00:00:00
Welcome, welcome to Wharton Moneyball
00:00:03
here on the Wharton podcast network.
00:00:05
This is Eric Bradler, professor of
00:00:06
marketing, statistics, and data science
00:00:08
here at the Wharton School. And some
00:00:10
combination of four of us today. We have
00:00:13
Shane Jensen, professor of statistics
00:00:14
and data science. AI Winer, professor of
00:00:16
statistics, and data science, the three
00:00:18
of us, and Cade Massie, some of us are
00:00:20
here every week here on Wharton
00:00:22
Moneyball. Uh, as always, guys, uh, we
00:00:25
always have an open line segment, which
00:00:26
we'll do first. As you know, we're
00:00:28
working out some technical issues with
00:00:29
our guest today, and we'll see if we
00:00:31
have one, but there's a lot going on in
00:00:33
sports that can easily take our
00:00:35
listeners through a full hour of sports
00:00:37
and analytics here on Morton Moneyball.
00:00:39
So guys, why don't we start with, you
00:00:41
know, we always like to start with what
00:00:42
caught our eye, Shane, since you're on
00:00:44
my left and I reread from left to right.
00:00:47
Shane, what caught your eye in sports
00:00:49
this week? And there could be a lot.
00:00:51
>> Yeah, I mean, I'm sure we'll talk about
00:00:52
baseball, so I'll wait on that. But um
00:00:55
what caught my eye obviously is the
00:00:57
Patriots uh and potentially having a
00:01:01
real quarterback again, a franchise
00:01:03
quarterback. I don't want to get too
00:01:05
hyped up too early. We're still kind of
00:01:06
he's only in his second season, but
00:01:08
Drake May was absolutely incredible. Has
00:01:11
been incredible so far this season.
00:01:12
He's, you know, I mean, again, it's
00:01:14
early to discuss MVP, but he would be in
00:01:17
that conversation if we were to have
00:01:18
that conversation at this point. Uh so I
00:01:21
guess I'm just really excited about the
00:01:23
Patriots. They they're looking like a
00:01:24
playoff team.
00:01:25
>> Yeah. What kind of credit do you give
00:01:27
honestly Mike Vrabel very successful
00:01:30
coach just took over as coach. Um remind
00:01:33
me their offensive coordinator Josh
00:01:34
McDaniel right is back as their
00:01:36
offensive coordinator. So I hate to look
00:01:38
I'm not taking anything away from Drake
00:01:41
May but you've got an established head
00:01:44
coach that's a winner.
00:01:45
>> You've got an established offensive
00:01:47
coordinator that's a winner. Well,
00:01:49
winner Mike Fable. I mean, when you say
00:01:52
winner, uh, uh, Mike Fable is a head
00:01:55
coach. I mean, not not not his overall
00:01:57
Tennessee record. You're not pointing
00:01:58
to, right? You're not pointing to the
00:02:00
fact that he's led a team to the
00:02:01
playoffs before.
00:02:03
>> Yeah. And I'm not trying to take
00:02:05
anything away from him. I mean, it's
00:02:06
obviously an impossible tell. I mean, we
00:02:08
had like, you know, 15 seasons of Belch
00:02:11
and Brady co-observed and people were
00:02:14
still arguing about whether it was Brady
00:02:15
or Belich. I don't think we're going to
00:02:17
ever, you know, it's going to be hard to
00:02:19
keep kind of deconvolve those,
00:02:22
you know, certainly early on in his
00:02:24
career.
00:02:24
>> Look, we all believe in, you know,
00:02:26
uncertainty, right? And I think for 15
00:02:29
seasons for Bellich and Brady, there was
00:02:31
obviously a very strong confound. Are
00:02:34
you Mr. Patriot, Mr. I love both Bellich
00:02:38
and Brady. Are is there any are you
00:02:41
starting to at least put a small
00:02:42
probability
00:02:45
given what Bellich did at New England in
00:02:48
his last few seasons, what you're now
00:02:50
seeing at North Carolina,
00:02:52
is there even a small part of you that's
00:02:56
starting to give more credit to Tom
00:02:57
Brady?
00:02:59
>> I've always given a lot of credit to Tom
00:03:01
Brady. I mean, that's that's a softball.
00:03:03
Yeah. I mean, all the credit in the
00:03:05
universe goes to Tron R. But but I mean,
00:03:07
again, let let me stop short. I don't
00:03:09
want to take any I I what what the
00:03:11
Patriots did, which no other team has
00:03:13
ever done kind of for the the length of
00:03:15
time that they did it, was a collab, a a
00:03:17
successful collab between one of the
00:03:19
great the greatest quarterback of all
00:03:21
time and one of the greatest coaches of
00:03:23
all time. You know, I I do think Belch
00:03:25
has done more in the post collab period
00:03:29
to kind of take away from I guess his
00:03:31
leg whatever, but I still think Belch's
00:03:33
legacy is is is kind of set in stone
00:03:35
based on what what he did with the
00:03:36
Patriots. Would I still hire him as a
00:03:39
head coach? No. But I mean, Audi, you
00:03:42
want to jump in on this or on Well, you
00:03:44
if you want to jump in on this, please
00:03:45
do and then you'll get your own topic. I
00:03:47
>> I do want to jump in on this. I don't
00:03:48
think that that it's important to
00:03:50
recognize that as a football outsider
00:03:52
for all these years who was trying to
00:03:54
build my knowledge base. One thing I
00:03:56
have learned it's it's almost impossible
00:03:58
to be successful as a coach without a
00:03:59
great quarterback. So the fact that that
00:04:03
Bich hasn't been so successful in the
00:04:04
post Brady years is not necessarily that
00:04:08
particularly damning of Bichc's talents
00:04:10
over the years.
00:04:11
>> How do you define success, Audi? For
00:04:12
example, I'm just making up some names.
00:04:14
So for example, someone won a Super Bowl
00:04:16
with Brad Johnson. Someone won a Super
00:04:19
Bowl with Joe Flacco. Someone won a
00:04:21
Super I mean there are I'm not saying
00:04:23
you get Brady's level of success. Let's
00:04:26
not I'm not trying to compare.
00:04:28
>> Audi said great coaches. So I think
00:04:30
there's got to be some kind of sustained
00:04:32
success built into that. I don't want to
00:04:34
speak for you Audi though.
00:04:34
>> Audi, go ahead.
00:04:35
>> No, I mean I mean my question is is it
00:04:37
what what is it? What is the
00:04:38
information? Think of it from a from a
00:04:41
posterior updating. We saw years and
00:04:43
years of of Bellichic with Brady.
00:04:45
Fantastic. Right. And we said, by the
00:04:46
way, remember we saw years I you may I
00:04:48
know you know this audi we saw years of
00:04:50
Bellich prey too and it wasn't that
00:04:53
great
00:04:55
>> right? But the thing is is that there
00:04:56
must have been more to just Brady to
00:04:59
years and years of all that success with
00:05:01
the with the
00:05:02
>> Yeah. I know. And I mean, again, again,
00:05:03
we we kind of I think um Eric, you
00:05:05
talked I mean, or we we we think back
00:05:08
probably most recently to their kind of
00:05:09
end end dynasty kind of run where Brady
00:05:12
was in god mode and in the early kind of
00:05:15
dynastic year, you know, when they were
00:05:16
winning very successfully in kind of
00:05:18
Brady's very early career.
00:05:20
>> Yeah. Brady was kind of the game
00:05:21
manager. He did he was kind of wore what
00:05:23
we regard as an adequate quarterback
00:05:25
that was fine that and it was a
00:05:27
world-class defense which Bich basically
00:05:30
won those championships. So I I I mean
00:05:32
it's kind of I I do think we kind of in
00:05:34
in retrospect want to kind I I certainly
00:05:37
give Bichc a very big part of you know
00:05:40
that that success especially in the
00:05:42
early years.
00:05:42
>> The other thing to think about and this
00:05:44
is uh something very very much a feature
00:05:46
I think of football. the game changes.
00:05:49
>> Um, and it does go through all kinds of
00:05:52
shifts and non-stationarities in the way
00:05:53
the way it's played and that in some
00:05:56
level
00:05:57
>> you may have been a fantastic co coach
00:05:59
because you've mastered the skills and
00:06:00
in the environment of a certain era and
00:06:03
then if that if you just continue in
00:06:04
that that you slide into your old age
00:06:06
doing that same thing, you may not have
00:06:09
it anymore. in which case it's not to
00:06:11
that reflects badly on the past Bellich
00:06:13
check but this current Bellich check is
00:06:15
not going to do it in today's game.
00:06:18
>> Well, first you guys have brought up two
00:06:19
parts and I want to get to what caught
00:06:20
your eye but I would two parts. One is I
00:06:23
completely agree with Shane's part which
00:06:25
is you have to give Belichc at worst if
00:06:28
you wanted to be critical of him which
00:06:29
there's no reason to be but if you
00:06:31
wanted to be you have to give him credit
00:06:33
for those first few championships where
00:06:35
Brady was not the Tom Brady of age 30 to
00:06:38
40 or even 30 to 45. Um the defense was
00:06:42
amazing. His numbers were they were
00:06:44
good.
00:06:44
>> He was pretty clutch in the playoffs. I
00:06:46
mean he did win a couple
00:06:48
but like but yes I mean I think it was
00:06:49
definitely that was a team known for its
00:06:51
defense. would have been,
00:06:53
>> you know, kind of guess Russell Wilson
00:06:54
and with Seattle type scenario, but also
00:06:58
a your point about the non-station of
00:07:00
the game I think is also it's another
00:07:02
reason. It's not like I don't think
00:07:03
Bichc has the energy or the mind or the
00:07:06
you know it's age doesn't matter to me.
00:07:08
It's not like I think he's lost his
00:07:10
ability to coach people, but I think the
00:07:13
point is is that the game has changed in
00:07:16
a way that he would need to
00:07:18
significantly change the style in which
00:07:20
he coaches to be successful today. And
00:07:22
that's just hard to do. That's hard of
00:07:24
anybody. It's not. And also, it's hard
00:07:26
to do for someone that won six NFL
00:07:28
championships. And I'm not even counting
00:07:29
the ones he won as the Giants defensive
00:07:31
coordinator. And how many Super Bowls
00:07:33
the man has been to? Like, I don't know,
00:07:34
half of them in the last 35 years. I
00:07:37
don't know what the number is, but it's
00:07:38
some big number. But Audi, what caught
00:07:40
your eye besides the Patriots? By the
00:07:41
way, I watched a lot of the Patriots
00:07:42
game sh game. Drake May is for real, but
00:07:46
Audi, what are your thoughts?
00:07:47
>> Okay, so of course the last time we were
00:07:50
together, the Yankees and the Phillies
00:07:51
were still in it and now both have been
00:07:53
eliminated. Um so big questions going
00:07:56
ahead um of course is particularly with
00:08:00
the Yankees and I think the general
00:08:02
question the analytics revolution if you
00:08:04
if you want to condense it today and
00:08:07
into its biggest impact factors would be
00:08:10
the hitting revolution that has led
00:08:12
towards the large numbers of strikeouts
00:08:15
um and and the kind of three true
00:08:18
outcome hitting or whatever you know
00:08:20
>> they call that all that kind of crap
00:08:22
>> the three two outcomes hitting, but it's
00:08:24
essentially, it's interesting because it
00:08:26
it it's plays off the idea that a strike
00:08:28
out is no worse than a regular out,
00:08:31
which conditional on an out is true.
00:08:34
Conditional on an out. That's true. And
00:08:36
in fact,
00:08:36
>> no, that's true.
00:08:37
>> Well, it's tricky because a strikeout
00:08:39
obviously does not move the runner over
00:08:40
and is generally a sacrifice, but it's
00:08:42
al also not a double play. So for and
00:08:45
from someone who watched a lot of
00:08:46
Stanton grounders into double plays, I'm
00:08:49
thinking why couldn't just strike out? I
00:08:50
do think there's kind of there's an
00:08:52
analog push against almost like against
00:08:54
balls in play, you know, more than
00:08:56
anything. And and and I think certain
00:08:58
certain teams have kind of I think
00:09:00
there's almost a counter push on that
00:09:01
half.
00:09:02
>> Well, so it's interesting. You look
00:09:03
through the Yankees who with the
00:09:04
exception of Aaron Judge, which I'll get
00:09:06
back to in a moment, which is basically
00:09:08
awful at the plate and massive numbers
00:09:10
of strikeouts. No two two strikes
00:09:12
approach. I mean, I want to pick on
00:09:13
Trey.
00:09:14
>> High slugging, high slugging, low
00:09:16
batting average.
00:09:17
>> Well, that's during the season, during
00:09:18
during the playoffs. No, no, I mean,
00:09:20
okay, right. Um, so very low batting
00:09:22
average, but also very high strikeout
00:09:24
rates. And one one of the things you
00:09:26
noticed quite vividly in the playoffs is
00:09:29
no two strike approach, right? So,
00:09:31
particularly with runners on, I mean,
00:09:33
Trent Gisham is swinging for that porch
00:09:35
with everything he has.
00:09:38
>> What is Vulpi like doing like
00:09:40
>> and Vulpi is the same way, right? So,
00:09:42
Vulpi's hits nearly, you know, about 20
00:09:44
home runs.
00:09:45
>> Let me let me ask you a question. Let me
00:09:46
ask a question. It's not like these
00:09:48
hitters just became who they became.
00:09:51
It's also to Shane's other point, it's
00:09:53
not like whe you know whether it's
00:09:55
Cashman or Boone can't see, huh, these
00:09:59
guys bat 180 against really good
00:10:03
pitching. Huh, maybe a good player in
00:10:06
the playoffs. They can see the data. I I
00:10:09
do I do think playoffs are mostly coin
00:10:11
flip, but I'm starting to kind of
00:10:12
deviate away from that and that I do
00:10:14
think this kind of three true outcome it
00:10:16
works well over very large samples. You
00:10:17
hit a lot of bombs, etc. Uh but I do
00:10:21
think in in the in the playoffs when
00:10:22
you're only facing good pitchers and you
00:10:26
have to kind of do stuff in high
00:10:28
leverage kind of situations or or like
00:10:30
you know the little things seem to
00:10:31
matter or would get magnified. I don't
00:10:32
even know.
00:10:34
But but anyway, that that's when a sort
00:10:36
of a like kind of hitting to contact
00:10:38
like putting balls in play actually an
00:10:41
ability to do that as a team is really
00:10:43
helpful. I see Toronto and Milwaukee
00:10:46
doing it
00:10:46
>> for sure.
00:10:48
>> Absolutely. I mean, one of the things we
00:10:49
noticed about Toronto is that they were
00:10:51
just an impossible out. They were I
00:10:53
mean, eight, nine pitches, fouling
00:10:55
things off and often striking out. Yes.
00:10:57
But but tiring the pitcher, putting the
00:10:59
ball in play. But I will say um so one
00:11:01
of the things about so two things that
00:11:02
caught my eye specifically in the
00:11:04
thought around the game first of all is
00:11:06
how unbelievably awesome Aaron Judge is.
00:11:08
Let's get that out there. Um and I just
00:11:10
noticed he's about the two to one
00:11:11
favorite in the betting markets to win
00:11:13
the MVP. Um although one of my insiders
00:11:16
suggested that no Raleigh is going to
00:11:18
win it and and it's you can make a
00:11:19
strong case for either. I think um
00:11:22
>> let me ask you a question Audi. Do they
00:11:23
do something for the MVP like they do
00:11:26
for the Hall of Fame or No, there is no
00:11:27
pre-bound. It just gets announced. There
00:11:29
is no
00:11:30
>> well there's no prevalence. It just gets
00:11:31
announced. That's
00:11:32
>> just gets announced. Okay.
00:11:33
>> And I don't think anybody tells you who
00:11:34
their vote is. Um but I'm not sure about
00:11:37
that. But so it's you know you know
00:11:38
Judge had a season for the ages, right?
00:11:40
So he'll he had a probably about a 10
00:11:43
war. Um but just a batting line would
00:11:45
just which is absurd.
00:11:47
>> Feel like we talk about his MVP
00:11:48
candidacy a lot on this show. It's not
00:11:51
worry about it guys. He's your guys'
00:11:53
regular season champion. Don't worry.
00:11:55
>> No no but he had an incredible post. He
00:11:57
had an incredible postseason and one of
00:11:59
the things that we did see was that
00:12:01
possibly one of the most
00:12:03
unbelievably impressive home run I've
00:12:05
ever seen hit 100 mph fast ball 4 in
00:12:08
inside.
00:12:08
>> Yeah, that was uh very few very few
00:12:10
people could have done what he did there
00:12:12
for sure
00:12:12
>> just to have that kind of strength and
00:12:14
speed to bring that bat around and keep
00:12:16
the ball in in fair uh in such a
00:12:18
circumstance. But here's something that
00:12:20
I just toss up to you guys. I was
00:12:22
thinking like what is it? I was thinking
00:12:24
to look to see um teams that are
00:12:28
explosive in their in their in their
00:12:30
offense. Maybe that's a strike against
00:12:32
them to have to just have this sort of
00:12:34
large variance in awesome in in offense.
00:12:37
So I thought I would rank the teams by
00:12:39
their explosiveness which I counted as a
00:12:41
number of either six or six runs games
00:12:43
or more or seven run games or more. I
00:12:45
tried both. And then I ranked every team
00:12:47
one to 30.
00:12:47
>> Number of times they scored that many.
00:12:50
>> Yes. Right. So it's and and one of the
00:12:52
things that was struck me as
00:12:53
unbelievable
00:12:55
this single counting statistic listed
00:12:58
the top eight teams.
00:13:01
I've never seen that that the top eight
00:13:03
the eight teams that are currently in
00:13:05
the playoffs are the eight teams that
00:13:07
line up with the statistics with one
00:13:09
exception depending on whether you do
00:13:11
six or seven the Arizona Diamondbacks.
00:13:13
Now they have a great part for hitting
00:13:14
they were they bounce around the
00:13:16
Detroit. I mean I'm trying to think like
00:13:17
I mean we're not just selecting for high
00:13:19
scoring team like I I I feel like it's
00:13:21
kind of explosive you'd want is are you
00:13:23
looking at the variance or just like the
00:13:24
me like
00:13:25
>> just a number just something triviously
00:13:26
as simple so I would have expected that
00:13:28
the eight teams would be in the top all
00:13:30
in the top half but they're all in the
00:13:33
row right here they are clear the way
00:13:36
just to follow up on Shane's question
00:13:37
what I think you're describing is is
00:13:40
that let's say the average number of
00:13:42
runs per game in the MLB is four and a
00:13:43
half or something like that
00:13:45
>> and so what you're describing is teams
00:13:47
that seem to have a lot of games above
00:13:50
the average amount of scoring and
00:13:53
therefore by that definition they're
00:13:55
explosive. You don't mean another No,
00:13:57
I'm saying another way you could define
00:13:59
explosive meaning, you know, it's how
00:14:02
much variance is there in their in their
00:14:04
runs per game. What sometimes they're an
00:14:06
11 run team, sometimes they're a
00:14:08
three-run team that averages to seven.
00:14:09
Like I when I when I think of the
00:14:11
Yankees or this kind of like the bad
00:14:13
variance that comes with like the three
00:14:15
outcome approach is like 3-2 outcome
00:14:18
approach. It's like you you you know you
00:14:19
you you hit an average of like three
00:14:21
runs a game but it's like zero and seven
00:14:23
zero and seven zero and seven zero and
00:14:25
seven type of thing. And the zeros
00:14:26
really don't work well in the playoffs
00:14:28
type. So that's so you know
00:14:30
>> in the regular season that's not it
00:14:31
because the Yankees have the fewest
00:14:32
numbers of two runs or more uh or fewer
00:14:35
and they also have the most number of
00:14:36
six or sevens. They really had the best
00:14:38
offense. But I was surprised that by
00:14:40
ranking just on an off simple offensive
00:14:42
statistic, I I thought that maybe the
00:14:45
Padres's would be up there or you know
00:14:47
other teams that were competitive all
00:14:48
season long, the Cubs, they um um they
00:14:52
just teams that were eliminated top
00:14:54
eight.
00:14:54
>> The Cubs were last Cubs were in it. They
00:14:56
were in it, but the the other teams that
00:14:58
were not um the ones that were
00:14:59
eliminated, Cincinnati, Red Sox. Um
00:15:02
>> you're not talking about playoff teams,
00:15:03
though. Yeah. Yeah. Yeah. So they were
00:15:05
>> No. I mean, I thought maybe these guys
00:15:06
would would be Yeah, of course. But
00:15:08
other teams, but eight right right there
00:15:10
in order. It was just this odd oddity
00:15:12
that that that I mean it's probably just
00:15:13
randomness. But I've been looking for
00:15:15
some explanation for what leads you to
00:15:18
the playoffs. And now secondarily, what
00:15:20
allows you to win in the playoffs and I
00:15:22
think those are different phenomenon. Um
00:15:25
getting to the playoffs is almost
00:15:27
certainly unbelievable offense. Winning
00:15:30
in the playoffs, I'm not sure. I think
00:15:31
you have to be more Yeah, I think
00:15:33
playoffs have to be more kind of a situ
00:15:36
You you have to have an adaptive
00:15:37
situational approach and I guess
00:15:40
>> swinging for the fences every pitch is
00:15:42
that that they're kind of trained to do
00:15:43
over the 162 game season. That's I think
00:15:47
what what I guess goes falls apart maybe
00:15:49
for the Yankees specifically or for the
00:15:50
kind of these three true outcome teams
00:15:52
more generally. I mean the Yankees also
00:15:53
have defensive pro ongoing defensive
00:15:55
problems that are
00:15:56
>> absolutely well humorous from my side
00:15:58
but need to be taken care of. Jeez,
00:16:00
guys. Since we're talking about
00:16:01
baseball, we're going to do a 153015
00:16:05
today where we've been talking open
00:16:07
lines here uh for about 15 plus minutes.
00:16:10
Um we do have a guest today.
00:16:11
Unfortunately, he's only audio only, but
00:16:13
for many of our listeners, um they're
00:16:15
doing audio on us anyway, but we're
00:16:18
honored to be joined by Dan Savorski
00:16:20
today. Dan has been on the show before.
00:16:22
Dan is a senior writer at something I
00:16:25
look at quite a bit, which is fan
00:16:27
graphs. He's also developer of the Zips
00:16:30
projection system. Uh Dan, welcome back
00:16:31
to Wharton Moneyball.
00:16:33
>> Hey guys, how's it going?
00:16:35
>> It's going great. Well, not for us
00:16:36
Yankees and Phillies fans, but in
00:16:38
general, it's going great. Um so what
00:16:40
before we get into what's even going on
00:16:42
today in the playoffs. Um could you just
00:16:45
tell us Well, I was going to start with
00:16:47
Zips, but let's forget that for a
00:16:48
second. What do you do at Fan Graphs?
00:16:50
cuz there's so many of our listeners
00:16:52
that, you know, want to know continuous
00:16:54
time win probabilities, uh, projections.
00:16:57
What do you do? What do you specifically
00:16:59
do at Fan Graphs?
00:17:01
>> Well, if if you ask people on social
00:17:03
media, I ruin baseball with math. That's
00:17:05
that that's what I do at Fan Graphs. Uh,
00:17:07
but but seriously, I do I I've developed
00:17:09
the ZIPS projection system, which I've
00:17:11
used uh for the last 20 years. Uh, I did
00:17:14
it while I was at ESPN at Fang Graphs.
00:17:16
That that's obviously a big part of what
00:17:17
I do. and I just generally write
00:17:19
coverage from an analytical standpoint.
00:17:21
I am a statner nerd and I have been
00:17:23
since the 80s. Uh and and that's that's
00:17:26
pretty much what I do.
00:17:28
>> Well, what is why don't you tell us what
00:17:29
is the uh ZIP's projection system?
00:17:31
>> Well, ZIPS is a set of predict
00:17:33
predictive algorithms. Uh on broad
00:17:35
terms, it kind of works a little like
00:17:37
weather forecasting tools in a way. The
00:17:39
idea of it is to get kind of an idea of
00:17:41
where a player is and where they're
00:17:42
going. uh the uncertainty around the
00:17:44
future because the future is very
00:17:46
uncertain as I can tell you from very
00:17:48
very many wrong projections in the past.
00:17:51
Uh if you've ever seen like a hurricane
00:17:53
forecast where they show the little
00:17:54
image of the hurricane on the graphic
00:17:56
and that cone of ignorance that comes
00:17:58
out beyond it, that's kind of the same
00:18:00
thing that I do, just not quite as
00:18:03
sciency as meteorology.
00:18:05
>> So, can you give us a sense like what
00:18:07
are some of the things the ZIPS
00:18:09
projection system predicts? Forget
00:18:11
what's under the hood. forget the actual
00:18:12
algorithm itself. Like if I was to go on
00:18:15
to fan graphs or I don't know, maybe
00:18:17
there's a website just for the ZIP's
00:18:18
projection system, would you be
00:18:20
predicting the number of wins of a team?
00:18:22
Would you be predicting who's going to
00:18:24
win tomorrow's game? Would you be
00:18:25
predicting who's going to win a series?
00:18:26
What are the things that you predict
00:18:28
with the system?
00:18:29
>> Well, generally speaking, the the player
00:18:31
season aligns uh also in season. There's
00:18:33
a a simpler inseason update that can uh
00:18:36
be run every morning uh by the Fanraft
00:18:38
system itself. Uh so what you see if you
00:18:41
go into fan graphs you see the
00:18:42
projection that's kind of the median
00:18:44
line. Uh I do when I do the team by team
00:18:46
rundowns or I write an article uh I get
00:18:49
you know the percentiles that things the
00:18:51
the error the uncertainty uh because
00:18:53
there's quite a lot of that uh uh and
00:18:56
the idea is just to kind of peer through
00:18:57
the fog a bit. Uh I also simulate uh
00:19:00
teams during the season. Uh that
00:19:02
requires a little little more work. I
00:19:04
simulate the season a million times. uh
00:19:06
there's a little bit of linear algebra
00:19:08
involved when I do a generalized model
00:19:11
of injuries to kind of capture that
00:19:12
better than our kind of naive depth
00:19:15
chart projections do. Uh so really if
00:19:18
it's something I can logically model
00:19:20
then I try to do it. Uh and you know
00:19:24
it's it's more work than any sort of
00:19:28
brilliance on my part. It was mostly
00:19:30
effort and time.
00:19:32
>> AI please.
00:19:33
>> All right. Well, let's talk about
00:19:35
projections and and and also
00:19:37
reflections. Um Trent Gisham probably
00:19:40
was the biggest surprise of of the
00:19:42
Yankees year. Um never hit many many
00:19:45
home runs in his history and hit 34 this
00:19:47
year. And uh so what what did your Zips
00:19:50
think and what does it think in the
00:19:52
future for this guy? It's funny, my
00:19:55
editor, one of my editors just wrote an
00:19:57
article about the Yankees and he asked
00:19:58
for the the the the percentile
00:20:02
projection that Gisham's season was
00:20:04
overall and I believe it was the 87th
00:20:07
percentile projection. Zips has always
00:20:09
kind of liked him quite a bit. Uh when
00:20:11
we talk about OPS plus that's kind of
00:20:13
what we were talking about. Uh but when
00:20:15
you look at Zip's projection, uh I can
00:20:18
slowly open it up and run.
00:20:20
>> Wait, hold on a minute. You're saying
00:20:21
that before the season that his this
00:20:24
year's performance would have been in
00:20:25
the 87th percentile. This is a guy
00:20:27
probably never hit more than one home
00:20:28
run every 50 at bats. Well, that was for
00:20:31
OPS plus generally generally speaking,
00:20:33
not for home runs specifically, but I am
00:20:35
opening it and I'll tell you in a minute
00:20:37
>> because I mean I I have to say one of
00:20:39
the calculations that we did on our show
00:20:40
early on, I just did a simple binomial,
00:20:43
you know, uh p value for his rate of
00:20:47
hitting home runs mid-season and it was
00:20:49
in the it was in the upper tail of the
00:20:51
of the distribution, maybe one. And
00:20:52
>> you're saying if you used his rate as
00:20:54
just a P and said, what's the
00:20:56
probability under that with I'll make up
00:20:58
300 at bats. you have 22 home runs or
00:21:01
whatever the number
00:21:02
>> it was at least one in a 100 and now
00:21:03
it's way in the 1,000s. I mean cuz the
00:21:06
guy never hit more than just a a handful
00:21:08
of home runs a season and I hit 34. I
00:21:10
mean wow
00:21:13
sorry go ahead.
00:21:14
>> Oh but one thing to remember of course
00:21:16
is we don't actually know the underlying
00:21:17
probability. This is more of a beta
00:21:19
binomial thing than a pure binomial
00:21:21
thing. We don't actually know the
00:21:22
underlying probability of what it is. We
00:21:24
only know where he's been. We don't
00:21:27
actually know how predictive that is as
00:21:29
a as a recurring probability of that he
00:21:31
hits a home run at any given year.
00:21:33
>> Yeah. So then the right. So then the
00:21:35
question is you know how much weight do
00:21:37
you put on the you know beta
00:21:39
distribution which is you know the prior
00:21:41
which you could use just the raw number
00:21:43
of at bats he has and that as a rate
00:21:45
then you could use the current one. So
00:21:47
the question is
00:21:47
>> well no the the question I was doing is
00:21:49
just looking at a a t test for the
00:21:51
difference right is this is he different
00:21:53
than last year? You're just using P hat.
00:21:57
>> Yeah. Is he a different ball player and
00:21:59
and uh and that's I mean it just he
00:22:02
seems to be too inconsistent from what
00:22:04
he's seen in the past to be to be to
00:22:06
unless it would have to be very use I'm
00:22:09
going to use my acting as host today and
00:22:10
move away from this dreadful Yankee
00:22:12
discussion. Dan, let me just let me just
00:22:15
ask you a general question.
00:22:18
It involves the Yankees sort of, but
00:22:19
like are you surprised at all if you had
00:22:22
used your pre-season projections? Are
00:22:24
you surprised that the four remaining
00:22:26
teams are the Brewers, the Dodgers, the
00:22:30
Mariners, and the Tigers? And if yes,
00:22:32
>> Blue Jays.
00:22:33
>> Sorry.
00:22:35
Blue Jays.
00:22:36
>> Yeah, you just
00:22:37
>> Blue Jays.
00:22:37
>> Blue Jays.
00:22:38
>> You had it right for the Blue Yeah. Blue
00:22:40
Jays over the Blue Jays over the Tigers
00:22:42
>> and the Brewers.
00:22:43
>> It's the Blue Jays against the Mariners.
00:22:45
>> Blue Jays. Mariners. Sorry. Yeah, Blue
00:22:46
Jays. Sorry. The Mariners beat the
00:22:47
Tigers. Blue Jays, Mariners, Brewers,
00:22:50
Dodgers. Would you be surprised that
00:22:52
those were the four teams? And if Well,
00:22:54
certainly there's probably low probably
00:22:56
any sim would have maybe given almost
00:22:58
zero probability that those were the
00:23:01
exact four teams, but what has surprised
00:23:04
you about what we're seeing right now? I
00:23:06
guess if you ask from a projection
00:23:07
standpoint, Zips is moderately surprised
00:23:11
by the exact four teams because like
00:23:14
like most systems and most people, it
00:23:16
picked the Dodgers uh to win the NL
00:23:18
West, though I should know with a lot
00:23:20
less certainty than most people did uh
00:23:22
or projection systems. Zips only
00:23:24
projected the Dodgers to have a 73%
00:23:26
chance of winning the division, which is
00:23:28
probably lower than most people would
00:23:29
have gauged coming into the season. Uh
00:23:31
it projected the Mariners to be second
00:23:33
behind the Astros. uh it projected the
00:23:35
Blue Jays fourth, but there was a lot of
00:23:38
uncertainty in the AL East projections.
00:23:40
There was only uh six wins separating uh
00:23:43
the teams in in the AL East. Uh when you
00:23:46
when you talk about the win projections,
00:23:47
it was really really uncertain about
00:23:49
what was going on. So, it did think the
00:23:51
Blue Jays would bounce back, but it had
00:23:54
him at a 40% chance of making the
00:23:55
playoffs. It wasn't going to be the f
00:23:58
they weren't going to be the favorite
00:23:59
team to make the ALCS.
00:24:00
>> So, Dan, is it fair to say the
00:24:02
following? Um, since we're a stats show
00:24:04
here on Wharton Moneyball here on the
00:24:05
Wharton Podcast Network, um, since we
00:24:07
are a stats show, I can say you're not
00:24:10
saying that the top these four teams,
00:24:13
the Brewers, the Dodgers, the Blue Jays,
00:24:16
and the Mariners, you're not saying that
00:24:18
it has a high probability of that four
00:24:21
tuple. You're just saying it's not that
00:24:23
surprising given the width and the
00:24:25
uncertainty of the distribution. Yeah, I
00:24:28
think when you look at it, no one team
00:24:31
being where they are is surprising. Uh
00:24:34
it's just that these weren't necessarily
00:24:35
the most likely teams. If you asked Zips
00:24:38
what the most likely teams uh would have
00:24:40
been, it would have been the Houston
00:24:41
Astros. Now, I would have been surprised
00:24:43
because the Baltimore Orioles did not
00:24:45
come anywhere close. Uh that is one of
00:24:47
the larger projection misses from
00:24:49
preseason. Uh the other one, of course,
00:24:51
is the Atlanta Braves. Zips would have
00:24:53
guessed that the most likely NLCS would
00:24:56
have been Braves versus Dodgers or
00:24:58
Phillies versus Dodgers. Uh, so
00:25:02
nothing crazy happened, but a lot of
00:25:06
these events were at least moderately
00:25:08
unlikely.
00:25:08
>> Let me ask you a question. When you
00:25:10
think about, you know, we all try to
00:25:11
measure success of our forecasts. um
00:25:14
when you tend to think about the whether
00:25:16
it's the you know the zips forecasting
00:25:18
system or more generally what's done at
00:25:19
fan graphs like you could say well you
00:25:22
know uh we got the Braves wrong or we
00:25:25
got the you know as the Braves didn't
00:25:27
even make the playoffs so like do you
00:25:30
think about the mag like as you're
00:25:32
thinking about developing your system do
00:25:34
you literally think 01 loss like you
00:25:36
know wrong's wrong or do you think wow
00:25:38
the Orioles wrong is really wrong
00:25:40
>> yeah and actually just to follow up on
00:25:42
that because I was thinking of very
00:25:43
similar thing to Eric I think like are
00:25:46
there particular teams or like sort of
00:25:47
circumstances like I mean obviously the
00:25:49
system's going to miss every year so
00:25:51
it's got to miss something but like you
00:25:53
know are there kind of either teams or
00:25:54
types of teams or like is it errorprone
00:25:57
in some way that you guys haven't been
00:25:58
able to kind of get at yet?
00:26:00
>> We the the the problems tend to be less
00:26:03
bias and more accuracy. Uh accuracy of
00:26:06
course is the hardest thing to pin down.
00:26:08
Once you've had so much experience
00:26:10
projecting things, you kind of can get a
00:26:13
pretty good idea of if there's certain
00:26:16
aspects that you're that teams are
00:26:18
systemically being overrated or
00:26:20
underrated. For example, if if teams
00:26:23
with a high percentage of fast players
00:26:26
is being uh uh consistently underrated,
00:26:28
then you kind of have an idea and from
00:26:30
the data uh however you choose to do
00:26:32
that any kind of, you know,
00:26:33
dimensionality reduction of any kind of
00:26:35
model. Uh so there isn't really a most
00:26:39
of these easier lowhanging fruit things
00:26:41
have been picked which is why you see
00:26:43
most good projection systems tend to
00:26:46
cluster around each other pretty well
00:26:48
with accuracy uh because the easy stuff
00:26:51
is done and now we're kind of in very
00:26:54
very small difference territory.
00:26:55
>> I have a question about but Shane I saw
00:26:57
your hand up please. Well, I guess just
00:26:58
to kind of follow up on that cuz I mean
00:27:00
I I totally I get what you're saying.
00:27:02
Like I I guess what's burned my question
00:27:04
is I I I wouldn't mind your thoughts
00:27:05
from somebody who like looks at the data
00:27:07
quite a bit. How are the Brewers doing
00:27:09
what they're doing as like, you know,
00:27:10
kind of a bottom payroll team,
00:27:14
you know, is like do is there something
00:27:16
kind of like that that is it how are
00:27:19
they doing what they're doing? I it's
00:27:22
funny that one of the things I because
00:27:24
people ask about this with the Brewers
00:27:25
because they're actually the team that's
00:27:26
had the least accurate projections over
00:27:28
the last five years uh is that if you
00:27:31
look at the team if you look at the
00:27:33
player projections they actually aren't
00:27:35
systemically missing uh low on the
00:27:38
Brewers but the team projections are.
00:27:41
Uh, and what I deduced, uh, is that the
00:27:44
biggest problem with projecting team
00:27:46
standings with the Brewers, at least for
00:27:48
Zips, I can't speak for any other
00:27:49
projection system, which has had the
00:27:51
same problem, is that I am doing a poor
00:27:54
job projecting who the Brewers will
00:27:56
actually use. Uh, for example, Zips gave
00:27:59
a very positive projection for Joey
00:28:01
Ortiz in his rookie year, but I wasn't
00:28:04
convinced that they were going to go
00:28:05
allin with him last year. Isaac Collins
00:28:07
had a decent projection this year and I
00:28:09
was very conservative about how many how
00:28:11
many play appearances I gave him in the
00:28:13
majors. It seems that I'm underrating
00:28:15
the Braves more than the projections
00:28:17
are. And that's kind of a problem when
00:28:18
you're doing something like team
00:28:20
standings because
00:28:23
a a model such as this is not going to
00:28:25
be able to really do a great job
00:28:27
figuring out who will play.
00:28:30
>> Yeah. I mean, you could have known
00:28:31
Andrew Vaughn would go there and do
00:28:32
whatever he's been doing as well.
00:28:34
>> Yeah. I mean, perhaps someone can more
00:28:36
accurately
00:28:38
model decisions that general managers
00:28:40
and coaches are making with with setting
00:28:42
uh the lineups, but that's not something
00:28:44
that's in my skill set. At least I have
00:28:46
figured out how to cross those problems
00:28:48
with any meaningful way. But Dan, let me
00:28:50
ask you a question that I'm channeling
00:28:52
my inner Shane Jensen here, even though
00:28:54
Shane is here.
00:28:55
>> So, this is, by the way, the odds I'm
00:28:57
about to tell you are before last
00:28:59
night's game. I before last night's
00:29:01
game. So, my understanding is the if I
00:29:03
got this right, I even got the team
00:29:05
draw. Seattle beat Toronto last night,
00:29:07
right? Yeah. Yes. Yeah, they beat him.
00:29:09
Okay. But this was before last night's
00:29:11
game.
00:29:12
How can you explain any team
00:29:16
Shane can jump in after you? The Dodgers
00:29:19
were plus 120. Now, let's all be clear.
00:29:22
There's four teams left. That
00:29:24
essentially gives them double the
00:29:26
probability of the average team left.
00:29:30
How would you explain that? Like does
00:29:32
that Dan maybe that seems normal to you?
00:29:34
And the Brewers by the way are plus 850
00:29:36
which gives them like you know 1/8 of
00:29:38
probability or 1 nth of probability
00:29:40
which is you know a quarter what they
00:29:42
should have roughly or you know how
00:29:44
would you explain that and do you think
00:29:46
there's enough precision or is that just
00:29:48
way too out of line? Well, of course,
00:29:50
one of the fundamental things is that
00:29:52
book makers are also seeking to maximize
00:29:54
how much money they earn. And there's
00:29:56
obviously going to be a great deal of
00:29:57
correlation between the one true
00:29:59
underlying probability, the wisdom of
00:30:01
crowds and and where they set these
00:30:04
things. Uh, but all I could do, I mean,
00:30:06
is is my work. Zips at least is less
00:30:10
sanguin about the Dodgers.
00:30:12
>> But what does it have the Dodgers win
00:30:13
probability at right now? Uh right now
00:30:15
it has the Dodgers uh to just to just to
00:30:19
get past this round. It only has the
00:30:21
Dodgers as 5446
00:30:24
>> uh over that's very different. Pretend
00:30:27
like Shane's not here. Shane would agree
00:30:29
with that.
00:30:32
>> So they can't have and then the Winter
00:30:34
World Series. I don't know. Let's even
00:30:35
give him 54 again.
00:30:36
>> Yeah. Actually, Dan, since since you're
00:30:39
talking about that 5446 and you've been
00:30:41
doing this obviously for many seasons,
00:30:42
is that kind of is that even unusual?
00:30:45
Because, you know, you would I think
00:30:46
most people would kind of naively look
00:30:48
at the Dodgers Brewers as kind of a mis,
00:30:50
you know, just based on persona like a
00:30:52
real kind of mismatch. 5446. Is that
00:30:55
about as much as you stretch away from
00:30:57
5050 for like once you get to something
00:30:59
like the conference series or like
00:31:02
historically after you've done this, is
00:31:03
that like a big spread? I'm working from
00:31:06
memory here simply because I don't have
00:31:07
one easy to access spreadsheet but I
00:31:10
don't believe that any of the
00:31:11
championship series have been
00:31:13
>> at least 6040 or or or if they are it's
00:31:17
been very close like maybe 6139 I
00:31:20
generally speaking I mean teams in
00:31:22
baseball are very very close it's not
00:31:24
like the NBA uh I can't think of the
00:31:27
paper right now who the three analysts
00:31:29
who
00:31:31
analyzed best of seven series and the
00:31:34
NBA and estimated that for MLB to have
00:31:39
the same record of better team advancing
00:31:41
as the NBA did that they would have to
00:31:44
play something like best of 76 series.
00:31:47
>> I think that was Michael Lopez who did
00:31:48
that.
00:31:49
>> Yes, that's it. I couldn't remember.
00:31:50
>> He he basically said I mean he's listen
00:31:52
because there's so much more spread in
00:31:53
the NBA you have uh you don't need that
00:31:56
many games to resolve who's better but
00:31:58
because the the differences in quality
00:32:00
in in any given game in MLB are much
00:32:02
more compact. you need way way more more
00:32:05
games to to
00:32:06
>> or not even any I mean certainly the
00:32:09
spread I mean in any given game I don't
00:32:11
know you'd probably go above 6040 but by
00:32:13
the time you get to the conference
00:32:14
series
00:32:15
>> selecting for teams where I doubt I mean
00:32:17
I'm I'm not surprised to hear 6040 is
00:32:19
kind of the limit of what you kind of
00:32:21
differ from the coin flip of course I
00:32:22
wouldn't go beyond 40 55 45
00:32:26
>> I'm going to play the game with you Dan
00:32:27
that I always do with Audi and Shane so
00:32:30
Dan let me do it with Shane and Audi and
00:32:32
and you just listen on and then I'll get
00:32:33
your comments. So guys, we just heard
00:32:36
5446. I'm going to give you another
00:32:38
piece of information and you tell me how
00:32:40
this would move things if at all. The
00:32:42
Brewers were the best team in baseball.
00:32:46
They were. And they had a better wreck
00:32:48
than the Dodgers. That move your
00:32:49
probabilities at all? No. All right. How
00:32:52
about this?
00:32:52
>> Over over an even schedule?
00:32:55
>> No.
00:32:55
>> Okay. How about the following?
00:32:57
The Brewers were 6 and0 this year
00:32:59
against the Dodgers.
00:33:02
No.
00:33:05
Audi, is that worth anything to you?
00:33:08
>> Yeah, I think that's worth a little bit.
00:33:10
Tiny bit, but a little bit. I mean, the
00:33:12
real question, Shane, is it nothing or
00:33:14
is it is it I think it's worth a
00:33:16
percentage point.
00:33:16
>> Yeah, I guess. You know, again, when I
00:33:18
think about Dan, I want to hear actually
00:33:21
I want to hear what Dan has to say
00:33:22
first.
00:33:22
>> Dan, what's your thought? Is that worth
00:33:24
any like in the I'm not saying it
00:33:25
should. You heard Audi and Shane say if
00:33:27
it has something, it's tiny. Is there
00:33:29
somewhere in the zip system that has the
00:33:32
fact that the Brewers are 6 and0 this
00:33:34
year against the Dodgers?
00:33:36
>> Well, directly no. But that's not
00:33:38
necessarily
00:33:40
right. That's just something that it's
00:33:42
probably too small for me to model in a
00:33:44
meaningful sense because Zips is trying
00:33:45
to get the underlying reasons why a team
00:33:48
might be better. But I can't say it
00:33:50
doesn't really move it at all because
00:33:51
Thomas Baes will will come back and come
00:33:54
out of the grave and and punch me in the
00:33:56
face.
00:33:57
>> Right. Of course, it has to move it a
00:33:59
little some way. The question is by how
00:34:00
much. I mean, listen, I to Toronto was a
00:34:03
beast against the Yankees all season and
00:34:05
they beat them in the playoffs. And I I
00:34:08
mean, sometimes that the teams are I
00:34:09
mean, we we have a hard time putting our
00:34:11
finger on exactly what it is, but there
00:34:14
are different there are archetypes of
00:34:15
players and pitchers and game styles and
00:34:18
>> yeah, I guess
00:34:19
>> it must matter a little bit and we have
00:34:21
some revealed information.
00:34:22
>> If I told you if I told you again, I'm
00:34:25
let me frame it a different way.
00:34:28
If I told you that team A was 6 and0
00:34:30
against team B, I'm not going to tell
00:34:32
you who the teams are. Team A had a
00:34:35
better record in the regular season than
00:34:36
team B.
00:34:38
Team A and team B are part of the last
00:34:40
four teams remaining.
00:34:42
Team A and team B are playing each
00:34:44
other.
00:34:45
And team A is plus 850 and team B is
00:34:51
plus 120. You'd be like, there's some to
00:34:54
do here. I mean, somebody would say
00:34:57
something about something seems
00:35:00
miscalibrated. Dan, let me give you the
00:35:01
first comment on this on my simplistic
00:35:04
analysis. I'm stripping off the names of
00:35:06
the teams. I'm just giving you the data
00:35:08
from this season. It must be because,
00:35:10
and told us this before, if you
00:35:13
correlate payroll with winning, it's
00:35:16
very correlated. If you if you look at
00:35:19
priors, obviously the Dodgers had a much
00:35:21
stronger prior than the Brewers. Is
00:35:23
there some other explanation I'm
00:35:24
missing?
00:35:27
>> Not necessarily. At least one that I
00:35:29
think that a model can actually
00:35:30
meaningfully address. Uh because these
00:35:33
generalized models are are very useful,
00:35:36
but again, we're we're talking about all
00:35:38
models are wrong, but some models are
00:35:39
useful. Uh and there might be something
00:35:43
that let's say there's a particular
00:35:45
style issue that causes the Brewers to
00:35:47
have a better record against the
00:35:48
Dodgers. Even if we on some level can
00:35:51
feel that, I'm not sure we can
00:35:53
necessarily isolate it because it's it's
00:35:55
hard to isolate these reasons for
00:35:57
things, it's what I always tell people
00:36:00
coming into the season about a
00:36:01
projection is that if a player hits 300
00:36:04
or I project a player to hit 300 and
00:36:06
they actually hit 300, I still don't
00:36:07
actually know if I was right. I still
00:36:09
don't really know if they were truly
00:36:10
that was the underlying probability in
00:36:12
any given atbat was 30%. or if they were
00:36:15
actually truly a 27% hit guy who got
00:36:18
lucky or a 33% guy who who who showed
00:36:21
real decline.
00:36:23
So, you never really know the real
00:36:25
answer in a way. I mean, there's lots of
00:36:26
tools to to, you know, evaluate our
00:36:28
conditional probabilities, but I don't
00:36:31
necessarily think that
00:36:33
it's it's in the gut. I hate to use the
00:36:35
word gut, but there's just some little
00:36:39
nudge here and there that models
00:36:40
necessarily can't really pick up super
00:36:42
well. Let me let me ask you one last
00:36:44
question then we'll uh kind of let you
00:36:46
go here. Um what do you see happening?
00:36:48
There's a lot that could be going on in
00:36:50
the off season, but like is there any
00:36:52
player when you look at their prior on
00:36:55
what fan graphs or the zip system might
00:36:57
project for them and now you look at
00:36:59
their posterior cuz the 2025 season,
00:37:01
let's forget the playoffs just for a
00:37:03
second. The 2025 season is over. What
00:37:06
players if any have moved the most
00:37:09
between your prior and posterior? like
00:37:11
who like and possibly who may cash in on
00:37:14
that in free agency. They could be those
00:37:16
could be independent questions, but I'm
00:37:18
just wondering who have you updated the
00:37:20
most. Peter Lonzo has actually improved
00:37:23
quite a bit when you look at the he's
00:37:24
kind of arrested a a gradual decline. Uh
00:37:28
but it's interesting in a way because I
00:37:30
still don't really have a real good
00:37:32
feel. I mean, I've talked to agents on
00:37:34
this issue of just how much his market
00:37:36
has recovered. Uh because last year Zips
00:37:39
was actually pretty close on the
00:37:40
contract he eventually got with the Mets
00:37:42
and I was really really surprised by
00:37:44
that. I thought, oh, Zips is going to
00:37:46
miss by like $70 million on him. Uh and
00:37:49
it didn't. And I was a little confused
00:37:51
by that because there's certain things
00:37:53
you expect to be wrong about when
00:37:55
there's the vibes. Uh and I think he's
00:37:58
going to get a bigger contract than he
00:37:59
did last year. Also, there's there's no
00:38:01
free agent compensation attached. But
00:38:04
when he's talking a seven-year deal as
00:38:06
he has initially, I don't know if it's
00:38:09
recovered that much. So, I'm really
00:38:11
fascinated uh to to see what actually
00:38:13
happens with his contract this off
00:38:15
season. He'll do better, but I would be
00:38:17
really surprised if he got seven years,
00:38:19
but I've been surprised before just
00:38:22
maybe actually while we have you might
00:38:24
as well get Do you have any, you know,
00:38:26
stated projection, maybe you've made it
00:38:28
already on how you see the uh playoffs
00:38:31
going?
00:38:32
>> Uh yes. Uh, Zips actually likes the
00:38:34
Mariners before before the game.
00:38:36
Obviously, it likes it them better now.
00:38:39
Uh, right now Zips has them at SE as
00:38:42
Mariners at 7228 over the Blue Jays. Uh,
00:38:45
and then the Brew the Dodgers 5446 over
00:38:49
the Brewers. Uh, it has depending on
00:38:52
what the matchup is. Uh, if we just look
00:38:54
at Blue Jays Dodgers, uh, Dodgers go up
00:38:58
5545. Uh, Zips actually likes the
00:39:00
Brewers a little more than the Blue
00:39:02
Jays. Uh, but you know, there's a lot
00:39:04
that could go wrong for a team in that
00:39:06
45%.
00:39:08
>> Well, Dan, we'd like to thank you for
00:39:10
joining us here on Morton Moneyball.
00:39:11
We've been talking to Dan Sorski. Dan is
00:39:14
a senior writer at Fan Graphs, the
00:39:15
developer of the ZIPS projection system.
00:39:18
Dan, as always, thanks for your insights
00:39:20
and thank you for joining us here on
00:39:21
Wharton Moneyball.
00:39:22
>> Always fun. Thanks for having me.
00:39:25
>> Welcome back. Welcome back to Wharton
00:39:27
Moneyball, the Wharton podcast network
00:39:28
edition. Today's been an interesting
00:39:31
show in the way we've taped it. We had
00:39:32
15 minutes of open segments of which
00:39:35
Shane was going on appropriately about
00:39:38
potential of Drake May and the New
00:39:40
England Patriots. Ai was lamenting as he
00:39:43
has a right to about the Yankees and the
00:39:46
high volatility and a little bit about
00:39:48
Trent Gisham. And then of course for the
00:39:49
last 25 or 30 minutes or so we've had
00:39:51
Dan Sorski on. Dan is a senior writer at
00:39:54
Fan Graphs, the uh developer of the ZIP
00:39:57
system. So now we're going to go back
00:39:58
for the last 15 or 20 minutes to the
00:40:00
open line segment here on Wharton
00:40:02
Moneyball. Um I'll go next. So um I'm
00:40:07
going to go to tennis because everybody
00:40:09
knows I love tennis and um
00:40:13
this was crazy what happened in tennis.
00:40:17
So, just to remind all of our listeners
00:40:19
here on Morton Moneyball, there are
00:40:21
multiple levels of tournaments in
00:40:24
professional te tennis. Everybody of
00:40:26
course knows the majors, right? There's
00:40:28
Wimbledon, US Open, French Open,
00:40:30
Australian Open. You might as well call
00:40:32
those 2,000 level events. The reason I'm
00:40:35
call I'm not I'm using the number 2,000
00:40:37
is if you win them, you get 2,000
00:40:39
points.
00:40:41
Then there's the one that's just below
00:40:43
that of which there's about eight or
00:40:44
nine in a year and those are called
00:40:46
masters 10,000 events. Notice it's half
00:40:48
but a thousand points is a lot in
00:40:51
tennis. Then there's what's called the
00:40:53
ATP 500 events. Then there's even 250
00:40:56
events. Notice I'm having there are 125
00:40:59
events but the major players don't
00:41:01
really play in them. Those are the major
00:41:03
level events. So, there are players that
00:41:05
you guys know, many players that you
00:41:08
guys know that have never won a Masters
00:41:11
1000 event. It's actually quite hard,
00:41:14
especially now, like for 20 years, you
00:41:16
are going to have to beat at least one
00:41:18
if not more of the big three because
00:41:21
they're all and I'll include Andy Murray
00:41:23
and Stan Roinka and other guys because
00:41:25
they're playing all playing the Masters
00:41:27
1000 events. And now, of course, you
00:41:29
have Alcarz and S. And so, that's not
00:41:32
going to be easy. and Jookovic is still
00:41:33
playing. So, you're going to have to
00:41:34
beat one of those guys. Well, I don't
00:41:37
know, AI, if you saw this.
00:41:39
I did not. Okay. What would be I know
00:41:42
Shane has. What would be a surprising
00:41:45
ranking? Like, how low a ranking do you
00:41:49
think you could get for someone to win a
00:41:51
mass 10,000? Let me just give you a
00:41:52
piece of information.
00:41:56
Alcarz chose not to play in this one, so
00:41:58
that helps. He did not play in it. He
00:42:00
had won just in China. and he won a 500
00:42:02
event in China. Tweaked his ankle a
00:42:04
little bit, decide not to play in this
00:42:06
Masters 1000.
00:42:08
Dinner played but had to retire in the
00:42:10
second round injured. Bookovich played
00:42:13
in the event, but like what ranking do
00:42:17
you think someone who could win the
00:42:19
event might be? 20, 30, 50. What do you
00:42:22
think?
00:42:22
>> All right. So, we actually discussed
00:42:24
this um um not too long ago with one of
00:42:26
our tennis guests. Okay. And my
00:42:28
hypothesis was that um there's still a
00:42:32
lot of dominance um outside of the top
00:42:34
three um relative to the bottom of the a
00:42:38
half of the top 20. But the our experts
00:42:41
seem to think that that wasn't the case
00:42:43
that you really get um tremendous
00:42:45
dominance at the upper top and then it
00:42:47
and then it and then it's it's it levels
00:42:50
out.
00:42:51
>> So I think probably 20 in a masters
00:42:54
event I would say. Yeah, the number is
00:42:57
204.
00:42:59
Oh my god. The guy that just won,
00:43:01
Valentine Vashro,
00:43:03
was ranked number 204 in the world. He
00:43:07
beat Holgaruna, who's number six. You
00:43:10
may have heard of him. He beat
00:43:12
NovakJokovic. You may have heard Yeah,
00:43:15
you may have heard of him. And then,
00:43:17
interestingly, he played his first
00:43:19
cousin who's ranked about 50 in the
00:43:22
world, this guy Render Kanesh. They're
00:43:26
literally first cousins. They're sisters
00:43:27
or mothers. They're they played each
00:43:29
other in the finals. Render Kesh had
00:43:31
beaten five seated players to get to the
00:43:33
finals, but a person ranked 204
00:43:40
won a Mast's 1000 event. Now, let me
00:43:42
just say he's now ranked number 40 in
00:43:45
the world, just to show you how much a
00:43:46
thousand will get you. And of course, I
00:43:49
say this is life-changing, not just
00:43:51
because of the $1.2 million he gets.
00:43:54
That's it's twice the amount he had won
00:43:56
in his entire career. The guy's 26. He's
00:43:58
not 18. It's 26. It's twice the amount
00:44:01
he had won. But he now gets into every
00:44:03
event. So, he gets into every major. He
00:44:08
gets into every mast's 1000. He gets
00:44:10
into any 500 event he wants. He doesn't
00:44:12
have to go through qualifying, which by
00:44:13
the way, he wasn't even in qualifying
00:44:16
for this. Just let me remind you guys
00:44:18
for a 128 person event like this they
00:44:22
let usually like 110 people in based on
00:44:24
rankings and the last 18 have to play
00:44:26
their way in. He was an alternate. He
00:44:30
wasn't even in the list of people that
00:44:33
could play their way in if somebody
00:44:35
dropped out. They just had some local
00:44:37
guy who was kind of around.
00:44:39
>> This would be like a Would this be like
00:44:40
an amateur winning the US Open?
00:44:44
like like the analogy like yeah in yeah
00:44:47
in golf I'm trying to kind of give an
00:44:48
analogy to like you know because that's
00:44:49
something at least we've seen amateurs
00:44:51
come on and
00:44:53
>> do well you know like you know I kind of
00:44:55
feel like
00:44:57
>> I you know I don't know if that's the
00:44:58
right analog or if it's even more rare
00:45:00
what just happened is even more rare
00:45:01
than
00:45:02
>> you should know Audi no masters events
00:45:04
been around for about 30 40 years this
00:45:05
is the lowest rank to ever win
00:45:07
>> right so I I have two remarks first of
00:45:08
all golf has more variance way more
00:45:10
variance than tennis there's I mean you
00:45:13
have the putting
00:45:14
randomness which is unbelievable uh and
00:45:16
and can make a huge difference. So um I
00:45:19
would have imagined that amateurs have
00:45:21
won made professional tournaments in
00:45:22
golf every rarely but every now and last
00:45:25
amate there
00:45:26
>> this is the US Open though.
00:45:27
>> Yeah there have been amateurs that have
00:45:29
won professional golf events but I don't
00:45:32
think if an amateur's ever well Bobby
00:45:34
Jones in 1920s won the major
00:45:36
>> No. Yeah. I mean I more recently I feel
00:45:37
I feel like every every few years
00:45:39
there's an amateur that like I mean of
00:45:41
course the amateurs aren't going to be
00:45:42
playing in the Masters or whatever. It's
00:45:44
kind of you know it's the conditions for
00:45:46
this to happen are
00:45:47
>> the person that wins the person that
00:45:48
wins there is an amateur championship
00:45:51
>> that person automatically gets to play
00:45:53
the Masters. So actually there are and
00:45:55
there's like a couple others like I'll
00:45:57
make it up the South American amateur
00:45:58
champion. There are three or four people
00:46:00
that amateurs that do get to play the
00:46:02
Masters every year. So the real question
00:46:05
about about this is what as a post let's
00:46:08
be bas what's my posterior I would
00:46:11
imagine that something we misranked him
00:46:13
in some way.
00:46:14
>> Yeah. Well that's that's I was going to
00:46:15
ask like what was his trajectory before
00:46:17
he got to 204. I don't know how far the
00:46:19
but if this guy was just on crazy ascent
00:46:21
or something.
00:46:22
>> So he'd been playing for years and yet
00:46:24
he's still I mean this doesn't make
00:46:26
sense. Tennis is a game with lots of
00:46:28
ladders of of levels of of quality. 204
00:46:33
is nothing like a 50 usually nothing
00:46:38
like and 50 isn't like 10. So that's a
00:46:42
lot of ladders to jump. And so
00:46:44
>> I tell you I'm reading it right. This is
00:46:46
look it's we I know you guys have some
00:46:48
skepticism of AI which you should but
00:46:50
let me just here's some data from AI. So
00:46:53
um he's increased steadily in January of
00:46:56
2022 he was in the top 500 in the world.
00:47:00
Then in January of 2024, he was in the
00:47:02
top 200 in the world. And then by May of
00:47:05
2024, he had reached 116. That was the
00:47:08
highest he had ever gotten. Then he
00:47:09
slipped back down to 204 because of
00:47:12
injuries.
00:47:14
So that's it. That's his That's his
00:47:17
entire
00:47:18
All right. Well, at least he had been
00:47:20
one 116 and he had a pretty strong
00:47:22
improving trajectory which was set back
00:47:24
by injuries. So you
00:47:26
>> Yeah. Maybe maybe he just had like a
00:47:28
long-term kind of crime like some kind
00:47:29
of injury like we're looking at him post
00:47:32
some like correction that really made a
00:47:34
huge difference.
00:47:34
>> I know you're doing something that a
00:47:36
good basian would do which is that look
00:47:38
we can't refute the data. He won. Now
00:47:41
the question is what's an explanation
00:47:43
for it? You've given a couple. One is
00:47:45
the distribution of talent is really
00:47:47
really flat. We don't really believe
00:47:48
that but that is one possibility right?
00:47:51
Another possibility is his true ranking
00:47:54
is not 204 and that you know had he not
00:47:57
been injured for part of 2025 maybe he
00:48:00
would have been 50 or 60 in the row. Now
00:48:01
that still would have been unlikely but
00:48:03
not the four sigma unlikely or more we
00:48:06
might be talking about here and this is
00:48:08
what scientists do. You you you you come
00:48:10
up with alternative hypotheses some of
00:48:13
which are more plausible than another.
00:48:15
And this one seems quite plausible. Can
00:48:17
I can I ask a sample size question in
00:48:19
tennis, Eric, cuz you know these ranking
00:48:21
systems better. Like how many now
00:48:23
tournaments in a row of like first place
00:48:26
like first first round exits would he
00:48:28
have to pile up? You said he's now at 40
00:48:30
or somebody said he's like at 40 in the
00:48:32
world now
00:48:33
>> to get to to knock him back down to 200.
00:48:35
How many bad performances does he like
00:48:38
how how much was this performance worth
00:48:39
and kind of like and sort of in the
00:48:41
almost like I guess the inertia of the
00:48:43
ranking system?
00:48:44
>> Well, I'll tell you I'll give you an
00:48:45
example right now. That's a great great
00:48:47
question. Here's what I can tell you
00:48:48
with certainty. He cannot go below for
00:48:52
the next year. Let's assume that the
00:48:54
number of points are fairly stable
00:48:57
across the different rankings, right?
00:49:00
So, um he pretty much can't go below
00:49:04
about 50 in the world or 55 because
00:49:08
that,020 points he has stays on for a
00:49:11
year. So even if he gets zero zero zero
00:49:15
zero zero zero and remember he's going
00:49:18
to have he's not going to play 51 weeks
00:49:20
but he's going to have that many weeks
00:49:23
of opportunity to get I mean cuz he's in
00:49:26
every tournament he wants to play. So it
00:49:29
would take a long time now for him to
00:49:32
drop cuz that it just happened and it
00:49:34
stays on for a year. He's gonna he's
00:49:37
he's changed. That's what I meant when I
00:49:38
said he's changed his career. Like even
00:49:40
if he has a bad year, he's probably
00:49:43
going to stay in the top 100, which will
00:49:45
also get him into tournaments for
00:49:46
another year. And I hate to say it, the
00:49:48
guy's 26, he ain't 18. You know, maybe
00:49:51
he's not. Yeah. Everyone says, "Well,
00:49:53
Jookovic plays till he's 38." Yeah,
00:49:55
right. Okay. Well, Pete Sampas played
00:49:56
till he's 31. Most players are done by
00:49:58
27, 28. This guy has changed the last
00:50:01
three or four years of his career, for
00:50:05
sure.
00:50:06
And that's amazing.
00:50:08
Please.
00:50:09
>> Uh yeah. Well, I was actually one of the
00:50:11
things we actually kind of touched on it
00:50:12
a little bit. Um that the tennis ranking
00:50:15
systems in particular have a lot of uh
00:50:18
sort of built-in randomness that are due
00:50:21
to, you know, decisions to play certain
00:50:23
tournaments and not play them. They're
00:50:25
not really what we would call power
00:50:27
rankings. Um which
00:50:29
>> they have both. They have they have the
00:50:31
live rankings which tend to have a
00:50:32
smaller window to them and then they
00:50:35
have the one that I just described to
00:50:36
Shane where the points stay on for a
00:50:38
certain period of time.
00:50:40
>> Right. But I'm I'm curious to know what
00:50:42
like what
00:50:43
>> like what an ELO system would do.
00:50:44
>> Yeah, an ELO system or or or just
00:50:47
someone's power ranking a regression
00:50:49
based uh forecast that says this is how
00:50:51
good we really think you are. And I
00:50:53
would I would imagine that his 204 was
00:50:55
quite probably quite different from what
00:50:57
his power ranking or ELO based or
00:50:59
whatever system
00:51:00
>> like what residual would that have in
00:51:01
like an ELO or something like that.
00:51:03
Yeah. No, that's a great question.
00:51:04
>> And this is actually interesting because
00:51:05
this is one of the problems with these
00:51:06
systems is that um it it sometimes takes
00:51:10
into account lack of play um or a new
00:51:14
player when he comes online, we don't
00:51:15
know what to do with him, so we give
00:51:16
them often a very low score. Well, here
00:51:18
I can tell you right now again this is
00:51:21
not I I can't see the number based
00:51:24
before this tournament but I just went
00:51:26
to the ATP rankings which is an ELO
00:51:28
based system tennis abstract.com. I've
00:51:31
talked about them before. Um it has
00:51:33
Vashiro this guy at number 43.
00:51:37
Now that's not very far from his other
00:51:39
ranking but how far that move Oh, let me
00:51:42
click on his name. Let me see if it has
00:51:43
a time series and like like let's see
00:51:45
how far it probably does. Hold on a
00:51:47
second here.
00:51:51
Well, it I if I'm reading this right, I
00:51:54
may be reading this wrong. Fans here on
00:51:56
Morton Moneyball. No, it had him at 204.
00:52:00
It had him like at 200. I assume. Oh, I
00:52:02
don't know what. Let me I I don't know
00:52:04
what V rank is. So, maybe this is
00:52:06
different. Maybe his rank was different
00:52:09
before this. Oh, I see. So, it's
00:52:12
actually changing his ranking. Oh, this
00:52:14
might be his opponent's rank.
00:52:17
I think these are his opponent's ranks.
00:52:19
Yeah, no, no, sorry. That's just who he
00:52:20
beat. Well, let me just tell you
00:52:22
according to the ELO system in the round
00:52:24
of 64, right? He beat the 82nd player in
00:52:27
the world, then 17, then 23, then 31,
00:52:32
then 11, then five, and then 54.
00:52:37
>> That's not bad. Like what what again?
00:52:40
Player A, player B. If one was at at 20
00:52:42
and one was at 200, what would you say
00:52:44
the odds were for a single match?
00:52:48
>> 95% at least.
00:52:50
>> First thing that came into my mind was
00:52:51
99%. I would say 2% chance that the
00:52:54
200th player in the world beats crazy.
00:52:57
>> There's no way he was a he was a go into
00:52:59
that
00:53:00
>> tour. But we we we've we've definitely
00:53:03
the null hypothesis that he's a right
00:53:05
>> Well, that's what I was about to say,
00:53:06
Shane. All for our listeners out there,
00:53:08
Audi's computing the P value under the
00:53:10
null. Audi does not believe, as he said
00:53:12
many times in 11 and a half years, a one
00:53:14
in a million events happen. And so
00:53:16
there's no way the 200th ranked player
00:53:19
beats five, six players in the top 20.
00:53:22
Not no way, but like with very very
00:53:25
small odds. Therefore, you have to then
00:53:27
say maybe the null of him being 200 is
00:53:30
just not true.
00:53:32
>> Yeah. I mean the question is is it
00:53:34
should you have revised it in the middle
00:53:36
of the tournament right I would have
00:53:39
>> well that's well that's another thing
00:53:41
>> certainly advance after that first round
00:53:42
or something like that also once he beat
00:53:45
Djokovic at that point you know I don't
00:53:47
know I mean I understand Jookovic has
00:53:48
good days bad days but even the bad
00:53:50
Jookovic is still pretty darn good and
00:53:52
so
00:53:52
>> you know so Eric let me ask you a
00:53:54
question because you go to a lot of
00:53:55
tournaments and watch a lot of tennis um
00:53:59
are you able to see with your eyes when
00:54:03
you're watching two players that one is
00:54:06
substantially better skilled than the
00:54:08
other as opposed to just looking at the
00:54:10
score.
00:54:10
>> Yeah, it's a great question. Let me tell
00:54:13
you what you can tell. You can tell a
00:54:16
lot about who's hitting the ball harder
00:54:18
because you a it's just the speed off
00:54:20
the racket. You can also hear the
00:54:23
difference. Like that's what I do, Audi.
00:54:24
Sometimes I'll close my eyes and I'll
00:54:27
try to guess which side is hitting the
00:54:28
ball. You can see that. What you can
00:54:31
also see and this is where I can see
00:54:33
that Jookovic I understand now I look we
00:54:36
can get tennis experts at some point
00:54:37
we're going to have Paul Anakone on he
00:54:39
would answer this as well one thing I
00:54:41
know for a fact and the data would
00:54:43
suggest this that Jookovic is doing
00:54:45
worse is what they call getting out of
00:54:47
the corners which means player a
00:54:50
different his opponent hits the ball
00:54:51
deep into the corner historically
00:54:54
Jookovic would stretch hit the ball back
00:54:56
and be back at the tea be back at the
00:54:58
center he can't get out of the corners
00:55:01
now because for him to stretch out far
00:55:03
enough to get to a ball. By the way,
00:55:05
same is true for squash. The the best
00:55:08
players get pinned in the corners but
00:55:11
come out of the corners fast. Jookovic
00:55:14
just at his age now. It's movement. It's
00:55:16
not that he can't hit the ball hard and
00:55:18
it's not conditional on him getting
00:55:20
there. He doesn't hit as good a shot. He
00:55:22
does not get to the ball. He's a step
00:55:24
slower out of the corners and that means
00:55:27
everything in the world of tennis. The
00:55:29
answer is yeah. That's the wonderful
00:55:30
thing about tennis. There's things you
00:55:32
can do that um at the live event, you
00:55:35
can tell who's hitting it harder, you
00:55:36
can tell who's moving better, you can
00:55:38
tell who's able to recover from being in
00:55:40
a defensive position more. All of that
00:55:43
is available and motion tracking data
00:55:46
would be able to answer all of that.
00:55:48
Well guys, it's been kind of an hour
00:55:50
here on Wharton Bunny Ball and the
00:55:51
Wharton podcast network. Uh it's been a
00:55:54
great show today with 15 minutes of open
00:55:56
segments. Um, I don't know how much
00:55:58
baseball we're going to be talking about
00:55:59
now since we know the Reds
00:56:01
>> Oh. Oh, yeah. Phillies knocked out.
00:56:03
You're not going to be paying attention.
00:56:05
I'm going to be talking baseball for the
00:56:06
next couple weeks.
00:56:07
>> All right. Well, we we'll talk some
00:56:09
baseball, but I think we got to talk
00:56:10
some NFL today. We got to talk about the
00:56:12
Patriots. We didn't even get to talk
00:56:14
about the 5-1 Bucks who absolutely beat
00:56:16
who have the best quarterback in the
00:56:17
NFL, Baker Mayfield. Right now, we
00:56:19
didn't get to talk about that, but we
00:56:21
got to talk with Dan Savorski about the
00:56:23
zip system and projections for the
00:56:25
season. We got talked about tennis. So,
00:56:27
lots of great stuff. So, on behalf of
00:56:28
myself, Eric Bradlo, on behalf of my
00:56:30
co-host today, Shane Jensen and Aie
00:56:32
Winer, on behalf of Kate Massie, who
00:56:34
will be back next week. Thanks for
00:56:35
joining us here on the Wharton Podcast
00:56:37
Network here on Wharton Moneyball.
00:56:40
[Music]
