Dear Fantasy Speed Skating Managers,
For the third weekend in a row, my train itinerary home was mangled
by cancellations and delays. By now, I should know that buying a ticket for a
German train is about as reliable as catching a flight out of Quebec City the Sunday
night after racing. On the plus side, Deutsche Bahn did a great job of emailing
me reminders about my unsalvageable itineraries. It’s nice to know that
someone
cares.
Caption: An actual screenshot
from Stepdad’s inbox
Fortunately, unlike the CTrain, it’s completely legal to
drink beer on the Regional Express 11 from Luxembourg to Koblenz (or any of the
37 other trains I took this weekend). I basically had no choice but to crack
open a couple of cold ones, boot up MS Excel, and crank out a column.
For those who missed draft night, your Stepdad and two other
managers (who are actual parents, unlike Dad, Stepdad, Your Biological Father,
and Mom) met in Metz, France to establish a temporary European headquarters for
FantasySpeedSkating.com. We spent most of our time buying baguettes, discussing
baguettes, and eating baguettes. However, we also held two formal workshops on
key Fantasy Speed Skating topics: the accuracy of Dad’s Power Rankings and the
validity of multi-manager teams. If you care about either of these things,
please read on. If you don’t, I recommend looking at some
Gary Larson cartoons
instead.
Are Dad’s Power Rankings Worthless?
Dad has published annual Power Rankings since the second
season of Fantasy Speed Skating. (In the first season, Dad just badmouthed
other managers’ teams in person). According to Dad’s recent column, he is
“usually right about this stuff.” With six seasons of Power Rankings behind us,
there’s enough data to look into this and ask the question on everyone’s mind:
Are Dad’s rankings actually any good?
An unnamed manager compared Dad’s Power Rankings to the
final standings over the last six seasons. In these seasons, Dad has never
successfully predicted the league’s winner, although he has been within two
places twice (Christian in 2019-2020 and Jake + Dylan in 2023-2024). Dad has
only successfully predicted the loser once (Anders in 2020-2021). If Dad’s
prediction for this year’s loser is correct, Darius and Anders will be the only
managers to have both drafted each other AND been accurately predicted to lose
the league.
However, looking at predictions of winners and losers in
isolation doesn’t really answer whether Dad’s Power Rankings are any good. A
better option is to use Spearman’s rank correlation coefficient, which measures
the similarity of two lists of ranked items. Spearman’s rank correlation
coefficient ranges from -1 to 1. A coefficient of 1 means that Dad nailed predictions
perfectly. A correlation of 0 means that Dad’s predictions were no better than
random chance. A correlation of -1 means that Dad’s predictions were perfectly
wrong (i.e., the order was the perfect opposite of what Dad predicted; see
footnote 1).
The correlations between Dad’s Power Rankings and the final
league rankings range from 0.42 in 2024-2025 to 0.6 in 2022-2023. All
correlations except for the 2019-2020 season are statistically significant at
the 0.01 level.
Caption: Rank correlation coefficients for Dad’s
pre-season Power Rankings and final league standings
Season
Rank
Correlation Coefficient
p-value
2019-2020
0.53
0.03
2020-2021
0.60
<0.01
2021-2022
0.48
<0.01
2022-2023
0.60
<0.01
2023-2024
0.59
<0.01
2024-2025
0.42
<0.01
On average, the correlation between Dad’s Power Rankings and
the final rankings is 0.54, which would usually be considered in the “moderate”
correlation. Basically, this shows that Dad’s Power Rankings are better than
completely random rankings. In other words, a drunk monkey throwing darts at a
list of managers would likely not do a better job than Dad. A sober monkey with
access to a pot of coffee and MS Excel might stand a chance though.
So is Dad a total shitter? Well, yes. But his Power Rankings
are not totally worthless. They certainly don’t guarantee you a win (or loss),
but they will likely be moderately related to your performance (sorry, Darius).
Do multi-manager teams have an unfair advantage?
The second workshop at European Fantasy Speed Skating headquarters
grappled with multi-manager teams. The first two-manager teams emerged in the 2020-2021
season. These days, there are several long-standing multi-manager teams, including
MCV, Kiersten + Hayleigh, and Eric + Emil. Peter and Brooke even have a team
together, although most people consider them more of a joke than an actual
team.
One manager has argued for years that multi-manager teams
should be banned. Their argument goes something like this: One of the hardest
parts of the draft is making a list of 80 skaters for draft night. With two
managers, coming up with a list of names is easier because each person only has
to come up with 40 skaters. Multi-manager teams thus have to put in less
effort, making things unfair. Furthermore, the manager argues that it’s unfair that
some managers essentially coast on their co-manager’s knowledge and don’t put in
the arduous labour required to draft and manage a team.
To be clear, your Stepdad doesn’t necessarily agree with
these arguments. Under their logic, if a team had 80 managers, each person
would only have to think of one skater, which would make generating a draft
list trivial, assuming that everyone thought of a different skater. I also
don’t really care if someone else coasts on their co-manager’s skills; I did
exactly that during the 2020-2021 season and it was great (thanks, Mom!). As an
anonymous manager on a three-person team noted, “a beautiful thing about
fantasy speed skating is people coming together for the love of speed skating
and multi manger teams create even more togetherness.” Dad might tell you that
Fantasy Speed Skating is about winning, but the real reward is getting to play
the game with your friends.
Leaving aside the question of whether a multi-manager team
is “total bullshit” or “chill and fun,” another interesting (and more testable)
question is if multi-manager teams perform better than their single-manager
counterparts.
To look into this, I ran a linear regression of normalized
rank on the number of managers. In theory, this should reveal if more managers is
more better. If you don’t care about the statistical hack job I committed, skip
the next paragraph and admire the beautiful graph below.
To run the regression, I normalized the results for every
year by dividing rank by the number of managers plus one to ensure that all
values were scaled between zero and one. I then regressed normalized rank on
number of managers. I can already hear haters (i.e., Dad) in the comments complaining
that “You really shouldn’t do that with ordinal data,” “Why aren’t you using a
Mann-Whitney U-Test?,” and “What the fuck man, the premise of this doesn’t even
make sense.” If anyone cares that much, I’m happy to lend them my TI-84
graphing calculator to sort this stuff out on their own. I’m a Stepdad, not a
mathematician.
Anyways, the results are shown below. We can see that there
is a small positive relationship between the number of managers and a team’s
performance. The R^2 value indicates that about 1% of the variance in ranking
can be explained by the number of managers. In other words, the number of
managers on a team is not a strong predictor of a team’s performance.
Caption: Team
Performance vs. Number of Managers
However, if we suspend disbelief and accept the trendline as
gospel, a team of between 7 and 8 managers basically guarantees a first-place
finish. Of course, the arguably optimal solution is just to have the whole
league compete as a single team, and then we’re guaranteed to win. I’ll run the
idea by the league organizers for next year.
Wrap-Up
Well folks, that’s a full recap of the sessions at Fantasy
Speed Skating’s European headquarters last week. If you disagree with anything
that you read today, please feel free to start a flame war in the comments or
DOX me personally.
I hope that everyone enjoyed draft night (or early morning)
and is looking forward to an exciting season of Fantasy Speed Skating. Thank
you for choosing Deutsche Bahn.
Yours in cancelled trains,
Stepdad
Footnote 1: Yes,
this would mean that Dad predicted the manager in the middle of the rankings
exactly perfectly. This is because Spearman’s rank correlation coefficient is a
measure of monotonicity (i.e., the tendency of two variables to go up and down
together). Go figure.
Footnote 2:
I have no idea if it’s legal to take an unweighted average of these things
across years. Sorry.