though he's confused. The top 4 teams are likely going to be the top 4 seeds in the playoffs. Additionally, the top 2 teams in terms of expected wins (IND and NO) are both in the top 4 for luck. However it stops there, after that, there is actually a slightly negative correlation between expected wins and luck.

    Anonymous should also remember that part of the reason we consider those top 4 teams to be excellent teams is their superior record, which is a result, in part, of their luck.

  18. Unknown says:

    Not sure if this has been mentioned before, but isn't it a problem that no team is expected (according to your model) to win 10 games? Does this indicate that all teams will likely win a few games due to luck, or are there simply typical game elements not being measured here? That said, I can certainly agree that the Vikings owe some of their success to luck (49ers game and Ravens game come immediately to mind).

  19. Jack says:

    So when you say a team is "lucky", is this a good predictor of future events? Are the Vikings going to regress toward their GWP, or is it possible that they are in control of their luck? Which specific scenarios during a game count as lucky?

  20. Brian Burke says:

    I don't have much time, so I'll try to answer as many questions as I can quickly.

    "Clutch" is definitely part of it. Whether someone can chose to be better than he otherwise is in the clutch is another discussion.

    This is based on GWP, which is adjusted for schedule strength. As I wrote in the original post, teams do not control their schedules. If a team wins a lot of games because they've been served up easy opponents, that's part of luck.

    The "expected" part of expected wins refers to the statistical expectation at this point in the season (12 games in). It doesn't mean I'd expect them to win that many by the end of the year.

  21. Brett says:

    It's hard to deny the luck of the Saints and Colts this season. If you don't believe the stats, just look at the anecdotal evidence. How many come-from-behind victories does a team need before you call it luck? If this list is flawed, it is not because the best teams are at the top. That's just the way this season has been.

    However, there are certain players who make clutch plays week after week, year after year, like Manning and Favre, and off the top of my head I would add Troy Polomalu and the Freeney/Mathis duo on the defensive side.

    Notice, for example, how many Steelers losses this year have come on last-second defensive collapses in Polamalu's absence, and their win over the Titans was in no small part due to Polamalu's big plays in the clutch. I am wondering, Brian, if it would be possible to come up with some sort of individual clutch rating based on your existing data?

  22. Anonymous says:

    2 things I don't like about the reasoning expressed in the original post:
    1) Kick/punt returns for TD's aren't anymore lucky as to when they occur than any other long TD's; therefore, the Jets' missing "expected win" can be totally accounted for by their horrible kickoff return coverage in the Miami game; and
    2)Made/missed kicks at crucial times are the key metric in evaluating the quality of a kicker...two kickers with identical made %'s overall but very different %'s with the game on the line are by definition good or bad kickers...a 16 game season is enough to judge that...
    Jon Silverberg

  23. Anonymous says:

    Why is the difference between your model and reality termed "luck?"

    Is it not just a reflection of the inaccuracy of the model?

    If you model was "perfect", then there would be no lucky teams. There would be no difference between actual and predicted performance. "Luck" by your definition would cease to exist if you had a better model.

    But the concept of "luck" or "luckiness" has nothing to do with your model or how well it models reality.

Leave a Reply

Note: Only a member of this blog may post a comment.