Oct 13, 2020
This episode features journalist Michael Salfino; who writes for The Wall Street Journal, FiveThirtyEight, The Athletic and other top shelf publications. In a spirited and wide-ranging discussion, we cover all of this (and yes, we use the word “multiverse” a few times):
Here's some handy dandy links for you:
Michael Salfino's Twitter
Michael's article on Data In
Politics
Michael's article on Facial
Recognition and Race
Michael's article on Movie
Ratings
Episode Transcript:
Rob Collie (00:00:00):
Welcome friends for our second episode and our first non co-host
guest. We're very fortunate to have the one and only Michael
Salfino. Now this guy writes for some really prestigious outlets,
including the Wall Street Journal and FiveThirtyEight. And he's
primarily known as a sports writer, but I think you'll see that
he's a bit of a Renaissance man. And of course he's into data,
otherwise we'd just called the podcast raw, wouldn't we? I've known
Mike seemingly forever, even though we've actually never met in
person. And I've been following his work since probably the mid two
thousands. He and I even collaborated on a research project back in
the day when I was still at Microsoft and I was working on the
original, great football project. And that was the first time I
aimed these BI tools at football data, which was my hobby at the
time, but that was years ago.
Rob Collie (00:00:46):
And we recorded this episode. We recorded this at the end of week
three of the NFL season. And so you're going to hear some dated
references in there from a few weeks ago. And don't worry if you're
not into sports or football. The key thing to pay attention to is
really how his mind works. He blends data with his subjective
experience in a really, really, I think professional and smooth
way. One of the things I really like about him is how curious he
is. And of course, he's pretty funny. We had a really, really,
really good time with Mike and I think we're definitely going to
have to have him back on the show regularly. So without further
ado, let's get to that intro music.
Announcer (00:01:20):
Ladies and gentlemen, may I have your attention please?
Announcer (00:01:27):
This is the Raw Data by P3 podcast, with your host, Rob Collie, and
your co-host Thomas LaRock. Find out what the experts at P3 can do
for your business. Go to powerpivotpro.com. Raw data by P3 is data
with the human element.
Rob Collie (00:01:45):
Welcome, Michael Salfino. I've really been looking forward to this
for a long time. I've been reading and listening to you for well
over 10 years now, probably 13 years. You write for and contribute
to a number of different projects and outlets. Why don't you just
kind of give us a rundown of that?
Michael Salfino (00:02:02):
Sure. I guess the best way to say it is I'm more of a reporter and
journalist. So I'm just trying to use some of the more advanced
analytics tools to communicate the players and the teams that are
actually good and likely to remain good for the foreseeable future,
as opposed to some of the methodology usually employed by
journalists, which is more descriptive rather than predictive in
terms of forecasting.
Michael Salfino (00:02:31):
So I don't consider myself in any way some sort of like data guru
or even statistician. But I think I have a good sense, from my
background that originally began in the corporate world, of how to
actually translate some of these higher minded concepts and to
ferret them out so that readers are able to, I think, glean from
them a better understanding of the game. So that's all I'm trying
to do.
Michael Salfino (00:02:55):
Like I want to give my readers their eureka moment that I had when
I was a kid, when I discovered the yards per pass attempt statistic
in a book about professional football. And despite following
football so intensely from grade school through high school, I had
never heard of this statistic. I had never heard of it on the radio
shows that I listened to or the television broadcasts or in any of
the articles. So I was extremely annoyed that there was a statistic
that correlated to winning, they said 75% of the time, irrespective
of any other statistic.
Michael Salfino (00:03:29):
And it was an efficiency stat, which I sort of intuitively knew at
that time. So it was much better than a total yardage stat. And I
felt that I was cheated, as a person who is passionate about
football, to not know something so important and not to have been
told this through all of those years, thus far in my life. So I'm
trying to actually provide some of that information to the people
who read my stuff so that they can maybe better understand the game
that they're so passionate about.
Rob Collie (00:04:00):
At our company we have a saying, which is your metrics that you're
using, half of them should have words in the name of the metric,
like the word per.
Michael Salfino (00:04:10):
Yeah.
Rob Collie (00:04:11):
Or adjusted or indexed. If you're not doing it that way, if you're
just sort of using raw metrics, there's no denominator, there's no
correction in it. You're probably breathing your own exhaust in a
number of ways. But before we even get into that, I want to give
you a chance to name drop a bit. I'm just going to cue this
softball up there, right down the plate. What are some of the
outlets you write for, Mike?
Michael Salfino (00:04:33):
Well, I started a syndicating nationally in newspapers for fantasy
sports. That was after I wrote for, then it was Baseball Weekly,
but it became Sports Weekly. So they wanted somebody to cover
fantasy football. Then after the fantasy football, I started
working for some of the regional sports networks on the content
side. And from there, I was discovered by a Wall Street Journal
editor who liked article that I did on evolution and pitching
mechanics, as it related to a Yankee pitcher at the time, Joba
Chamberlain. He was criticized for suffering an injury and losing
effectiveness because he was unable to precisely repeat his
throwing motion when pitching.
Michael Salfino (00:05:17):
But the biomechanics on the evolutionary side said that that is
something that's just natural in athletics because nobody was able
to hunt the mammoth the same exact way. You know what I mean? And
so since sports is more akin to something like that, the ability to
precisely repeat movement is not something that comes naturally to
athletes. And just because you look at it when somebody gets hurt
and you see that they're different, we're not looking at it at all
the times that somebody doesn't get hurt or lose effectiveness. So
there's no way to actually know whether this ability to precisely
repeat motion, with slow motion video evidence is actually
predictive of any kind of success, or even if it describes
success.
Michael Salfino (00:06:02):
So the Journal liked that article and they wanted somebody. They
wanted, what they said was somebody who they said was like a real
nerd to write analytics driven pieces on the sports side. So I did
that for about 10 years. I still do it occasionally for them, but
now I mostly write for The Athletic on the fantasy side of things,
and for FiveThirtyEight on the analytics side of things. When an
editor from the Wall Street Journal went over to FiveThirtyEight, I
kind of went over there with him.
Rob Collie (00:06:33):
Just kind of minor publication, you know. Wall Street Journal,
FiveThirtyEight.
Michael Salfino (00:06:36):
I'm very fortunate. Yeah. I don't know like how this happened, but
I think I found a niche because I was able to produce reader
friendly content because of my journalism experience. And I always
had a passion for data, just coincidentally. So it was just a
perfect fit to be at a publication like the Journal where you had
to convey some of these higher minded concepts in a more sort of
earthy and lay person kind of way. So it was just a fortunate fit
really on my part. I've been lucky.
Rob Collie (00:07:11):
I think that's actually really the thing that was compelling. I've
been much more into fantasy football 13 years ago than I am today,
but I'm still doing it.
Michael Salfino (00:07:19):
Yeah.
Rob Collie (00:07:20):
But I mean, I used to read basically like the entire internet of
fantasy football content.
Michael Salfino (00:07:25):
Right.
Rob Collie (00:07:25):
And so I had exposure to plenty of different people. And I think
the thing that drew me to you and your oftentimes partner, Scott,
is the blend, sort of, of the nerdery, of the technical stuff with
also human element, which is, by the way, the tagline of this
podcast, data with the human element. So that blend and the ability
to explain it. One of Einstein's famous quotes is, "If you can't
explain something simply you don't actually understand it well
enough."
Michael Salfino (00:07:56):
Right.
Rob Collie (00:07:56):
It's not just a communication problem. It's even a symptom of not
having a good model in your head. So I think it's more than just a
communication skill. I think it's communicating those things simply
to yourself is the first part, before you can communicate it to
others.
Michael Salfino (00:08:12):
Exactly.
Rob Collie (00:08:13):
I've always enjoyed that. I understand from your perspective that
it would seem like fortunate accident and all of that. But as an
observer, as an outsider, I think it makes total sense. I mean,
you're able to take important concepts and make them broadly
accessible, and also not scare everybody off with it. It's not
elitist.
Michael Salfino (00:08:33):
Yeah. Maybe when I say that I'm fortunate, I think I'm fortunate in
really not having both feet in either world. Like having one foot
in each world may make me somebody who, in one of those sides of
the sports world would seem less than super qualified to be talking
about whatever it is that I'm talking about. But I think the fact
that I tow both of them enables me to kind of find a sweet spot in
communicating to the reader. You know? If that's clear.
Rob Collie (00:09:03):
Yeah.
Michael Salfino (00:09:03):
So the people on the analytics side, the people who go to like the,
the Sloan meeting and stuff, those people would maybe interpret my
understanding of some of the statistical concepts to be lacking.
But I'm still able to report on the stuff that they're doing, I
think, to the lay reader, in a way that's clearer and concise and
compelling, which sometimes they lack with their non journalistic
background.
Rob Collie (00:09:32):
Yeah. What I've discovered, and I didn't expect this. I've sort of
been on a similar journey in some ways. You can be 90th percentile
in a few different fields and feel relatively inferior because you
only tend to hang out and sort of consume the work of the 99th
percentile, and things like that. But then when you realize that
your particular profession or life isn't a single event sport, it's
more of a decathlon. And if you're 90th percentile in a bunch of
different things, the overall decathlon sport, you might be 99th
percentile when it's time to blend them.
Michael Salfino (00:10:06):
It's really interesting that you say it that way, because one of
the things that I've always done, and it's unusual, I think, in my
profession, is that I don't really consume a lot of sports
information. I voraciously consume all kinds of news and articles
and analysis, but it hardly ever relates to sports. And I was
talking to my friend, Cade Massey, who I've worked with in the past
and who is a professor at Wharton. And we were at lunch. I told him
how I didn't read sports very much anymore. I mean, obviously I did
prior to writing about it. And he said that his mentor, Richard
Thaler, from the University of Chicago, always told his students do
not read about economics, read about everything else. And then you
will learn the things from all those other fields that you can then
and apply to economics.
Michael Salfino (00:11:00):
And you reduce, I guess, being so sort of cloistered in whatever it
is that you're reading that you feel that you can only pursue those
paths. So in other words, if you're reading only about your field
of expertise or what you're actually writing about, you may feel
that that field of play is limited to only those ideas that people
have kind of presented. And you may feel that you don't really have
the reign, creatively, to maybe focus on things that you would have
naturally if you didn't feel that you're kind of stuck with what
everybody else is talking about. If that makes any sense.
Thomas LaRock (00:11:36):
I wanted to mention something real quick, because Rob had just
talked about the 90th percentile and that struck me. It's like
describing somebody who's consistent, slow, steady, not flashy or
over the top and excelling in any one area. And when I saw your
name come up as our guest, so I did a little research of course,
and I'm finding you in all these great places, such as
FiveThirtyEight. Just what came out today here, the little
transcript of the conversation you're having, this resonated with
me. You made this comment just today. It just says the NFC East is
now 2-9-1, by the way.
Michael Salfino (00:12:14):
Yeah.
Thomas LaRock (00:12:14):
And I love it because you're just stating a fact as a journalist, a
sports writer. And a lot of times I think these sports writers kind
of have hidden agendas to some degree, or they want to be a little
more flashy because they need that, the ratings, the hits.
Michael Salfino (00:12:28):
Yeah.
Thomas LaRock (00:12:29):
They want to be at 95 percentile. So they would just say, the AFC
is 2-9-1, and then they would go on and on and on about how
horrible everything is. You didn't do that. You just stated the
fact. You say, by the way, draw your own conclusion. And I love
that. And I've been reading more and more of your stuff today. And
I really like the way you communicate. Like you said, you don't
want the reader to be cheated. But you're also not going to bring
them to a place where you want them necessarily to go. You're going
to let them get there through your words.
Michael Salfino (00:12:58):
Exactly.
Thomas LaRock (00:12:59):
And, and I think that's just great.
Michael Salfino (00:13:00):
It's kind of like when I've taught in the past. The university that
I attended when I was a kid wanted me to come back and teach a
couple of journalism classes, and I did it for two semesters. But,
similarly, that was like sort of the approach, right. Where you're
just trying to make sure when you're editing, like somebody's
journalism paper, I feel compelled almost to just fix it. But what
you have to do is you have to kind of tell them what they're doing
wrong so that they understand it and that they can fix it. And
that's the same thing that I kind of do, especially with the
fantasy analysis and the player and the team forecasting, is that I
try to show what I'm doing so that they can then take those things
and apply it to whatever player it is that they're looking to
analyze. You know?
Michael Salfino (00:13:50):
So it's kind of the old saw about teaching somebody to fish, rather
than just giving them fish. But I really think Twitter is probably
... with what you were saying, as far as the concision of some
analysis, Twitter is just such a great tool for really sharpening
your ability to say things in a very condensed way, which I think
really heightens the ability of a reader to understand what it is
that you're saying.
Rob Collie (00:14:19):
Rob Scribbles in his notebook, must start using Twitter more.
Because I'm famously long-winded. There's a corner of hell where I
will be forced for all eternity to write conference abstracts, like
the abstract of a talk. It must be 150 words or less. Writing those
for all eternity would be pretty much my worst case scenario.
Michael Salfino (00:14:42):
Well it's weird because, one of the things that people do in my
field is they'll promote something that they've just written on
Twitter by the word count. And a greater word count is supposed to
equal a greater level of sophisticated analysis. And to me, it's
almost like a confession. Like if you just said, I just finished
2000 words on Kenyan Drake and the Arizona Cardinals running game,
it's like, why the hell do I want to read 2000 words on that? You
know? And what I always say, and this is, again, I think I have
been fortunate because at the Wall Street Journal, I was on the
print side. So I had to say these very complicated things in about
275 words or less. And sometimes when the subject matter would be
extremely involved when I was talking to like Stanford biomechanics
people about thoroughbred racing and why Secretariat was sort of a
unicorn and why horses haven't gotten faster, like human athletes
have. It's like, how am I going to do this in 275 words?
Michael Salfino (00:15:45):
But I did it. It was hard, but I did it. And I always tell people
who complain about a short word count that Lincoln saved the
Republic in 275 words. So if Lincoln can do that in 275 words, you
should be able to tell somebody why they should either own or not
own Kenyan Drake in 275 words or less.
Rob Collie (00:16:06):
I feel so indicted, you know? Yeah. But you know, it's another
saying, right? The mark Twain quote, "I apologize that this is so
long. I did not have time to make it brief."
Michael Salfino (00:16:20):
Exactly right.
Thomas LaRock (00:16:21):
Yes.
Michael Salfino (00:16:21):
That is exactly right. It's so much harder. And you know, what's
weird is a lot of times writers get paid by the word. And it's
like, guys, it's harder to write shorter than it is to write
longer. You know? So that's just such a stupid way of compensating
writers.
Rob Collie (00:16:37):
Yeah. And there's actually a parallel to that in the old Revenge of
the Nerds documentary with Steve Jobs and Ballmer and Gates were
their talking about how IBM was measuring their output by KLOCs,
thousands of lines of code. And if you go back and look at sort of
like the origins of Microsoft, it's these competitions between Bill
Gates and Paul Allen to see who could write a certain routine in
the least code possible.
Michael Salfino (00:17:04):
Right.
Rob Collie (00:17:05):
Like viciously. Like a single byte was enough to make a difference
for these guys. IBM was measuring it like you were lifting bales of
hay or something. It was just-
Michael Salfino (00:17:15):
Exactly. Yeah, that's interesting, how the same thing could apply
to fields that on the surface appear to be completely different,
the same objectives.
Rob Collie (00:17:26):
Yeah. We see that everywhere. Clients or perspective clients are
very often asking us, "Well, do you have any experience in our
industry?" And the answer's always yes. I mean, we have experience
in every industry.
Michael Salfino (00:17:36):
Right.
Rob Collie (00:17:37):
But the harder thing to appreciate is that it turns out that our
experience in other industries is going to be just as important.
I'll give you an example. Many, many years ago, I did a project for
one of the big four accounting firms on, I forget if it was FIFO or
LIFO, the first in first out or last in, first out accounting. So
we were doing these accounting financial models, basically to
assess tax liability. And then years later was talking to someone
who worked in a food warehousing business about their spoilage
model, when food would go bad in the warehouse. And it turned out
it was exactly the same problem.
Michael Salfino (00:18:15):
Right.
Rob Collie (00:18:16):
It used exactly the same formula, the same techniques. And we just
like copy pasted that whole pattern. And the industry isn't
actually the thing. There's almost like these forms of problems
that you do start to see.
Michael Salfino (00:18:29):
Right.
Rob Collie (00:18:30):
And it's fascinating.
Michael Salfino (00:18:31):
In my field, the less I know about a subject, usually the less
likely it is that I will make a mistake in that article. So in
other words, if I'm writing about something that I'm intimately
familiar with, I may rely a little bit on my memory and just flag
something very easy, or just not explain it well enough, or skip
over something. But if I don't understand. So when I write about
hockey, which I really don't have a passionate interest in, those
articles tend to be the clearest examples of my writing because I
almost have to explain it to myself before I could explain it to
the reader. And that's what I try to do you anyway, that's my
process. But it's so much easier to do when you literally do not
know much about the subject matter than it is when you are so well
versed in it.
Rob Collie (00:19:20):
Well, I'm feeling a little bit less indicted now. Because I have
been, for example, critical and took a lot of heat, within our
little community anyway, for a number of years about being so
critical about the way our industry operates, which is, at least
traditionally, you would always first go and build a data warehouse
before you'd ever put a dot on a chart. And I've been calling BS on
that for a very, very long time. And I had enough of sort of
bonafides from my background at Microsoft that I sort of felt
okay.
Rob Collie (00:19:49):
But there's like this 1,200 page bible written by Kimball. And you
must always say his name in hushed tones. It's like a
priesthood-
Thomas LaRock (00:19:59):
Wait, you're a Kimball guy, not an Inmon?
Rob Collie (00:20:02):
Exactly, right. I have no idea. I consider the fact that I've never
read this 1,200 page book to be a qualification her than the
disqualification. And that goes back to some things you were saying
earlier, Mike, which is that if you consume too much of your own
industry, you become much more vulnerable to groupthink.
Michael Salfino (00:20:22):
Yes.
Rob Collie (00:20:22):
And groupthink isn't always bad, but it almost always is. So I like
that. I like that you don't have a problem differing from the
consensus. Every year going into NFL season, there is a consensus,
one through 10 of who the top players are.
Michael Salfino (00:20:40):
Right.
Rob Collie (00:20:41):
That the industry just happens to just magically arrive at as if
it's the right thing. But it's really just like the kids walking in
Dead Poet's Society, how they just automatically fall into a
marching pattern with each other, right?
Michael Salfino (00:20:52):
Well you'll notice, once you're sort of in the industry and like
completely OCD about fantasy sports, that it's the drafts that take
place at the very early part of the season, before ADP, which is
average draft position, is even remotely cemented where you have
such a freedom. And a lot of people are intimidated by it and a
little bit afraid of it because they don't want to go on record
over-drafting or under-drafting a player, especially over-drafting
a player, like taking him too high based on where he eventually
settles in his ADP. But I find it liberating to just be able to
draft the players based on the value that I think they have rather
than what that sort of consensus groupthink is.
Rob Collie (00:21:36):
Groupthink runs business, largely.
Thomas LaRock (00:21:38):
Absolutely.
Michael Salfino (00:21:39):
Yeah.
Rob Collie (00:21:39):
Escaping that is a little bit scary. It's kind of like the coaches
that used to not, and a lot of them still don't, the coaches in
football who, or really in sport, who didn't follow analytics. We
used to always think why don't they go for it on fourth down, the
percentages are great. It's just that their goal isn't necessarily
to win the most games. Their goal is to not get fired.
Michael Salfino (00:22:02):
Oh yeah. Scott and I ... Scott Pianowski of Yahoo, who's my partner
on the Breakfast Table, where you came into contact, I think, with
our work. We've always written about that, about how coaches ...
and this happens in business as well. They'll go for the quiet loss
rather than take the chance, increase their win probability. But in
the process, in defying the conventional wisdom, create a scenario
where they will be the subject of much more intense criticism if
they fail to win. Which, in either case is going to be the extreme
likelihood anyway.
Rob Collie (00:22:39):
Mm-hmm (affirmative). Yeah.
Michael Salfino (00:22:40):
But they'd rather just go down in a way where the reporters aren't
going to ... I always say it's like when you're playing a game. Say
you're playing Strat-O-Matic, or even Madden or a video game or
whatever it is, with sports. You want to win. You really want to
win. But you're going to do things that just maximize your win
probability, intuitively. It's not that complicated to know that if
it's fourth and two, you should just go for it. Right?
Michael Salfino (00:23:05):
But teams ... We don't have to worry when we're playing a game that
somebody is going to say that we should be fired and never be
allowed to play the game again. So I think that that is so
constraining on ... and I wouldn't even call it creativity or
courage, but it's almost like a willful ignorance of maximizing win
probability just because you are afraid of losing your job, which
maybe is rational.
Rob Collie (00:23:32):
Yeah. It is. It is. The lesson here isn't for the individual. In
business, for instance, if we were going to take a lesson away from
this. It's not for the individual. We're not saying, "Hey, you
should be more courageous." This message is actually more for the
managers, that they need to create an environment in which the
right outcome is optimized for.
Michael Salfino (00:23:51):
Right.
Rob Collie (00:23:52):
It's much more about the environment, because the environment does
create different incentives. It creates incentives that are
different. Don't stand out, don't get fired. Right. I think people
are intuitively doing the math in their heads and coming to the
right conclusion, in general. Don't stand out, you know?
Michael Salfino (00:24:08):
Yeah.
Thomas LaRock (00:24:08):
There's a reason Norv Turner held onto his job for as long as he
did.
Rob Collie (00:24:15):
We got to give Mike a chance to react to that. You know?
Michael Salfino (00:24:18):
Well, I think with the birth of analytics in sports that maybe the
coaches are going to become, especially in the other sports, maybe
not necessarily in football, are going to become more like middle
managers, right. And the general managers and the analytics
department are going to be more of the key decision makers. And I
think that's a double-edged sword. I think the teams will generally
be better. But the problem with analytics as it relates to a field
like sports, is that it fosters an environment where really teams
start playing the same way. So some of the variability that existed
previously, especially like in a sport like baseball or basketball,
where teams had stylistic ways of playing the game that was
inherently interesting to watch. And now analytics is kind of
smoothed that out so that teams generally play the same way.
Michael Salfino (00:25:11):
They may not play the same way as well. And that's obviously the
difference between success and failure in the individual sports.
But everybody is basically playing the same game in the NBA. And
they're definitely playing the same game in Major League Baseball
now. And for a while that was the situation in hockey with the
Devils where everybody just dumped and chased the puck. And
football, I think we're being saved right now by the Ravens, who
have managed to find a high level of success in playing a
fundamentally different brand of football, mostly because of a
quarterback that has unique skills.
Thomas LaRock (00:25:46):
So re you in one way saying analytics is making sports boring?
Michael Salfino (00:25:50):
Yeah. And I talk about this with my colleagues all the time. I
think baseball has a natural ability to defy the analytics on a
team level and maybe find, ironically, better analytics and being
better able to exploit market inefficiencies because there's the
variability of the ballpark geometry. And so you can theoretically
create a ballpark that is going to hurt every other team that is
playing the same analytics way while just helping you. You could
have 470 feet to center field, like the old Polo Grounds, 360 feet
down the line, 390 feet in the power alleys. And it's like, just
try to hit home runs here, guys. It's not happening.
Michael Salfino (00:26:36):
And then you build a team based on the things that are no longer
valued in Major League Baseball, or as valued, which would be
defense and base running and stealing. Strikeouts would obviously
still be valuable, but you could even have more of a contact staff,
especially a fly ball staff with all that expanse in the outfield.
But nobody really wants to do that. The Mets were sort of situated
like that with with Citi Field when it first opened. And then the
hitters complained and they just shortened the fences.
Rob Collie (00:27:09):
You need a ballpark that's got those movable walls, you know? So
every single game you're adjusting the outfield fence. You're like,
"Oh, this team's got a lot of lefties."
Michael Salfino (00:27:19):
Yeah Exactly right. Right. Yeah.
Rob Collie (00:27:23):
A variable geometry ballpark.
Michael Salfino (00:27:24):
Which is the way ballparks used to be, just because of the fact
that they were in neighborhoods. And they were limited as far as
the geometry is concerned to whatever the surrounding areas were.
But it made for, I think, a more interesting game. Then we got into
the cookie cutter thing with the ballparks in the 70s, which were
all astro turf and cut the same way for dual purpose football and
baseball. And now we're at a point really where the home run is
king. And so all these parks are much easier to hit home runs in,
generally speaking, than they used to be.
Rob Collie (00:27:54):
Let's go back a little bit. We started off talking about yards per
pass attempt. Not everyone here is into football, but I still
think, again, going back to that thing we were talking about, the
cross domain experience is oftentimes even more valuable than
in-domain experience. For about 20 years, at least 20 years, as
long as I've been paying attention, in the football world there's
been this ongoing argument. And I think it's largely been resolved
now, but it took forever. And it used to go like this. So there's
basically two kinds of plays you can run in football. You can hand
the ball to someone and have them run, and try to block for them.
Or you can drop back and you can throw the ball through the air. So
there's passing plays, which is throwing. And then there's rushing
plays, which is just running.
Rob Collie (00:28:40):
Obviously you can make bigger chunks of yards. You can throw the
ball a lot farther. And a really good pass is going to generate a
lot more ... you're going to go a lot further down the field than a
really good run, on average. And yet conventional wisdom forever
was that a powerful running game was the number one thing you could
have in football. It was the most valued thing, by a lot of people
anyway. And if you went and you looked at the statistics, you would
see that the number of yards you accumulate rushing per game did
correlate very strongly with how often you won. But there was a
problem with that, right? What was the problem?
Michael Salfino (00:29:21):
Well, first of all, there were a couple problems with that. But the
main problem that is just maddening to me as a football fan in
2020, that you get professional analysts and coaches still talking
about when Emmett Smith or now I guess Ezekiel Elliot gets 20
carries or more i a game, the Cowboys never lose. Or their record
is incredibly like a .900 winning percentage. But that's just a
product of already having won the game, usually with your passing
game. And so you're just increasing the rushing attempts. And it's
just so obvious. I swear I knew this when I was 11 years old that
well you're winning, so that's why you're running. But the
interesting thing is if you look back in the history of football
and there's this guy, Bud Goody who's I guess sort of like the Bill
James of football, he passed away probably within the last decade
or so. But he worked until like his nineties, working with coaches
and with teams. And starting in the mid-60s he developed the yards
per pass attempt stat and found its correlation to winning.
Michael Salfino (00:30:29):
And if you look back, even on the Vince Lombardi Packers in the
super bowl era, and that's the epitome of like running to win, is
Vince Lombardi. He's like the Supreme example of that. But they
actually dominated, if you look at the statistics, at an historic
rate in yards per pass attempt for, minus yards per pass attempt
allowed, which is sort of that net YPA stat. They were about four
yards better gaining than what they allowed. Which if you're like
two yards better, you're one of the 98th percentile teams in NFL
history. So that's just a ridiculous margin to have. So even going
back to the Dawn of the Super Bowl era, it really was a pass
dominated league, but nobody really knew it. But now there's just
no excuse not to know it.
Michael Salfino (00:31:24):
So one of the things that I did with the Journal, we had access to
a stat service. So I wanted to see, in the intervening period,
which was many decades, whether this 75% correlation of winning
yards per pass attempt in a game, and we call it net yards per pass
attempt because it includes sack yard in passing yards, and sacks
as pass plays. I wanted to see if that 75% threshold was still
true. And so they went all the way back to the merger and it was
like 74.6%. So if you don't know anything else ... and that's way
better than turnover differential, which is another stat that's
cited by people who want to find correlations to winning. Because
turnover differential is, as you know, so much more extreme than
any differential. And not even relevant to every game, because you
could have the same number of turnovers, where net yards per pass
attempt is always going to be won, even by the smallest margin, by
one of the teams.
Michael Salfino (00:32:23):
So this includes every game, and no matter what the margin, 75% win
probability for the team that wins that stat in each game. So
obviously teams, if they were rational, would be building teams to
win the battle of the passing game. They would be focusing on
receivers, pass protection, and quarterback on offense. And they
would be focusing on pass rush and cornerbacks and defensive backs
on defense. And they wouldn't really care about anything else,
especially in a salary cap situation where your resources are
limited by the rules of the game.
Thomas LaRock (00:32:57):
There are so many factors, and especially in football. I remember
Wayne Winston talking about how if you could ever figure out what
11 guys to put on the field for any one particular point in time,
you'd make a lot of money because nobody knows how to do that. But
it's also the game situation. I think you talked a little bit about
that. But just the net passing yards itself, there's so much that
go into it. the down, the distance, the clock, everything about it.
And unless you factor that in, you could talk about 75% of this,
whoever wins this stat wins the game. but I'm like, there's so many
points in time for you to get to that. There's more layers than
just looking at any one stat and correlating it with a win, in my
opinion.
Michael Salfino (00:33:39):
Sure. But there is going to be probably one stat that's more
important than all the other stats. So even though there are many
exceptions and there are ways where maybe you can distort that
statistic. Garbage time, by the way, is not one of them, which is
something I learned in working with Rob's data back in his
Microsoft days, which I found very interesting at the time, which
has never changed where it's actually harder for teams to generate
yards per pass attempt in garbage time than it is in regular time.
So that's why the efficiency stats are the stats that you should be
looking at.
Rob Collie (00:34:13):
We're kind of microwaving this meal, and I want to slow cook it for
the listeners. So let's slow it down just a little bit. Right? So
we started off by saying that, Hey, in the dark ages 20 years ago,
and it does persist to this day, which is still maddening, I agree.
But 20 years ago, except for a few outliers, everyone believed that
a powerful running game was the key to winning. And if you looked
at the surface of it, like hey the teams with the most rushing
yards do tend to win more.
Rob Collie (00:34:44):
But as you said, I want to make sure that the non-football crowd
knows this. In a game, once you've got a sufficient lead on the
other team, you start running the ball a lot more often than you
start passing the ball, simply because of this dynamic in football
where the clock keeps running at the end of a running play. But a
lot of passing plays result in the clock being stopped. So the way
to run out the clock is to run the ball.
Michael Salfino (00:35:10):
Right?
Rob Collie (00:35:10):
And so you tend to ... every rushing play tends to be positive. So
if you're running the ball more, you're just going to amass more
rushing yards.
Michael Salfino (00:35:19):
The way it's said is, and I ascribe to this, is that you pass to
beat the opponent and you run to beat the clock.
Rob Collie (00:35:28):
That's right. That's right. Okay. So this is an example of a metric
that if you just looked at it ... and keep in mind, this is an
industry that's been around for a very, very long time, with a
tremendous amount of money on the line.
Michael Salfino (00:35:40):
And scrutiny.
Rob Collie (00:35:42):
You would think that "the market" would've figured this out a lot
faster than it has. And as you point out, there are still some
stalwarts who get on there and parrot the old line, the old line
about the running game. Okay. So people started to slowly figure
out, okay maybe it's not the running game, maybe it's the passing
game. And so then you would naturally go, okay, it wasn't rushing
yards. Let's look at passing yards. But this was also a little bit
misleading because what happens when you fall behind in a game, if
you're trailing, you tend to throw the ball a lot more, so you
generate more passing yards that way.
Rob Collie (00:36:22):
So this yards per pass attempt, it starts to factor in. And I think
this is the thing, Tom, that you were saying, yards per pass
attempt is actually multiple metrics. You're integrating a lot of
information. Because the number of attempts that you throw is
indicative, especially late in the game, is indicative of the game
situation and the time that remaining. A Lot of things just sort of
get accounted for. But then there's also the yards per pass attempt
that your defense allows. And so now you've got the defensive side
of the ball. And so if you have your offense yards per pass attempt
on the positive side, your defense is sort of negative, right? And
you balance those out. And if your net yards per pass attempt,
defense versus offense, is positive. That thing that Mike talked
about, the old Vince Lombardi teams had a differential between
their offense and defense yards per attempt was four, four yards
per attempt. That is a massive number. That is just ridiculous.
Michael Salfino (00:37:24):
And the thing about it is, that's great about the stat is that it's
a power stat. You maybe aren't utilizing that power, but that power
always exists for you to use. So if you are generally winning the
battle of the passing game or if you have great team passing
efficiency, even if you don't pass a lot, like the '72 Dolphins
famously were a running team, but they passed at an extremely
efficient rate. But they controlled whether they felt like doing
that or not. You know what I mean? You couldn't stop them really
from passing. It's just whether they chose to. And that's still the
case in football.
Michael Salfino (00:38:00):
Russell Wilson used to be a great example of this, not necessarily
anymore, because he's having what's on pace to be an historic
season. But here's a guy who was always extremely efficient as a
passer, but the Seahawks just wouldn't pass that much. But whenever
they needed him to pass late in games, when they were trailing,
they just unleashed Russell Wilson and would win. Now you could
say, why not unleash Russell Wilson early in the games and just
blow teams out of the water. So now they're being forced to, by a
subpar defense. But there was nothing stopping them from just
beating teams into oblivion with Russell Wilson early in the games
in the past, except for just poor decision making.
Thomas LaRock (00:38:38):
So I think another layer here that is especially true for football,
it's really your opponents matter a lot because the schedule is
balanced only by division. So you know, this year my Patriots get
to play the NFC West, and we have to go on the road to Seattle and
somebody else. So that's not the same. It's not even the same for
everybody in my division because the Seahawks will play at a couple
of AFC East teams. So it gets close. You get to play each other.
And maybe it'll be more equal without a crowd. But the opponents
matter so much when you're looking at a stat like this, there's
just no question.
Thomas LaRock (00:39:12):
But you mentioned the dolphins and I would just say that's a great
example of a team that was so good at running. Every other team,
all their opponents would try to focus on stopping the thing that
would kill them the most. And you would gamble that you wouldn't
get beat with the pass, but they did. And it's like that in every
sport. I watch my Celtics get beat. I say, you know what if Iggy
makes four, three pointers and we get beat, I'm willing to give up
that chance because I've got to guard these other guys. And you
know what the damn guy sticks those three pointers and we're going
home. And that's a huge thing because at some point you have to
take a chance.
Michael Salfino (00:39:46):
Because you're maximizing the probability.
Thomas LaRock (00:39:48):
Exactly.
Michael Salfino (00:39:48):
Yeah. So it's rational.
Thomas LaRock (00:39:49):
But it leads to this ridiculous stat where people try to tell me
how great the '72 Dolphins were at passing. And I'm like, but they
weren't.
Michael Salfino (00:39:56):
Well, I think they maybe could have been if they did it more. It's
very interesting, the whole chicken and the egg thing. A theory
that I subscribe to was that your ability to run well did not
increase your ability to pass well, that the two areas were,
despite conventional wisdom and how easy it is intuitively to
accept the fact that, well, if you're running well teams will jump
your running game and then you could throw behind them more easily
and you could trick them. But when you really think about it, if a
defense is stuffing your running game, there's nothing they want to
do more on earth than to keep stuffing your running game. So like
why wouldn't a play action work, even if they're having a high
level of success and limiting your running game. There's blood in
the water, man. Those guys just want to crush you.
Michael Salfino (00:40:41):
And it turned out, with the more sophisticated data that we have
now and utilizing my friends Cade Massey and Rufus Peabody and
their excellent Massey-Peabody site, that there really is no
correlation between run success and the ability to throw. There's
really no correlation, which is extremely counterintuitive, between
running the ball well in a game and the success of your play action
passing game, where you're faking a run.
Rob Collie (00:41:07):
Yeah. So I remember following you for a little while when you were
talking about this, this wasn't recently-
Michael Salfino (00:41:11):
Oh, I've been talking about this since like 2000.
Rob Collie (00:41:14):
Yeah. I'm going to harvest a memory from a while ago and this just
might no longer be state of the art for you. But it's something
that I remember you talking about a while ago, which is that it was
really just your willingness to try running it, that set it up.
Like if you were willing to run every now and then just to kind of
keep them honest.
Michael Salfino (00:41:31):
Yes.
Rob Collie (00:41:31):
Even if you were terrible at it, as long as they were guessing as
to what the play call was going to be, and they were off balance,
it didn't really matter whether you were getting three yards every
time you ran versus getting four and a half.
Michael Salfino (00:41:44):
And a lot of it was the quality of the fake.
Rob Collie (00:41:46):
Yeah. Yeah.
Michael Salfino (00:41:47):
You know, because it's a game of deception.
Rob Collie (00:41:50):
Here's a really nerdy thing that I love. And it's a bit of a reach,
but I have to bring it up. So if you have position ... We're
talking about derivatives for a moment like physics, calculus
derivatives. If you take the first derivative of position over time
is speed. The change in where you are in are, in a unit of time,
that's speed. Okay. The second derivative of position, with respect
to time, is acceleration. The first one's how fast is your position
changing? The second one's how fast is your speed changing? Now the
third derivative of this, how fast is your acceleration changing?
The only word we've got for it is jerk. If you're sitting somewhere
and suddenly the acceleration of your body changes rapidly, that's
when you tend to lose your balance. It's an unexpected change in
acceleration.
Rob Collie (00:42:36):
So only living things with brains and nervous systems care about
jerk. Jerk has no physical impact on anything. Buildings aren't
more likely to fall over from a high jerk. They only care about
acceleration. So it's just this interesting thing that even in
physics, when you start taking more derivatives, you get to a point
where the living thing's ability to anticipate becomes crucial. It
just like the ability to anticipate comes out, even in basic
physics, basic calculus, when you're running derivatives like this.
And it just seems like to me ... Again, I'm a 2:00 in the morning
in the dorm room kind of nerd, those kinds of conversations. And to
me that's the same thing, right? If you can't anticipate what's
coming, that's all that it takes. That's the real value of a
running game in football.
Michael Salfino (00:43:26):
And the same thing in baseball with the changeup and the fastball.
Or you could see now and you could see it even with the pitchers of
a generation ago, when the arm slot matches perfectly and the
release point is the same and they do the overlap of the two
pitches you could see, from a hitter's perspective, how pitches
that end up at wildly different places in the strike zone appear to
be exactly the same for 7/10 of the journey to the plate.
Rob Collie (00:43:56):
Yeah. Information processing.
Michael Salfino (00:43:58):
Yeah.
Rob Collie (00:43:58):
It's the same thing why in the history of warfare weapon has always
beaten armor. There's never been armor that came out that defeated
the weapons of the age because the weapon gets to choose where it
hits, and the armor doesn't. The weapon has an information
advantage, right?
Michael Salfino (00:44:16):
Yeah. It's kind of like what we were talking about before with
passing.
Rob Collie (00:44:18):
Yeah.
Michael Salfino (00:44:19):
Yeah.
Rob Collie (00:44:19):
So I got one other football thing that I thought would be neat to
talk about. And again, it's one of these counterintuitive things.
The most devastating single thing that a quarterback can do, even
this I might be uncertain of, but if you're a quarterback and you
throw a lot of interceptions, you're constantly throwing the ball
to the wrong team.
Michael Salfino (00:44:39):
Yeah.
Rob Collie (00:44:40):
That is really going to hurt you.
Michael Salfino (00:44:42):
Not all turnovers are the same. Interceptions have a much greater
... and that's one of the things that I talked about early on in
the Breakfast Table, which nobody really knew, could articulate
why. But interceptions are much more costly than fumbles. And it's
got nothing to do with return yards.
Rob Collie (00:44:58):
Right. Interceptions are a much better predictor of failure than
fumbles.
Michael Salfino (00:45:03):
Now I have a theory about that, but you never know if your theory
is right. I remember I went to a symposium at the Museum of Natural
History. And this was said by the physicist who was giving the
lecture. But his whole point was that facts are basically
meaningless. They're so low in the pecking order. They're just
things that are out there. And we tend to think that facts are more
important than theory, right? Because people will say, "Oh, you
have the theory of evolution, but it's not a fact." And it's like,
no, the theory is over the fact. facts need a theory, otherwise
they're just things that are just floating out there. So unless you
have a coalescing theory as to why those facts exist, the facts
don't even really matter.
Thomas LaRock (00:45:49):
Wow. That's deep.
Rob Collie (00:45:50):
It is. Yeah. So it's not the dots, it's the lines. It's the lines
you draw through the dots that are the important thing. Yeah. I
like that. I'm going to be going around saying that, with my
fingers steepled, for the next 18 months. That's great.
Michael Salfino (00:46:12):
There's certain things that happen in sports. For a while, Ben
Roethlisberger was horrible on the road. And it's like, well, okay.
Drew Brees being horrible on the road can be explained by the fact
that he plays indoors. Why would Ben Roethlisberger be horrible on
the road? But it went on for like two years. And then it ended in a
sudden flash of great road performances in 2018. But I never really
pushed it, like so many of my colleagues did, because they were
obsessed with the fact of how Roethlisberger was playing on the
road, where I could not find a rational theory to explain why he
was playing poorly on the road. So I just ignored it. I just
figured it was noise.
Michael Salfino (00:46:54):
But I was always open to the idea that somebody would come up with
something where I was like, "Yeah, that makes a lot of sense. I
could buy that."
Rob Collie (00:47:00):
Oh, interesting. You were open to the existence of a theory.
Michael Salfino (00:47:03):
Yeah.
Rob Collie (00:47:04):
But because we didn't have one-
Michael Salfino (00:47:06):
I just ignored the data.
Rob Collie (00:47:07):
You just ignored it. I don't know that I'd be that strong. I do get
fooled by randomness a little bit. Plus you oftentimes just don't
get time to wait for statistical significance.
Michael Salfino (00:47:18):
As far as the randomness, maybe it was in The Drunkard's Walk or
whatever, the book on randomness, I don't know, wherever I read
this. But it's basically that if something is random, it's always
going to have the appearance at some point of not being random. You
just have to have the discipline to just ignore it, unless you
could figure out a reason why something is predictive and not
merely descriptive. So if you're flipping a coin, you know you're
going to have a stretch where it comes up heads or tails like 80
out of a hundred times, if you flip it long enough, right. Or eight
out of 10, let's call it. Maybe that's more likely. But it won't
mean anything, because we know it shouldn't mean anything.
Michael Salfino (00:48:00):
In the real world we don't know if those things are random. So if a
CEO has been a CEO of one company and they have a great product,
was he just randomly in the right place at the right time? Or was
the company successful because of his or her leadership? It's kind
of like the Pat Mahomes situation, right? Is Pat Mahomes a 95th
percentile hall of fame quarterback, no matter where he would've
been drafted? Or is he a hall of fame 95th percentile quarterback
because he was put into a 95th percentile situation for a young
quarterback?
Thomas LaRock (00:48:39):
That's also deep. And I'm hoping that he is 95% because I have him
starting tonight.
Michael Salfino (00:48:44):
Well that's the thing. The environment is still 95th percentile,
right? Like he's got great players all around him and a great
offensive coach. And it's harder to fail in a situation like that.
See, like in football, there's no true skill level, I maintain. Or
maybe, obviously, Pat Mahomes is highly skilled. So I shouldn't say
that there isn't any. But what I'm saying is it's not transferable,
like it is say in baseball .I could put Mike Trout on any team and
he is Mike Trout. I put Pat Mahomes on the jets, he may not be Pat
Mahomes.
Rob Collie (00:49:16):
That's true.
Michael Salfino (00:49:17):
He probably won't be. We actually can say definitively, he would
not be Pat Mahomes. We just don't know the extent to which he
wouldn't be Pat Mahomes. I mean, would he be Sam Darnold or would
he be something much better?
Thomas LaRock (00:49:33):
Well, can we just use LeVeon Bell as the test case for this?
Michael Salfino (00:49:36):
Sure. Yeah. But a lot of people would say he took a year off, he's
older. But I think you're ... I would agree with you, that the
environment fundamentally changed. There was really no explanation
otherwise as to why his performance would decline as precipitously
as it did.
Thomas LaRock (00:49:52):
So can I ask one more football question?
Michael Salfino (00:49:54):
Sure.
Thomas LaRock (00:49:55):
Because I'm going to challenge you on something, because you
mentioned earlier how garbage time doesn't matter.
Michael Salfino (00:49:59):
Sure.
Thomas LaRock (00:50:00):
I was going to bring this up as well, because in basketball it, was
a Bill James? They have, the lead is safe. Is this lead safe?
You're up by so many points. There's so much clock and so many
possessions.
Michael Salfino (00:50:11):
Yes.
Thomas LaRock (00:50:11):
Football doesn't have that equivalent, is this lead safe, that I've
found. But let's just say that existed.
Michael Salfino (00:50:19):
Well, they do have win probability.
Thomas LaRock (00:50:20):
But let's just say this existed, that there's a way where you'd say
this lead is safe. I'm up by two plus scores with two minutes to
play and-
Rob Collie (00:50:29):
Guys, we have to make a Falcons joke here. Any assessment of when a
lead is safe has to start with the question, yes/no, are you the
Falcons?
Michael Salfino (00:50:36):
Here's a stat. Here's a stat that's amazing. The Falcons are the
first team in NFL history to lose two games where they had a 17
point lead in the second half, or at any point in a game, in the
same season. And they did it in consecutive weeks. Consecutive
weeks.
Thomas LaRock (00:50:51):
That's awesome.
Michael Salfino (00:50:52):
Is that random?
Thomas LaRock (00:50:56):
No, no. I think it's a trend. So here's my counter example for you.
And it goes back to the early 2000s. So maybe you might remember
this. But there was a quarterback in the league at that time. And I
was the commissioner for our fancy football league, and so I was in
tune with all the games and I would do writeups for the games. And
this quarterback excelled in garbage time.
Michael Salfino (00:51:20):
Mark Bulger?
Thomas LaRock (00:51:21):
No.
Michael Salfino (00:51:22):
Okay. Because that was the one that we studied back around that
time, with Rob.
Thomas LaRock (00:51:27):
This one that I picked up on was Aaron Brooks.
Michael Salfino (00:51:31):
Oh yeah. Okay. Aaron Brooks.
Thomas LaRock (00:51:32):
He played for the saints. He would have horrible stats for three
quarters, but in the fourth quarter he was the king of the fourth
quarter. He'd pick up 14, 15 points. It was all garbage time
points. By the end of the game, he'd have 20, 22 points. And as you
know, you compare quarterbacks, how did my quarterback do against
the other quarterback. I had 22, he had 25. So it was even. Aaron
Brooks always came out, at the end of the day, fairly competitive
with whoever he was lining up with. But his NFL team was horrible.
Everything about it was horrible. The only way he score points was
just in complete garbage time. So when you mentioned that garbage
time doesn't matter, I'm like, "I don't know."
Michael Salfino (00:52:08):
But you see how it could matter in fantasy. I'm never going to say
it doesn't matter in fantasy.
Thomas LaRock (00:52:14):
Okay.
Michael Salfino (00:52:15):
Because you want pass attempts. And we always say the greatest
thing that could happen to your fantasy quarterback, unless you
have like an insane tax for it in terms of negative points, like I
do in one of my leagues, is the pick six, because he comes right
back on the field and now he's got to throw even more.
Thomas LaRock (00:52:31):
Yeah.
Michael Salfino (00:52:31):
So it's like manana, right? But we're talking about different
things. You're talking about the scoring of fantasy points. I will
not deny that garbage time can significantly increase a
quarterback's fantasy scoring. Like Matt Ryan on the other end,
when he's trailing, is a perfect example of that. He's the king of
that right now. But what I'm saying is that prevent defense's work.
It's harder to throw. So your efficiency stats will be worse, not
your counting stats.
Thomas LaRock (00:53:02):
Right. So yeah, two different things. Gotcha.
Rob Collie (00:53:04):
Something you were saying earlier, Mike, that I really wanted to
dig into. The only thing, and I mean this, the single thing that I
learned in college that I considered to be useful, one thing. It
was in a psychology class and it's this thing called attribution
bias. And I think it might be called different things. But
basically it's like when we look at someone's performance, whether
in life or in relationships or in sports, we tend to human beings,
we tend to overwhelmingly allocate their success or failure to
attributes of the per person.
Michael Salfino (00:53:45):
Right.
Rob Collie (00:53:46):
We under count, dramatically, the input and the impact of their
situation.
Michael Salfino (00:53:54):
Right. So like with Mahomes, what we were talking about. Perfect
example. Yeah.
Rob Collie (00:53:58):
That's exactly right. It goes way beyond sports. It's
everywhere.
Michael Salfino (00:54:02):
Well, like the CEOs that I was ... the hypothetical CEO that I was
talking about.
Rob Collie (00:54:06):
Exactly. Right. And so this actually comes back to also the fooled
by randomness thing as well. Right? How do you separate? And I
agree with you that football is a wonderful example of this. Pat
Mahomes can only be Pat Mahomes in a situation that is going to be
conducive to him. With Jimmy Johnson of the Cowboys' dominant era,
would Jimmy Johnson have allowed Pat Mahomes to be Pat Mahomes?
Probably not. You know?
Rob Collie (00:54:35):
And also, here's one that Tom will appreciate. Tom Brady, who, as
far as I know, is not known for his levels of graciousness. I don't
think that's something he's known for, has said that Aaron Rodgers
would be amazing if he got to play for Bill Belichick.
Michael Salfino (00:54:54):
Yeah.
Rob Collie (00:54:55):
Better than me, way better than me. You have no idea how many
titles Aaron Rodgers would win playing for Bill Belichick.
Michael Salfino (00:55:01):
Which is why Mike McCarthy should be indicted for his head
coaching. And he's doing the same thing in Dallas you got Dak
Prescott throwing for like 7,000 yards a game.
Thomas LaRock (00:55:09):
Hey, just relax. We're okay with the job he's doing for Dallas.
Michael Salfino (00:55:14):
Because they should be 0-3. Think of how much happier the world
would be. There would be a much higher level of happiness worldwide
if the Cowboys, who are so close to 0-3, were actually 0-3 right
now. 1-2 is good, but we would be dancing in the streets if the
Cowboys were 0-3, like 95% of the world.
Thomas LaRock (00:55:34):
That's an example of the schedule though. They were fortunate to
play Atlanta. That's all.
Rob Collie (00:55:39):
Yeah. Two forces conflicted. Atlanta must always lose.
Michael Salfino (00:55:43):
That's one of the 2-9-1 wins. The 2-9-1 NFC East, that's one of
those wins. And the other one was NFC East versus NFC East, is the
Redskins of all teams. What if I told you the Redskins would have
the other win?
Rob Collie (00:55:57):
Wait, I don't-
Michael Salfino (00:55:57):
Well actually the Washington Football Team.
Rob Collie (00:55:58):
That's right.
Michael Salfino (00:55:59):
I did this in the chat too. I am so sorry because I am so ... I
totally believe in the politics of the Redskins name change. And I
actually like ... I made the joke on the FiveThirtyEight chat two
weeks ago that it's like The Band. They should just call it the
Washington Football Team.
Thomas LaRock (00:56:16):
Yeah. I agree.
Michael Salfino (00:56:16):
It's a great name. That is so cool. So I hate when I slip, but I
can't ... there's a ghost in the machine, man. It's been too many
years. I can't let that go as easily as I want to, morally.
Rob Collie (00:56:29):
Mike, you've also written, at least in the past, I think you've
written for some psychology magazines. Isn't that right?
Michael Salfino (00:56:34):
I did. I wrote an article about facial recognition and why witness
testimony me is completely invalid, because we're face blind to
different races. And I wrote an article about the movie rating
system and how basically everybody was gravitating towards PG-13,
and violence was not something that they used to assess the
suitability of a movie for an age group. So they were just jamming
in all the violence that they wanted to. And the only thing they
had to worry about was using the F word in a sexual context.
Rob Collie (00:57:16):
Or more than once.
Michael Salfino (00:57:17):
Or more once.
Rob Collie (00:57:18):
Because you're allowed one non-sexual F bomb.
Michael Salfino (00:57:21):
You are. You are. Exactly.
Thomas LaRock (00:57:24):
So, in the show notes, can we link to these articles? Can we
remember to do that?
Michael Salfino (00:57:29):
Those are old ones.
Thomas LaRock (00:57:30):
I don't care. They sound fabulous.
Michael Salfino (00:57:32):
I think the face blindness was related to the Duke ... It was
pegged to the Duke rape scandal where somebody was criticized of a
different race for not being able to recognize the faces of the
accused, white assailers. But that's just a natural product of the
face blindness with race that I cited, based on other people's
research.
Rob Collie (00:57:58):
So you also write for FiveThirtyEight. Do you stick to sports on
FiveThirtyEight?
Michael Salfino (00:58:02):
I do. I would love to write about politics, but I could not be
trusted. It would be like writing about the Jets. I'm too partisan
to write about politics.
Rob Collie (00:58:13):
Being partisan, it's so rare in this day and age. I mean, you're
such an outlier.
Thomas LaRock (00:58:23):
Most of us are just middle the road.
Rob Collie (00:58:24):
It's actually almost hard to trust someone that isn't partisan
these days. Right?
Michael Salfino (00:58:28):
Nowadays, yeah. You know, it's so weird. I won't say really like
what side I'm on. Although I guess you could kind of figure that
out.
Rob Collie (00:58:38):
You could probably infer it.
Michael Salfino (00:58:39):
And this may be a male, female thing. I am far from a social
butterfly. My wife is very socially outgoing. She had a big problem
in this hyper partisan environment with having a lot of friends who
were on the other side, and it was very upsetting. I had no
problem. Every single person ... and it's not that many. There must
have been some sort of test that had nothing to do with politics
that I used to ferret people out, so that there was no one I knew
who was on the other side. And you know, this is many generations
of friends and probably 20, 25 people in total, at least, who I've
stayed in pretty close contact with. To be 25 for 25 is kind of
crazy.
Rob Collie (00:59:28):
That's awesome.
Michael Salfino (00:59:29):
But there's got to be something. I don't know if it's a male,
female thing. I don't know if it's a social, anti-social thing. But
maybe it's being judgmental. Maybe it's a bad quality in most cases
that actually accrue to my benefit in this circumstance.
Thomas LaRock (00:59:47):
Doesn't that lead to confirmation bias? If you're surrounded
by-
Rob Collie (00:59:52):
Yeah, it sounds like groupthink to me.
Michael Salfino (00:59:52):
Probably. Probably. Yeah.
Rob Collie (00:59:56):
They're all taking McCaffery at 1.1. All right. So politics though,
right? Without the content of the politics, let's talk about sort
of the analytics side of it.
Michael Salfino (01:00:09):
Yeah, just the numbers. Yeah. I love the numbers, man. This is at
the sports level for me. So, go.
Rob Collie (01:00:16):
All right. So I've given you a little hint of this in the past, but
I have a bit of a controversial ... I don't think it should be
controversial, but its proven to be every time I bring it up with
smart people. I mean, I get attacked. I get labeled as
anti-intellectual and written off really quickly before they even
really have a chance, I think, to hear what I'm really trying to
say. Let me present the thesis in the least controversial way
possible by saying that analytics, when applied to presidential
elections, it's just bullshit. It just doesn't work. You know, it's
like that old saying, so when you slap a probability on an event,
like a 65% chance of this person winning. I think it's like The
Naked Gun joke of he's got a 65% chance of winning, but only a 40%
chance of that.
Rob Collie (01:01:07):
I think in situations where we have the benefit of many trials and
many sort of similar circumstances, I think we have a much better
chance at developing models that put a percentage chance on an
outcome that is about as good as it can be. I don't think that any
of our percentage predictions about something as big and as
uncommon as a presidential election, I don't think it's that kind
of game. I just don't think it is. I think the assumptions that the
human beings building the models, assumptions that they're forced
to make in order to build the models, are themselves the vectors of
so much noise. And then combine that with the fact that on the day
of the election, we'll say something like there's a 65% chance,
before the first votes are cast, right? There's a 65% chance of
this person winning.
Rob Collie (01:01:54):
It's not the same as a sports game. A Sports game, you don't know
whether that ball is going to be ... when it's spinning and the guy
tips it, is he going to hit the strings of the laces of the ball or
not. That's going to change the trajectory of the tip. And the
whole game can turn on something as small as that and completely
random. Whereas, on the morning of an election, who's going to win,
I think it does not hang on anything remotely like that. Who's
going to win is already sort of known to the universe. It's just
not known to us yet.
Michael Salfino (01:02:28):
I see what you're saying. But don't you think that the margin of
the outcome can be random or can be attributed to randomness? So in
other words, Trump won by 77,000 votes across the three states. If
we had that election again, without anybody saying, "Oh my God,
what did I do? I didn't even think Trump was going to win. I
would've definitely not voted for a third party candidate," or any
stuff like that. If we just had it again, without any fore
knowledge, do you think that Trump would've won those three states
by that combined margin again? Do you think that was just set in
stone for perpetuity?
Rob Collie (01:03:07):
Well, I mean, it's sort of like the law of large numbers, et
cetera, right? The handful of small, random events that lead to
someone voting or not voting or changing their mind-
Michael Salfino (01:03:18):
Maybe there were 100,000 people who wanted to vote, but just got
into an argument with a spouse or-
Rob Collie (01:03:26):
Right.
Michael Salfino (01:03:26):
Or the kid broke his leg. Who knows what could have happened, got
sick.
Rob Collie (01:03:30):
That's right. I believe those things happen. But I believe that
they happen in a way that tends to not sway the outcome very much
because Democrats or Republicans more likely to get arguments with
their spouses that morning. It's just like ... it's hard to have a
lot of those random events pull in the same direction.
Michael Salfino (01:03:46):
Definitely Republicans.
Rob Collie (01:03:47):
Okay. Yeah.
Michael Salfino (01:03:48):
I don't want to be partisan.
Rob Collie (01:03:48):
No. Don't. You know.
Michael Salfino (01:03:48):
No. I have no idea.
Rob Collie (01:03:50):
I mean, you know. I think that the range of initial conditions in
which the randomness that can happen in a single day would sway the
outcome from one to the other, I think that's incredibly tiny. I
just think that that happens like never. We never find ourselves in
a situation where the amount of randomness that can happen in a
single day will change the outcome from blue to red or vice versa.
I just don't think we find ourselves in those initial conditions.
It's just not that kind of thing.
Michael Salfino (01:04:18):
I thought the problem with the viability of the predictions for the
last election were extrapolating national margins and not really
focusing on the state polls, because historically the likelihood of
the national polls consensus winner correlated so well to electoral
success. But, I think even though that had happened previously with
Bush and Gore, I think it happening again has changed the nature of
the polling for this cycle.
Michael Salfino (01:04:52):
So even though the national margin that everybody relied on was
more or less in line with the forecast, I don't think you're going
to have that problem again this time.
Rob Collie (01:05:04):
Right.
Michael Salfino (01:05:04):
But the thing that interests me the most about what you said is
whether a sporting event can be analogous to a percentile or a
percentage prediction of a political outcome. And I guess that just
has to do not necessarily with being able to assess how the people
are going to vote at an individual level, but just creating some
sort of uncertainty regarding the polling metrics that were used.
Would that be a fair way to say what ... and plus, we only have a
single trial outcome, right?
Rob Collie (01:05:40):
Right.
Michael Salfino (01:05:41):
So if there was, like some people said, a 95% chance or whatever it
was of Hillary Clinton to win the last election, maybe in the
multiverse she won 95% of the times, but we just happened to be in
one of those 5% cases. But you rejected that when we spoke the out
it previously.
Rob Collie (01:05:58):
First of all, I respect any method by which we get the word
multiverse onto the podcast. This notion that quantum randomness
actually results in forking universes from every point in time, so
you get the multiverse. I don't believe that. I believe that even
though there is a lot of quantum randomness that happens every day,
that leads to unpredictability, you just can't predict the future.
That over the large volume of voters, it just gets all kind of
netted out to zero over the course of the day. So, no, I think in
2016, I bet there were almost like 99.999% of all multiverses,
trump won that election. The only one's-
Michael Salfino (01:06:38):
You're depressing me.
Rob Collie (01:06:40):
Again, we don't know which side you're on.
Michael Salfino (01:06:42):
No, no, no. Hypothetically.
Rob Collie (01:06:45):
Yeah. Hypothetically, that might depress you if you had that ...
The ones where he lost are ones where he just randomly had a heart
attack that day and died, right? That's an example of one where he
would lose, he's just not around to win. And eating that diet also
doesn't doesn't help his chances.
Rob Collie (01:07:04):
I just think that it's what you said earlier about we made some
mistakes in our modeling last time that we're not going to repeat.
That, to me, is the other big part of it? It's really two key
points. One is that yeah, we do. We have assumptions that we make,
and this election we're likely to discover other methodology
mistakes. There's going to be a new set of methodology mistakes.
It's kind of like if they only played a football game in the entire
world, they only played it one game every four years, the amount of
change that would happen in the players and strategies and coaches
and everything, it'd be so hard to produce a good model there.
Michael Salfino (01:07:38):
But don't you think of model was good in the sense that it did
pretty accurately forecast the national margin? And now there's the
theories or the models that Nate has developed over at
FiveThirtyEight, which extrapolate the margin of polling
differential into an electoral win probability. And so anything
below 4% gives Trump a chance. But then once you get up to 6%, 7%,
8%, it's a very small chance that that can be overcome.
Rob Collie (01:08:10):
Well, this starts to remind me of the CEO example, right? You have
a product that succeeds and then you have a product that doesn't
succeed. You just don't get enough trials really to evaluate
whether a certain methodology is effective. Like this is the
highest stakes, small sample size, low number of trials. I think
the number of trials in this science is essentially almost always
one.
Michael Salfino (01:08:32):
But what about ... can't you take ... Just from of polling,
couldn't you take all the polling results and sort of grade the
pollers, like FiveThirtyEight does? And come up with a model that
was reliable. In other words, it's not just the presidential
cycles, but polling is polling, right? As far as the election, all
of those elections wouldn't those be reasonable tests of polling
accuracy?
Rob Collie (01:08:56):
Maybe. But two things. Number one, I think I've dragged this out
too long. Probably boring everybody to death. It's only interesting
to me. And two, I don't know. I mean a presidential election is a
presidential election. It's got different things going on. And
we're going to find out, right? We're going to find out, especially
in this election, we really don't know qualitatively what all the
extra factors are going to be like. We're even wondering ... There
is even articles right now wondering whether or not the loser of
the election accepts that they lose.
Thomas LaRock (01:09:27):
We're not worried about the loser accepting the results. We're
worried about one person accepting the results.
Michael Salfino (01:09:33):
But that person doesn't ... We don't need that person to accept the
results. Where is it written that the loser has to accept
losing?
Thomas LaRock (01:09:39):
It's true.
Michael Salfino (01:09:39):
Can the Jets actually just not accept losing? Would their record be
better?
Rob Collie (01:09:44):
I've often wondered that. If a team just said-
Michael Salfino (01:09:47):
"We won this game. No."
Rob Collie (01:09:48):
"You ruled it a turnover, but we're not going to give you the
ball." I've always wondered how long it would go before they'd
actually have to pay up.
Thomas LaRock (01:09:56):
Fake interception. It just didn't happen. Mulligan. I get to do
that one over.
Rob Collie (01:10:01):
Atlanta will come back and say, "Listen, you see what just
happened? That's not actually possible. We all agree that that's
not possible."
Michael Salfino (01:10:07):
Or Atlanta could just say you should take the probability of each
game and turn that into a winning percentage, and that's our
record.
Thomas LaRock (01:10:13):
Sure.
Michael Salfino (01:10:14):
Totally rational argument.
Thomas LaRock (01:10:15):
They'd still be 0-3, I think.
Rob Collie (01:10:17):
All right. Well, one thing I want to make sure we do is I want to
give the people listening ... I want to give them a chance to track
you down. You're Michael Salfino.
Michael Salfino (01:10:24):
MichaelSalfino on Twitter. I'm going to be doing a analytics site
for gambling and also for player forecasting in terms of prop bets.
That site is called Bet Prep. But it's just a tool for people to
sort of mine data, as opposed to ... It doesn't give you the picks.
It just avails the database and all of the historical games and
statistics to you for your strategies. And I write for the
Athletic, occasionally for the Wall Street Journal. And at
FiveThirtyEight I do the chats every Monday on the NFL.
Rob Collie (01:11:02):
And everyone that plays fantasy football with me, you can just tune
out right now. Just turn off your ears. It's cool. The podcast is
over. But for those of you who stuck around, you also do the
Breakfast Table podcast.
Michael Salfino (01:11:16):
Yes. Oh, I should have said that. Yeah.
Rob Collie (01:11:18):
Yeah.
Michael Salfino (01:11:19):
Okay. Members only.
Rob Collie (01:11:21):
Yeah, members only. I'm in the big spender bucket. I'm in the $7 a
month.
Michael Salfino (01:11:25):
Yeah. $7 a month. You get everything.
Rob Collie (01:11:28):
And I even leave it active during baseball season. I just feel like
that's how it ends up being a fair price for it. I don't turn it
off during baseball season. I never tune in.
Michael Salfino (01:11:38):
We made more money when we just let people give us whatever they
wanted, because some people would just like give us $100 or even
more.
Rob Collie (01:11:46):
Right. Yeah.
Michael Salfino (01:11:48):
But I didn't know. I have no idea the way to do it. We're grateful
that in a world where there are so many fantasy football podcasts,
that people think that this is a worthy product that they want to
pay for is flattering.
Rob Collie (01:12:01):
I know I'm trying to wrap this up, but it reminds me of the ...I
think this was in Freakonomics where the daycare was tired of
parents picking up their kids late. So they decided to impose a $20
fine.
Thomas LaRock (01:12:13):
Right.
Michael Salfino (01:12:13):
Yeah.
Rob Collie (01:12:13):
And instead, late pickups exploded.
Michael Salfino (01:12:16):
Yes, because it was like, "20 bucks. I don't even have feel
guilty."
Rob Collie (01:12:19):
Yeah. It's like the $20 guilt tax.
Michael Salfino (01:12:21):
Yeah.
Thomas LaRock (01:12:24):
Wait, you'll watch my kid for another hour for 20 bucks?
Michael Salfino (01:12:27):
Right. It's way more of a motivation to not have to feel guilty
than it is to just feel like you could pay 20 bucks and get the
service.
Rob Collie (01:12:34):
Yeah.
Thomas LaRock (01:12:34):
That's awesome.
Rob Collie (01:12:35):
Yeah.
Michael Salfino (01:12:35):
As a former parent with kids in those situations, trust me, I
would've been spending that $20 liberally.
Rob Collie (01:12:43):
I've really enjoyed this. I know Tom has as well. I told Tom ahead
of time that he was going to love you. And he said, "Well, I'm
married, but you know, maybe."
Michael Salfino (01:12:52):
Look, if Jets and Patriot fans can get together.
Thomas LaRock (01:12:55):
We can. Absolutely. I have no problem. Of course, I've been
successful for two decades.
Michael Salfino (01:12:59):
Exactly.
Rob Collie (01:13:00):
So Tom still associates himself with the team. Earlier he said like
we have to fly out to the ... he's on the plane, you know?
Thomas LaRock (01:13:07):
Absolutely.
Rob Collie (01:13:07):
He sits next to Belichick, going over film.
Michael Salfino (01:13:10):
Really the Patriot ... How could the Patriots have had this success
without Tom being a fan?
Rob Collie (01:13:14):
They say the fans important. They're really talking about Tom.
Thomas LaRock (01:13:17):
Oh no, no, no. I'm sorry, but I've been a fan through all the lean
years too. The Rod Rust era.
Michael Salfino (01:13:24):
Oh, stop with your lean. I'm so tired. The Boston sports victimhood
is gone, forever.
Rob Collie (01:13:31):
Yeah.
Thomas LaRock (01:13:31):
I'm just saying, I survived the Rod Rust era. I was fan when they
were 1-15 and the only win was against the Colts. I remember that
season. So I'm not a bandwagon. That's why I guess I'm trying
say.
Michael Salfino (01:13:44):
Yeah. And I totally respect. Let's put it this way. I have no
respect for bandwagon fans. You have to stick with it. People say,
"Why are you still a jet ..." Like my wife all the time. "Why are
you a Jet's fan? You don't have to be." It's like, "You don't
understand." I don't even mind. I'll wear my Jets hat when the Jets
are like, now. I'm not going to do it when they're playing well.
You know what I mean? Because I don't want to seem like I'm on the
bandwagon with the Jets. Now, this is when the true fans come out,
now.
Thomas LaRock (01:14:15):
Agreed.
Rob Collie (01:14:16):
Yeah. This is when the true irrationality comes out, where we
associate ourselves with something that is totally beyond our
control.
Michael Salfino (01:14:24):
Name something else that people could have loved since they were
like seven years old, the same way they love right now, the same
way they love those things right now. There's really nothing else
other than a sports team.
Thomas LaRock (01:14:34):
Peanut butter.
Michael Salfino (01:14:34):
You know what? I agree with that.
Rob Collie (01:14:37):
Yeah.
Michael Salfino (01:14:37):
I agree with peanut butter.
Rob Collie (01:14:40):
I stipulate peanut butter.
Michael Salfino (01:14:43):
Non food.
Rob Collie (01:14:46):
I love sports, obviously. At the same time though, I just don't
think that we need any mechanisms growing up that teaches
tribalism. I really-
Michael Salfino (01:14:57):
But we're tribal. So why not have something really safe, like being
tribal for a sports team than something that's actually
dangerous?
Rob Collie (01:15:04):
I Agree.
Michael Salfino (01:15:04):
Excluding Eagles fans, of course. Because that's dangerous. They're
dangerous anyway.
Rob Collie (01:15:08):
No one ever gives us the lesson as well, which is ... and don't
extend this to other parts of your life. This is the one place in
your life that you get to be like this. But when you put this down,
don't be this way. I would love that.
Michael Salfino (01:15:22):
But it's not like I'm a face painter, you know. There's tribal and
there's tribal. I think you could be a fan without being completely
... I'm like Groucho Marx, man. I'm not going to be in any club
that would have me as a member. There's no way I'm going to go over
the moon like some people do. I have to keep a somewhat ... I have
to keep a safe distance for appearances. But inwardly, trust me, I
have my face painted.
Thomas LaRock (01:15:50):
I'll give you a quick story. It was years ago. I'm watching the
game. You remember Jerry Glanville?
Michael Salfino (01:15:55):
Yeah.
Thomas LaRock (01:15:55):
He had stopped coaching and he was announcing. And this defender
did everything right. Everything right. Just perfect defense. And
the guy still made the catch. And Jerry said, "Well, when he gets
the sideline, what you say is the offense gets paid too."
Michael Salfino (01:16:11):
Yeah.
Thomas LaRock (01:16:12):
And the thing is, Jerry, he wouldn't be screaming at the guy. He
would just say, "You know what, we got beat by a good play." You
shake somebody's hand. And we're never taught that compassion, that
empathy for somebody else and just congratulate, "Hey, you know
what? You beat me." Instead. it's yelling and screaming. "You
suck."
Michael Salfino (01:16:28):
"You cheated. This is rigged."
Thomas LaRock (01:16:30):
Like, "I can't believe I lost at this." And that is what's really
missing. The tribalism isn't really the issue. It's having
compassion for the other person.
Rob Collie (01:16:38):
Yes.
Michael Salfino (01:16:39):
Yeah. So see, you Patriot fans should be more compassionate to us
Jet fans.
Thomas LaRock (01:16:44):
I do my best.
Michael Salfino (01:16:46):
Sure you do.
Thomas LaRock (01:16:48):
I have no issue with the ... I really don't. There are very few
teams in sports I hate. I don't hate the Jets. Why would I hate the
Jets? You're not good. There's no reason to hate you.
Michael Salfino (01:16:57):
And that's why we hate the Patriots. You just summed it up.
Rob Collie (01:17:01):
It's that condescending ... you know, that pity.
Thomas LaRock (01:17:02):
But no, it's true. I don't. There's very few teams I hate, and the
Jets are not on that list.
Michael Salfino (01:17:11):
Well, hopefully one day we will be worthy of your hatred.
Rob Collie (01:17:16):
We aspire. Yeah.
Michael Salfino (01:17:18):
Definitely.
Rob Collie (01:17:19):
Mike, we've enjoyed the hell out this.
Michael Salfino (01:17:21):
It's been a lot of fun. I hope it's good for you guys. I was just
winging it, so sorry if I-
Thomas LaRock (01:17:27):
No, you're good.
Michael Salfino (01:17:28):
Sorry if I talk too much.
Rob Collie (01:17:29):
Are you kidding?
Michael Salfino (01:17:31):
Okay.
Rob Collie (01:17:31):
Do you know where you came to?
Michael Salfino (01:17:33):
All right.
Rob Collie (01:17:34):
So Mike, we've got an election coming up. We'll definitely want to
have you back on after that.
Michael Salfino (01:17:39):
Oh my God. I may be ... you may have to contact me in Canada or
Italy. I found out I could move to Italy. They treat me as a
citizen. I could just go to Italy.
Rob Collie (01:17:48):
Really?
Michael Salfino (01:17:48):
Yeah. If you have any Italian blood, you can move to Italy, no
strings attached. Go there and they take you. My wife, get the hell
out. She's got to figure it out.
Thomas LaRock (01:18:02):
I have this. I can go?
Michael Salfino (01:18:03):
All you have to do is prove any Italian blood and you're Italian in
the eyes of Italy. It's like the mob.
Thomas LaRock (01:18:09):
I didn't know this.
Rob Collie (01:18:10):
I wonder if Luke has any Italian blood with the last name of
Pierazoli.
Thomas LaRock (01:18:18):
My great-grandparents came off the boat. I've got the blood, I've
got the documents to prove it.
Michael Salfino (01:18:22):
I'll see you there, November 5th.
Rob Collie (01:18:26):
As long as we're all not out in the streets fighting the second
American civil war.
Michael Salfino (01:18:32):
Oh my God.
Rob Collie (01:18:33):
We'll have you back on. How does that sound?
Michael Salfino (01:18:35):
Yeah, that'd be great. I appreciate it, guys.
Rob Collie (01:18:37):
Awesome.
Michael Salfino (01:18:37):
Okay.
Rob Collie (01:18:37):
Awesome. Looking forward to it.
Michael Salfino (01:18:39):
Thanks a lot.
Announcer (01:18:40):
Thanks for listening to the Raw Data by P3 podcast. Find out what
the experts at P3 can do for your business. Go to
powerpivotpro.com. Interested in becoming a guest on the show?
Email. Lukep, L-U-K-E-P@powerpivotpro.com. Have a day today!