88s and Heartbreak
(Bohm, avert your eyes. This gets a little rough.)
Comrades,
Once again this year Rob has given me the opportunity to
hijack WNQ so his disjointed and nonsensical ramblings are not the only public
face of our league. We cannot descend into a fascist state and allow this president
to suppress our voices. We must be heard and our glorious socialist paradigms
cannot go unheeded! Given Rob’s position of power and posting exclusivity, I
have limited confidence these published words are actually mine and not edited
beyond recognition. As a litmus test I offer the following: No one likes Rob
and he has small genitals. [Editor's Note: I'll have you know my genitals are in fact quite large and inflamed. - Bob]
Let Red October commence!
Alright, I can’t keep the Bolshevik stuff up any longer [Ed. Note: Cop out. If he was truly dedicated to this, he would continue]. I
opted to go analytical and data-heavy this year, so bear with me. This is
lengthy. Regarding the post title (forgive the Kanye reference) [Ed. Note: I approve of all Mr. Kardashian references, for the record], we’re taking a
historical look at the general luck of our league and reevaluating past
performance. I think we can all agree that fantasy is equal parts skill and
blinding luck. As a pretty established bottom dweller around here, it’s
convenient to blame my performance on bad luck. In my mind, there are really 2
types of misfortune when it comes to fantasy: the team-crippling-injury type [Ed. Note: As an aside, here is an interesting article discussing the rise in ACL injuries this year. Peter's players have been surprisingly absent. Is the curse broken?] and the week-to-week, my-opponent-flat-blew-up-and-I-lost type (aka you score
100 points a week and lose all 13 games if your opponents always score 101). We
all know I’m familiar with the former, but it turns out not so much with the
latter.
When I started crunching numbers, I was mostly interested in
this theory I had that your end of season, pre-playoff, rank was more dependent
on how well your opponents played against you (i.e. your Points Against) than
how skilled your team actually was (i.e. Points For). I was convinced that I
was frequently screwed by my opponent curb stomping me with banner weeks in
spite of my respectable point totals. The years in which I could barely field a
team with intact knee ligaments are clearly an aside. I set out to look at a
correlation between rank in Points For and Points Against as compared to
regular season final rank. For reference, I only used ESPN league years because
enough data just isn’t available from the Yahoo dark ages. Here’s how it played
out:
Rank In Points For To
End Of Season Rank
Rank In Points
Against To End Of Season Rank
In general, that worked out in line with my expectations.
The correlation coefficient for Points For is 0.577 and for Points Against is
0.671. Neither of those are great values, but at least the association seems
more linear for Points Against. So we can weakly conclude that it’s better to
be lucky than good. In my head that
confirmed it; I was getting screwed and that’s why I never make the playoffs.
Then I looked at the numbers a little deeper…
First some basic assumptions:
1) Your primary goal is just to make the playoffs.
Crazy shit happens in the playoffs and middling teams end up winning our
championship all the time. The variance is so high in fantasy and that luck
factor of how well your opponent performs on any given week is incredibly
strong. In single elimination playoffs someone gets lucky and wins. If you can
make the playoffs, you’ve got a chance at the title. [Ed. Note: Unless you're anyone besides Scott, Jason, Drew, or Ryan Good. Then you chance is zero percent.]
2) You only have to be average to make the
playoffs. Half of the league makes it (and for a long time it was 60% of the
league). Only once has a team with a better than .500 record missed the
playoffs and that was Kim with her endless ties last year. If you are
completely average and go 0.500 you’ll get in.
3) 88 points is your magic number (thus the title).
If you average our scores across all of the ESPN years, the weekly average for
points is 88.4. If your team averages more than that for the season you’ve got
a 84% chance of making the playoffs over the last 4 years (16/19 times). The
average weekly score of playoff teams is 93.3 points. If you get there or
better you’ve got 100% chance historically of making the playoffs.
4) You should assume your opponents will average
88.4 points per week as well. On any given week, if you score more than that
you will win more often than not. If you score less, you will lose more often
than not.
5) Sometimes you’re going to get screwed and
sometimes you’re going to get lucky. You might win scoring 75 points or lose
scoring 100. It happens, but it should even out at the end of the season right?
6) If it doesn’t even out for you over the course
of a season (or 4 seasons), you were either lucky or unlucky. You won too many
games scoring less than 89 or lost too many scoring more than 88.
7) If you were lucky or unlucky, your team was
probably better or worse than your win-loss record shows. I think point totals
are a better assessment of team quality than win-loss because win-loss forces
assessment of these minute differences in teams to a binary [Ed note: he used the word binary!] level. That binary [Ed. note: twice!] distinction of win or lose is super dependent on your opponent’s performance,
while in reality your team’s performance has absolutely no relationship to your
opponent (unlike real sports). This is that example of losing all 13 games
while scoring exactly 100 points every week. It’s possible and that would be some
bad luck for a pretty good team.
8) Understand that this year is a little different
now that we’ve added fractional points. Rob covered this once before, but I’m
estimating fractional points are worth somewhere in the range of an extra 3.5-5
points per week, based on 7 skill players earning fractionals which reasonably
average out to 0.5 points per player (with some wiggle room for extra
fractional points in that players non-traditional point scoring route; like
fractional running yards for a QB, or fractional receiving yards for a RB). The
average score through 7 weeks is 91.7, or 3.3 points more than the magic 88.4.
Good enough for me.
So
here’s what I did: I went back through the last 4 years and looked at how many
times a year each person got screwed (scored 89 or above and lost) or got lucky
(scored 88 or less and won). Theoretically that should even out over the course
of a season, but it doesn’t always. Yes I realize there are some confounders
like everyone doesn’t play everyone an equal number of times every year AND we
changed the number of players in the league diluting talent AND sometimes
people just give up on their season and it’s easier to beat them (see Godin,
Computer Buffalo Dung etc.) [Ed. note: I think we sometimes give Kyle too much credit for poor showing that year. While Computer Buffalo Dung and the Season of Doom was certainly fluky, at the time, since then Kyle has drafted his own team and been just as bad. Kim has computer drafted and made the playoffs before. I don't think Kyle is entirely to blame for his bad season, but he gets a lion's share.]. Get over it. I think it’s a decent way to look at
how lucky or unlucky we have been both overall and for individual years. Here
are some informative charts and my general conclusions (again, this is just for
the last 4 years). The luck aggregate is your balance of times lucky and times
screwed, with a negative number meaning you got screwed. The other 2 charts are
overall records with an adjusted record correcting for your luck aggregate.
Ties are a half win and a half loss, with adjustments in the luck aggregate for
a lucky tie (aka scoring 88 or less and tying) or vice versa. The data:
Player
|
Luck Aggregate
|
Drew
|
2.5
|
Travis
|
-2
|
Davis
|
2
|
Peter
|
0
|
Jason
|
0
|
Rob
|
4.5
|
Kim
|
-1.5
|
Good
|
1
|
Donel
|
-0.5
|
Kyle
|
5
|
Scott
|
-6
|
Team Specific Stats
|
Wins
|
Losses
|
Percentage
|
Adjusted
|
Drew
|
37.5
|
14.5
|
0.721153846
|
0.673076923
|
Ryan Good
|
32
|
20
|
0.615384615
|
0.596153846
|
Ryan Davis
|
8
|
5
|
0.615384615
|
0.461538462
|
Travis
|
30
|
22
|
0.576923077
|
0.615384615
|
Rob
|
28.5
|
23.5
|
0.548076923
|
0.461538462
|
Jason
|
26
|
26
|
0.5
|
0.5
|
Kyle
|
23
|
29
|
0.442307692
|
0.346153846
|
Scott
|
23
|
29
|
0.442307692
|
0.557692308
|
Kim
|
22.5
|
29.5
|
0.432692308
|
0.461538462
|
Peter
|
22
|
30
|
0.423076923
|
0.423076923
|
Donel
|
18.5
|
33.5
|
0.355769231
|
0.365384615
|
Team Specific Stats
|
Wins
|
Losses
|
Percentage
|
Adjusted
|
Drew
|
37.5
|
14.5
|
0.721153846
|
0.673076923
|
Travis
|
30
|
22
|
0.576923077
|
0.615384615
|
Ryan Good
|
32
|
20
|
0.615384615
|
0.596153846
|
Scott
|
23
|
29
|
0.442307692
|
0.557692308
|
Jason
|
26
|
26
|
0.5
|
0.5
|
Ryan Davis
|
8
|
5
|
0.615384615
|
0.461538462
|
Rob
|
28.5
|
23.5
|
0.548076923
|
0.461538462
|
Kim
|
22.5
|
29.5
|
0.432692308
|
0.461538462
|
Peter
|
22
|
30
|
0.423076923
|
0.423076923
|
Donel
|
18.5
|
33.5
|
0.355769231
|
0.365384615
|
Kyle
|
23
|
29
|
0.442307692
|
0.346153846
|
The
last chart just re-orders the same data ranking by adjusted win percentage and
hopefully gives a better sense for actual overall team quality. There are 11 of
us here because I cut out Godin as there isn’t much to draw from his
performance last year. So what can we take away from this?
1.
Kyle has been terrible. Clearly that computer-drafted
year is going to skew his overall record toward worse, but he’s also been
incredibly lucky. His luck aggregate adjusts him to 5 more losses and a drop in
winning percentage of almost 10%. He goes from 7th best record to
worst. If you look closely, it was really all 2010 when he somehow had the league’s
best record while picking up FIVE lucky wins out of 9. He scored 88 or less and won FIVE times. How is that possible?
That’s the total balance of his luck
aggregate right there. He was 8th out of 10 in points for that year,
while having the fewest points scored against him by a wide margin (almost 6
points a game less than the next closest person). That’s a bad team that just
got lucky. If there’s ever going to be proof that win-loss is not an accurate
descriptor of team quality, it’s that.
2.
The opposite of Kyle is Scott, who adjusts to 6
more wins and an improvement in his overall win percentage of 11.5%. He jumps
in overall rank from 8th to 4th.
That’s an unlucky man. He’s a net negative 1 so far this year also, continuing
the trend.
3.
I’m not unlucky at all, netting out to zero for
a luck aggregate. Thanks to Kyle’s awfulness I do jump one spot from 10th
to 9th in overall rank though. That’s something.
4.
The other luckiest player? Rob, with an
aggregate of 4.5 toward the lucky side. I’ve been wondering how his
consistently terrible teams keep sneaking into the playoffs. He’s already
planning to miss them this year [Ed. note: 4-3 bitch! Second place snitches!].
What
does all of this mean for this year? Probably not much at this point. Through 7
weeks, using the league average of ~92, Kim and myself are both a +1 on the
lucky side. Jason, Drew, and Scott are -1 to the unlucky side. There’s still
plenty of time for that to even out.
It’s
interesting to do some playoff prognostication looking at that those initial
scatter diagrams, though. If you decide to use Points For (historical positive
predictive value of 84% for making the playoffs if your team scores better than
the league average), Rob, Travis, and myself are good bets. Dane and Scott are
really really close to the threshold. If you believe it’s better to be lucky
than good and use Points Against, the math gets slightly less convincing. Your
rank in Points Against correlates a little better with your final position than
Points For (remember 0.67 to 0.57 from the scatter diagrams). But, if you use actual
values for Points Against (aka how likely are you to make the playoffs when
your opponents average less than the league average each week) the positive
predictive value is worse than Points For at 74% (or 14/19 times). By that
metric Scott, Travis, Dane, Good, myself, and surprisingly Kyle are better
bets.
I
think it makes sense to combine both for prognostic purposes. You end up with
Travis, Dane, Scott, and myself on both lists. I’d bet on that list [Ed. note: Travis, Dane, Scott, and Peter are just the worst teams to be in the playoffs. The only way to make it worse would be to add Jason.]. And really
it makes sense that those are the people scoring the most points and having the
least points scored against them. Of course they will win more often. Kim,
Jason, Donel, Drew, and Davis: You have been neither lucky nor good.
If
you made it this far - congratulations, you officially enjoy fantasy football
and math as much as I do. We both need help. That’s all for now. Rob sucks.
“All
power to the Soviets”
8=========> Peter <=========8
