Randomness and Luck

by Bruce D
Here's some data showing the randomness and luck referenced in a post by Bruce D. (me) concerning "What's the Deal with the Falcons?" on Advanced NFL Stats.
My version of luck is similar to Brian Burke's version in that we're both seeking to find the randomness in the NFL, yet its very different in that Mr. Burke calculates expected wins based on team ability, and the difference between expected wins and reality is considered "luck".
My version of "luck" calculations look at each "luck" type play and assigns points (good+ or bad-) associated with each of these individual plays. In each of these cases, the good luck team gains luck points, and their opponent losses the same amount of luck points. Here is an excerpt from my humble web-site as to what I've deemed as at least partially lucky plays:

"Bad Luck explained:
Points either prevented or gained by what is deemed as a "lucky" means are analyzed and valued. These are events that happen in every NFL game, and are at least partially a result of randomness and luck versus actual team ability. Since randomness and luck are not repeatable, teams with good luck so far this season are penalized, and teams with bad luck so far are rewarded in advantage values. This is done for 2 reasons:
#1. Teams cannot be expected to continue to be as lucky or unlucky as their records and points data indicate.
#2. The general betting public tends to use points and winning performances to judge a team's value, and they may not take into account that some of those wins and points were due to luck.

Though some "luck" is caused by aggressive defenses and team play, at least some of each team's luck is most likely due to, well, luck!

Things the I60200 System deems as at least partially due to luck.
Interceptions. More a product of the QB than the defense. You gotta be in the right place at the right time.
Fumbles lost. Anyone can fumble at any time, but if you lose the fumble, bad luck for you.
Kick off returns for a touch down. At least a portion of each one is luck, and they are game changers.
Punt returns for a touch down. Ditto what I just said.
Missed or blocked field goals. Though not automatic, if you miss more than the other guys you faced, you have the bad luck here.
Blocked punts. Hard to do, but always gives good field position to the blocking team.

In all these cases the "good luck" team is given a negative value, and the "bad luck" team is given a positive one. This is to point out what is more likely to happen in the future relative to current team performance, and not what has happened in the past."

In order to show some proof of this point to, I've calculated the 2009 NFL season luck points (total good luck points - total bad luck points) for each team, showing these in their first 8 games vs their last 8, and showing the difference between them.

Team	1st half Luck	2nd half Luck	Difference
ari	-7.5	7	14.5
atl	3.5	-5.5	-9
bal	1.5	5.5	4
buf	12.5	-2.5	-15
car	-34.5	45	79.5
chi	9.5	2	-7.5
cin	7.5	-0.5	-8
cle	-8.5	6.5	15
dal	11.5	-9.5	-21
den	7.5	2.5	-5
det	-24.5	-35	-10.5
gnb	17	27.5	10.5
hou	2	-20	-22
ind	20	-10	-30
jac	-15	-2	13
kan	12.5	-2	-14.5
mia	16.5	-20	-36.5
min	19.5	5	-14.5
nor	17.5	7.5	-10
nwe	15	-0.5	-15.5
nyg	-7.5	-15	-7.5
nyj	-11.5	7.5	19
oak	-17	-2.5	14.5
phi	27	14.5	-12.5
pit	-14	-6.5	7.5
sdg	1	17.5	16.5
sea	0.5	-7.5	-8
sfo	-12	22.5	34.5
stl	-20	-22	-2
tam	-2.5	-5	-2.5
ten	-10	0.5	10.5
was	-15	-7.5	7.5

more often than not it seems just random, or opposite. Carolina went from -34.5 luck points in the first half of the season, to +45 luck points in the last half. San Fran, similar, and so on.

Here are the luck factors(average luck points per game) for each team through week 11 of the 2010 NFL season (negative numbers are the luckiest teams).


As you'd imagine, these extremes happen game to game, and during any series of games during the season. This is where I look for "value" in predicting game outcomes where some may not know that a portion of a teams scoring, wins, and losses are due to just randomness and luck.